paperID
stringlengths 36
36
| pwc_id
stringlengths 8
47
| arxiv_id
stringlengths 6
16
⌀ | nips_id
float64 | url_abs
stringlengths 18
329
| url_pdf
stringlengths 18
742
| title
stringlengths 8
325
| abstract
stringlengths 1
7.27k
⌀ | authors
stringlengths 2
7.06k
| published
stringlengths 10
10
⌀ | conference
stringlengths 12
47
⌀ | conference_url_abs
stringlengths 16
198
⌀ | conference_url_pdf
stringlengths 27
199
⌀ | proceeding
stringlengths 6
47
⌀ | taskID
stringlengths 7
1.44k
| areaID
stringclasses 688
values | embedding
stringlengths 9.26k
12.5k
| umap_embedding
stringlengths 29
44
|
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
3c2c973e-69f7-4296-9ae5-9e8ae2ad141c
|
novel-algorithm-to-generate-shortest-edit
| null | null |
https://raw.githubusercontent.com/ppml38/shortest_edit_script/main/paper/shortest_edit_script_algorithm.pdf
|
https://raw.githubusercontent.com/ppml38/shortest_edit_script/main/paper/shortest_edit_script_algorithm.pdf
|
Novel algorithm to generate shortest edit script using Levenshtein distance algorithm
|
String similarity, longest common subsequence and shortest edit scripts are the triplets of problem that related to each other. There are different algorithms exist to generate edit script by solving longest common subsequence problem. This paper proposes an algorithm that uses string similarity problem to generate shortest edit script. For this we use the famous Levenshtein distance algorithm, which computes a numerical value that represents similarity between the strings from 0 to n, where n is the length of longest input string, and produce the shortest edit script which contains instructions of Insert, Delete and Substitute.
|
['P. Prakash Maria Liju']
|
2022-08-16
| null | null | null |
github-2022-8
|
['edit-script-generation', 'file-difference']
|
['computer-code', 'computer-code']
|
[ 5.60174346e-01 -3.23576391e-01 2.72349328e-01 -6.87414348e-01
2.18801796e-01 -1.10573804e+00 2.24916711e-01 4.69271123e-01
-4.00046259e-01 6.71965718e-01 3.43566060e-01 -3.44025850e-01
-3.47524256e-01 -1.03787029e+00 -4.05055761e-01 -4.35037538e-02
1.33881986e-01 1.74652100e-01 6.47111475e-01 -6.20885849e-01
1.15787244e+00 4.03159469e-01 -1.64822841e+00 3.38711441e-01
1.02666152e+00 2.11135492e-01 4.29715276e-01 1.03522015e+00
-6.65304482e-01 2.82537878e-01 -6.99666023e-01 -5.48824549e-01
4.12032336e-01 -1.23610735e+00 -8.12616050e-01 -5.54255426e-01
-7.44354948e-02 -9.02067199e-02 -9.01916921e-02 1.13452673e+00
3.45376670e-01 2.69301474e-01 8.23253334e-01 -1.14969242e+00
-9.57146883e-01 7.97656059e-01 -1.82022110e-01 3.70988995e-01
1.17572141e+00 -1.07244149e-01 8.03942025e-01 -4.50540572e-01
8.09351206e-01 9.71131146e-01 9.07319486e-01 1.72868833e-01
-7.57709563e-01 -3.44763219e-01 -6.24919832e-01 2.72392780e-01
-1.36532295e+00 4.47135746e-01 2.07783893e-01 -4.32395875e-01
1.28701508e+00 8.04081619e-01 5.50134778e-01 3.93728048e-01
7.40731120e-01 -8.16334412e-03 9.22738969e-01 -6.44263327e-01
1.65082311e-04 1.29069865e-01 2.76292622e-01 4.80728090e-01
-1.23610655e-02 -5.15508913e-02 -1.13078922e-01 -2.91190594e-01
5.33269286e-01 6.50236234e-02 1.37103051e-02 3.73851597e-01
-1.07178831e+00 8.17285597e-01 -3.57613266e-01 5.13254225e-01
8.25731829e-02 -8.61544684e-02 7.41939783e-01 9.08854663e-01
-4.95412946e-01 4.94706452e-01 -1.96931422e-01 -5.81877649e-01
-8.00125837e-01 4.17097300e-01 1.14068282e+00 1.12403500e+00
7.17493534e-01 -2.61573464e-01 -2.02401027e-01 5.17682195e-01
1.03902161e-01 3.97745162e-01 7.09597766e-01 -7.66372740e-01
2.82914132e-01 7.80221820e-01 -6.27731606e-02 -1.23889303e+00
9.29438472e-02 2.90761590e-01 -3.39533150e-01 2.23632127e-01
-9.88830253e-02 -1.05792344e-01 -3.63366216e-01 1.28391528e+00
3.30342740e-01 1.47063881e-01 -1.01784803e-01 5.20918131e-01
9.95691121e-01 9.47301209e-01 -2.65834838e-01 -2.95541704e-01
9.29940224e-01 -7.37826467e-01 -5.91270804e-01 5.95700085e-01
5.82014501e-01 -1.46421754e+00 1.12428975e+00 4.06224042e-01
-1.01990855e+00 -6.65925026e-01 -1.15409553e+00 -7.02499822e-02
-6.82738841e-01 -2.24852189e-01 1.88289762e-01 8.48796666e-01
-9.25146401e-01 1.24359679e+00 -9.83727723e-03 -5.95572054e-01
-6.71135426e-01 1.78927019e-01 -2.48838827e-01 3.74764770e-01
-1.16009700e+00 1.00581682e+00 9.08153594e-01 -3.72435898e-01
-4.03409868e-01 -5.30030668e-01 -5.24351180e-01 -2.02749580e-01
-1.24904275e-01 -5.41405678e-01 9.37835753e-01 -1.15576613e+00
-1.35585630e+00 8.14322650e-01 -1.22648031e-01 -2.12335855e-01
6.41595781e-01 2.72972494e-01 -5.13797224e-01 -2.77167737e-01
-6.82881027e-02 1.93215951e-01 6.21743441e-01 -4.45982784e-01
-3.25287312e-01 7.01135397e-02 -1.13852844e-01 2.15655535e-01
-9.86237228e-02 5.85947514e-01 -2.56998420e-01 -9.55206096e-01
-2.71560490e-01 -6.68701231e-01 -2.80236900e-01 -1.21365316e-01
-1.75104022e-01 -4.33084965e-01 6.31685674e-01 -9.23116684e-01
1.78063142e+00 -2.19471622e+00 3.51778090e-01 5.87997973e-01
-5.05333722e-01 1.59879729e-01 -2.78442979e-01 1.25571227e+00
-1.82289377e-01 4.98388186e-02 -5.82127333e-01 5.94268680e-01
-1.01938471e-01 3.86082411e-01 -3.00666392e-01 3.24301682e-02
-3.75717700e-01 3.09834599e-01 -9.50896919e-01 -7.38372087e-01
-1.08411156e-01 7.47669637e-02 -2.23992065e-01 2.72289068e-01
-2.43229061e-01 -2.25764960e-02 -2.81367362e-01 2.96110730e-03
5.93659341e-01 6.86071992e-01 5.64040095e-02 -2.59951472e-01
-6.26133263e-01 4.02184948e-02 -1.61236846e+00 1.59842491e+00
-8.77338424e-02 4.96223599e-01 -6.49873495e-01 -9.73106980e-01
1.65642357e+00 7.11868480e-02 2.40838736e-01 -4.48932171e-01
2.35528469e-01 3.44064176e-01 -2.84534305e-01 -1.10799289e+00
7.05254436e-01 -5.40067554e-02 -9.34815332e-02 8.62367630e-01
-2.87366927e-01 -7.32435405e-01 8.95740211e-01 3.12330127e-01
1.10798347e+00 2.81329572e-01 6.13713443e-01 -2.63623983e-01
1.09318888e+00 -7.30322599e-02 4.77777690e-01 8.84739518e-01
1.14542685e-01 4.57659245e-01 7.93078303e-01 -3.62535894e-01
-1.68085372e+00 -1.25198901e+00 3.19936164e-02 8.44626606e-01
9.23488066e-02 -5.66266596e-01 -9.20462370e-01 -3.87569517e-01
1.59433708e-01 8.57725799e-01 -4.06587213e-01 -2.20982403e-01
-8.62742662e-01 1.72065914e-01 8.38055313e-01 2.52581775e-01
3.42224658e-01 -1.20661664e+00 -7.04768717e-01 3.22865874e-01
3.58425856e-01 -5.21538198e-01 -9.95658159e-01 -2.90804774e-01
-8.80707383e-01 -9.85881925e-01 -4.05599684e-01 -1.15572190e+00
6.56758308e-01 -9.87548083e-02 7.44903564e-01 2.79433370e-01
-8.13190281e-01 1.37669936e-01 -8.05287421e-01 -2.47960597e-01
-8.66948426e-01 -3.21696401e-02 -9.90666896e-02 -5.21952569e-01
1.75222978e-01 -8.20260286e-01 -1.77713931e-01 3.13613653e-01
-1.03060484e+00 -1.92774519e-01 3.77686918e-01 2.61167325e-02
3.44814122e-01 -3.19295190e-02 3.75904262e-01 -7.43742585e-01
1.07388628e+00 -5.62023044e-01 -4.08038110e-01 6.11661851e-01
-4.32623982e-01 3.18758637e-01 1.19921327e+00 -1.30428165e-01
-8.11061442e-01 8.70586783e-02 -1.33610725e-01 3.83381575e-01
6.20875619e-02 3.95423204e-01 2.95388520e-01 -2.55607814e-01
7.27184534e-01 8.42284083e-01 9.57175270e-02 -5.53832650e-01
4.11386102e-01 1.06010413e+00 7.20344961e-01 -4.15427029e-01
6.50348127e-01 7.26087391e-02 1.11522591e-02 -5.86063683e-01
-1.93841442e-01 -1.73576698e-01 -6.20884776e-01 -3.51063073e-01
7.69276679e-01 -2.15141494e-02 -6.70612216e-01 1.93504646e-01
-1.46554863e+00 2.23116934e-01 -1.19894832e-01 3.14171255e-01
-6.57910526e-01 9.43165064e-01 -2.87665069e-01 -2.46570140e-01
-6.95457518e-01 -8.04764628e-01 3.06125015e-01 5.03134906e-01
-7.77362287e-01 -6.83775961e-01 5.23099840e-01 -3.09195787e-01
2.48344719e-01 5.59510589e-01 1.07951713e+00 -5.98890007e-01
-3.04259330e-01 -1.66622803e-01 1.66301191e-01 4.31476086e-01
2.65634298e-01 7.83094347e-01 3.00254196e-01 1.22208573e-01
-1.08177118e-01 3.52820307e-01 2.51908034e-01 -3.91059250e-01
1.27162600e+00 -3.22345346e-01 -2.57884026e-01 5.67212760e-01
1.56969798e+00 7.94630885e-01 8.39771628e-01 3.33682507e-01
1.40808567e-01 2.98724174e-01 8.22269320e-01 7.75366127e-01
-2.88106725e-02 3.96852970e-01 -1.09149165e-01 6.72966421e-01
1.81442440e-01 -2.96179414e-01 4.28262502e-01 1.19357443e+00
2.26003062e-02 1.02294207e-01 -7.88221121e-01 4.55404073e-01
-1.51702857e+00 -1.26596510e+00 -6.93356156e-01 2.27180505e+00
1.13115776e+00 1.30573004e-01 1.92923397e-01 3.98151100e-01
9.02380168e-01 -5.51886186e-02 -2.80021638e-01 -1.68575716e+00
2.86368467e-02 5.91275036e-01 4.90833968e-01 6.07195556e-01
-4.96838868e-01 8.03463519e-01 6.77928543e+00 7.60219395e-01
-7.74625599e-01 -9.10921991e-02 -3.95034164e-01 -2.91673969e-02
-7.74908960e-01 1.85679600e-01 -6.58458471e-01 9.77819443e-01
8.47335517e-01 -1.09919941e+00 6.76418960e-01 5.97996771e-01
3.27253699e-01 -2.79224217e-01 -9.73853648e-01 7.61474907e-01
4.04601991e-01 -1.23441780e+00 4.62819844e-01 -9.08073366e-01
1.17856836e+00 -5.09812653e-01 -3.28909218e-01 -2.41937727e-01
3.88183236e-01 -8.21459234e-01 5.29339433e-01 8.87314558e-01
4.09019887e-01 -9.29819882e-01 4.15410906e-01 1.24638990e-01
-1.48231304e+00 1.70975700e-02 -7.04692662e-01 1.52551522e-02
3.57431501e-01 6.12721682e-01 -4.75346774e-01 6.16088033e-01
4.92192805e-01 7.13087499e-01 -4.10631269e-01 1.52723992e+00
-1.83996797e-01 -7.77529627e-02 -1.73715547e-01 -6.46398604e-01
2.34972894e-01 -8.60195816e-01 5.59724569e-01 1.53615320e+00
8.20900023e-01 3.08099031e-01 1.48520491e-03 7.06679463e-01
1.94863841e-01 6.36686862e-01 -9.17787611e-01 -1.79923385e-01
8.77043426e-01 6.64579093e-01 -5.68521023e-01 -4.69244599e-01
-2.68301368e-01 1.42366135e+00 -1.83840469e-01 -1.37087211e-01
-1.25276279e+00 -1.68416381e+00 3.99871290e-01 -2.87090540e-01
2.27868602e-01 -5.46550930e-01 -4.33954149e-01 -1.54646590e-01
1.72734857e-01 -1.03025341e+00 4.09162164e-01 -7.49415934e-01
-7.12798655e-01 3.54926139e-01 -8.80659968e-02 -1.15714192e+00
1.49203733e-01 -2.00889811e-01 -1.31939626e+00 9.75527644e-01
-6.10674143e-01 -4.33270037e-01 -4.28641856e-01 6.11732364e-01
7.34061003e-01 -8.44387040e-02 6.90695226e-01 3.01041394e-01
-1.42240986e-01 5.35402179e-01 5.78609049e-01 -7.73157254e-02
7.36193120e-01 -7.91340470e-01 6.13596201e-01 4.29372072e-01
-3.22909772e-01 7.84995854e-01 1.09325290e+00 -1.05218577e+00
-1.29641426e+00 -6.38384461e-01 1.35991192e+00 -1.38843834e-01
5.13745606e-01 2.38587245e-01 -7.27950454e-01 6.91782355e-01
4.16162133e-01 -8.36308599e-01 8.07326019e-01 -8.54932487e-01
-3.54152083e-01 -1.35027185e-01 -1.29831195e+00 5.86793900e-01
1.20508313e+00 -5.00140429e-01 -1.03617239e+00 3.00029039e-01
6.68675721e-01 -3.46865773e-01 -1.12637544e+00 -2.97336672e-02
5.78530073e-01 -1.22752237e+00 6.76604807e-01 -6.76533580e-01
6.88115299e-01 -6.48717642e-01 -7.48252645e-02 -1.09366691e+00
-3.53094637e-02 -8.14320028e-01 8.36107314e-01 1.32269621e+00
3.17323476e-01 -4.71466810e-01 3.83430213e-01 8.08307528e-02
-1.58966854e-01 -6.16979897e-01 -5.51552176e-01 -1.24779665e+00
-2.42212098e-02 -6.28906768e-03 1.09382641e+00 9.95915473e-01
3.64637703e-01 -2.17182845e-01 -2.37867534e-01 -3.61559063e-01
3.40756446e-01 1.77873626e-01 5.06191790e-01 -1.07717907e+00
-2.66082823e-01 -3.98209780e-01 -6.63661301e-01 -4.50785995e-01
1.16143525e-01 -1.20449293e+00 -1.10628709e-01 -1.55046797e+00
1.69618264e-01 -2.04465270e-01 2.44598210e-01 -7.80151263e-02
8.66433084e-02 -2.19545722e-01 3.34340751e-01 -1.25958517e-01
-6.91212416e-02 3.04528236e-01 8.75206590e-01 3.21510597e-03
-3.27046841e-01 -3.64153057e-01 -1.03605367e-01 4.99396592e-01
9.54440832e-01 -7.48134553e-01 -5.00220239e-01 -6.32571042e-01
6.65082037e-01 -1.53247109e-02 -2.46544808e-01 -9.57426906e-01
3.08997452e-01 -7.08349407e-01 -1.89589545e-01 -4.96749729e-01
-4.31029469e-01 -7.32358634e-01 8.76170397e-01 8.54871869e-01
-6.58976674e-01 8.89951289e-01 -1.82716474e-01 1.24921009e-01
-7.61608928e-02 -1.41567647e+00 7.68238783e-01 -4.43705857e-01
-6.94411278e-01 -2.34796733e-01 -5.95022500e-01 -5.85551634e-02
1.59493017e+00 -7.41921186e-01 -2.99799815e-02 1.20763863e-02
-4.66538787e-01 -4.89389244e-03 5.61598420e-01 5.27203202e-01
6.83286667e-01 -1.28725874e+00 -6.67269468e-01 1.01123936e-01
-2.61211306e-01 -8.19209993e-01 1.62617525e-03 4.07685876e-01
-1.42590153e+00 2.35762745e-01 -7.35144973e-01 1.24795362e-01
-1.76973879e+00 4.01037067e-01 2.39660382e-01 3.24411601e-01
-4.13575053e-01 9.31171000e-01 -9.42267776e-01 -5.95753551e-01
1.02077171e-01 -1.21207073e-01 -2.43752941e-01 -5.83544597e-02
7.16204643e-01 9.91919994e-01 -2.40382105e-01 -2.88144141e-01
-2.64944732e-01 1.32719409e+00 4.77857515e-02 -2.63512153e-02
8.90141368e-01 1.39732227e-01 -9.80969667e-01 3.24619472e-01
1.58104229e+00 -1.68345273e-01 -2.78471798e-01 2.64519274e-01
1.22883737e-01 -8.76808286e-01 -8.61439347e-01 -3.88702035e-01
-4.24783796e-01 4.55480069e-01 5.69629312e-01 1.79108545e-01
8.38184655e-01 -3.42397034e-01 1.19500625e+00 4.95928109e-01
2.40344122e-01 -1.51155686e+00 -2.94637103e-02 8.50778520e-01
9.96087849e-01 -5.20595074e-01 1.57112051e-02 -3.56324881e-01
-4.85932678e-01 1.66503763e+00 6.03052616e-01 -4.05397266e-01
3.25562567e-01 3.37243766e-01 -2.94223070e-01 2.93269306e-02
-2.90829211e-01 2.59443402e-01 -1.61058754e-01 5.29240966e-01
6.26616418e-01 -1.56514928e-01 -1.72281218e+00 2.21932828e-01
-8.40841055e-01 4.10382122e-01 8.14996719e-01 1.19678473e+00
-9.91550744e-01 -1.49832797e+00 -5.35627365e-01 5.49936056e-01
2.57433392e-02 -1.82562377e-02 -7.85530150e-01 1.46669984e-01
4.14838940e-01 6.86397731e-01 2.17696264e-01 -4.77400005e-01
5.66698670e-01 1.00719057e-01 5.63034773e-01 -3.34168822e-01
-1.18909740e+00 -1.14044344e+00 -8.30190107e-02 -4.20193225e-01
1.45993501e-01 -6.87144816e-01 -1.66431010e+00 -8.63990963e-01
7.71566629e-02 4.02841926e-01 8.02988112e-01 7.74164855e-01
1.20252922e-01 1.23921096e-01 1.03333294e+00 2.63268426e-02
-8.67752612e-01 -4.16935116e-01 -4.18246150e-01 5.73912323e-01
-2.80719399e-01 3.10816038e-02 -2.24172637e-01 2.55991757e-01]
|
[4.989056587219238, 5.224582672119141]
|
4c5ed98e-283c-4e72-b2a8-705724881ab7
|
graphflow-exploiting-conversation-flow-with
|
1908.00059
| null |
https://arxiv.org/abs/1908.00059v2
|
https://arxiv.org/pdf/1908.00059v2.pdf
|
GraphFlow: Exploiting Conversation Flow with Graph Neural Networks for Conversational Machine Comprehension
|
Conversational machine comprehension (MC) has proven significantly more challenging compared to traditional MC since it requires better utilization of conversation history. However, most existing approaches do not effectively capture conversation history and thus have trouble handling questions involving coreference or ellipsis. Moreover, when reasoning over passage text, most of them simply treat it as a word sequence without exploring rich semantic relationships among words. In this paper, we first propose a simple yet effective graph structure learning technique to dynamically construct a question and conversation history aware context graph at each conversation turn. Then we propose a novel Recurrent Graph Neural Network, and based on that, we introduce a flow mechanism to model the temporal dependencies in a sequence of context graphs. The proposed GraphFlow model can effectively capture conversational flow in a dialog, and shows competitive performance compared to existing state-of-the-art methods on CoQA, QuAC and DoQA benchmarks. In addition, visualization experiments show that our proposed model can offer good interpretability for the reasoning process.
|
['Lingfei Wu', 'Yu Chen', 'Mohammed J. Zaki']
|
2019-07-31
| null | null | null | null |
['graph-structure-learning']
|
['graphs']
|
[ 2.36651316e-01 2.17717260e-01 -6.99649528e-02 -4.55703050e-01
-3.53532851e-01 -6.65849805e-01 6.64370000e-01 5.87515175e-01
-1.60772517e-01 5.82752883e-01 8.47952306e-01 -7.96598494e-01
-1.50264502e-01 -8.96477401e-01 -1.78962857e-01 -1.36070520e-01
7.99120441e-02 6.49815679e-01 4.51819509e-01 -7.45021522e-01
4.90127116e-01 -3.14625502e-02 -1.04287100e+00 5.88924646e-01
1.14986527e+00 5.76072335e-01 1.40125364e-01 9.33135092e-01
-9.95022833e-01 1.57240558e+00 -9.67384517e-01 -5.56348979e-01
-4.44093376e-01 -1.08254099e+00 -1.54761982e+00 -3.79849561e-02
1.30660057e-01 -3.11936975e-01 -5.56697428e-01 8.01158071e-01
2.70271271e-01 5.11809230e-01 4.40246195e-01 -1.26921916e+00
-4.16112125e-01 1.17057037e+00 -2.18184724e-01 5.78840971e-01
1.02132416e+00 -4.03193012e-02 1.34931600e+00 -1.88756004e-01
6.63353026e-01 1.61233866e+00 4.07432735e-01 5.48230886e-01
-6.47457659e-01 -3.65757942e-01 6.95195377e-01 8.79650354e-01
-6.64777160e-01 -7.71739781e-02 1.11566138e+00 -9.52539891e-02
1.14480221e+00 3.75110030e-01 8.81603718e-01 9.65463817e-01
2.48572171e-01 9.66144681e-01 7.52202332e-01 -4.21894372e-01
1.69749841e-01 -4.91152257e-01 8.02959681e-01 7.92366385e-01
-3.66142601e-01 -7.20447004e-01 -5.03915787e-01 -2.54268736e-01
2.59554178e-01 1.68209281e-02 -5.58574557e-01 -9.66764838e-02
-8.60471666e-01 9.99811172e-01 5.06041944e-01 4.91174430e-01
-1.53022304e-01 -7.71761090e-02 6.23944759e-01 4.93458599e-01
9.06322226e-02 3.25193197e-01 -1.08686954e-01 -3.71279031e-01
-2.95036972e-01 2.74543226e-01 1.41830242e+00 8.51551354e-01
5.29184282e-01 -3.05171013e-01 -3.53974193e-01 8.41144145e-01
2.21559256e-01 4.24393713e-02 4.20013845e-01 -1.01644242e+00
7.43624389e-01 1.11849570e+00 -1.94300979e-01 -1.32384586e+00
-7.23241568e-01 -1.09658735e-02 -7.38063514e-01 -4.60363746e-01
4.91500676e-01 -2.20934436e-01 -3.96975875e-01 1.60291433e+00
3.56029689e-01 1.71920612e-01 3.10306102e-01 8.48339558e-01
1.21094465e+00 9.39626694e-01 4.31994200e-02 -6.33154213e-01
1.51696181e+00 -1.35827565e+00 -1.30317199e+00 -3.23241502e-01
4.84844089e-01 -5.53133845e-01 1.18794298e+00 2.91416377e-01
-9.28224921e-01 -2.54283518e-01 -9.84110951e-01 -4.11523759e-01
-7.95593336e-02 -6.21572852e-01 9.38428223e-01 2.67335832e-01
-7.63314784e-01 3.40239823e-01 -5.02750933e-01 -4.07132626e-01
7.75841549e-02 6.10060059e-02 1.21348009e-01 -1.41189203e-01
-1.48031271e+00 8.08496356e-01 3.07638735e-01 2.71847367e-01
-5.37373245e-01 -3.05945218e-01 -1.00913644e+00 3.35924149e-01
9.12213206e-01 -6.85225606e-01 1.69366109e+00 -7.49395549e-01
-1.84310865e+00 2.97577143e-01 -5.23625135e-01 -4.84909356e-01
2.69485384e-01 -2.14807004e-01 -4.65129524e-01 3.78399372e-01
-2.87295818e-01 1.38251379e-01 4.92909104e-01 -1.02920210e+00
-5.72045565e-01 -2.32813656e-01 7.50573754e-01 5.92883050e-01
-2.18233407e-01 -5.55984266e-02 -5.22417665e-01 -2.44765013e-01
2.19509490e-02 -6.18709505e-01 -1.93045780e-01 -4.10887688e-01
-4.48487610e-01 -8.25596690e-01 7.98346698e-01 -7.78175831e-01
1.83358109e+00 -1.68084383e+00 4.47044969e-01 -1.17963821e-01
5.89134574e-01 3.79986286e-01 -8.02573860e-02 9.48600888e-01
3.28533888e-01 3.08204591e-02 -4.13433552e-01 -4.46072966e-02
-3.14442590e-02 4.54325497e-01 -2.72096246e-01 -6.96226358e-02
-1.64650798e-01 1.02290165e+00 -1.07956934e+00 -6.59888148e-01
1.14724040e-01 1.37000531e-01 -5.49200535e-01 5.71834326e-01
-7.82080650e-01 3.65006626e-01 -6.16791248e-01 1.64735258e-01
4.67621088e-01 -5.85164607e-01 9.11408305e-01 -1.56706963e-02
2.91875988e-01 6.23727620e-01 -7.53078043e-01 1.98132145e+00
-6.13908172e-01 6.79810643e-01 3.88845243e-02 -1.05448008e+00
8.22173834e-01 1.85752377e-01 -1.02771744e-02 -6.98437750e-01
3.36548299e-01 -3.40986341e-01 3.58084947e-01 -7.97913492e-01
5.03082931e-01 7.53356889e-02 -7.39943162e-02 8.10098648e-01
-5.23221539e-03 -2.69887950e-02 3.90919119e-01 6.85560763e-01
1.20246613e+00 -3.10602635e-01 4.53164786e-01 8.66888836e-02
1.14117849e+00 1.23266801e-01 4.44059521e-01 5.58663666e-01
-4.25043225e-01 1.11896746e-01 1.15720093e+00 -4.18173254e-01
-4.15828913e-01 -7.52081573e-01 5.80142200e-01 1.22475934e+00
4.62023497e-01 -8.35778594e-01 -9.60564315e-01 -9.56879199e-01
-3.18318456e-01 9.58606005e-01 -3.65913510e-01 -1.07078627e-01
-1.09414077e+00 -3.04841965e-01 3.94268572e-01 3.23659152e-01
8.60886157e-01 -1.36647236e+00 -2.53803879e-01 4.05202419e-01
-8.05734813e-01 -1.26955807e+00 -6.51429117e-01 -4.59258646e-01
-8.12170863e-01 -1.46640515e+00 -6.00043945e-02 -9.25157547e-01
1.86107978e-01 5.43853998e-01 1.35018325e+00 6.37501061e-01
-4.49250266e-02 4.94164228e-01 -7.40486264e-01 -1.19874030e-01
-6.91581368e-01 3.07264358e-01 -5.79821587e-01 -1.03765421e-01
5.69289029e-01 -7.11599648e-01 -6.87337220e-01 1.78362250e-01
-7.85391867e-01 7.43667856e-02 9.68676433e-02 8.32378507e-01
-4.71655093e-02 -1.55051172e-01 7.58910298e-01 -1.21492064e+00
1.40670431e+00 -5.17038465e-01 -3.01350206e-01 7.13367701e-01
-5.79867363e-01 1.40151113e-01 8.34170640e-01 -2.02654421e-01
-1.52439809e+00 -5.17611623e-01 -2.25041986e-01 4.54494320e-02
2.26678755e-02 7.67199159e-01 -6.76084384e-02 3.68379354e-01
5.12150943e-01 7.66075179e-02 1.64639056e-01 -2.46719286e-01
5.84002435e-01 6.79351032e-01 5.09215534e-01 -5.62037170e-01
3.02855492e-01 2.15672806e-01 -2.39573210e-01 -8.32165182e-01
-1.01288843e+00 -6.94061816e-01 -5.07440269e-01 -2.94955701e-01
9.96715963e-01 -4.55113798e-01 -1.34330702e+00 3.00461471e-01
-1.46800244e+00 -3.28670532e-01 2.46967807e-01 1.10346846e-01
-4.00603414e-01 9.16467190e-01 -8.95913601e-01 -9.73051786e-01
-6.29114926e-01 -7.65488029e-01 4.87393767e-01 5.02514839e-01
-4.11019742e-01 -1.36746883e+00 1.07459530e-01 8.76906812e-01
3.64317864e-01 1.81057304e-01 1.25832069e+00 -8.53559732e-01
-4.64371473e-01 2.33695686e-01 -1.15362592e-01 -6.07639328e-02
3.08794588e-01 -1.60479411e-01 -6.10995173e-01 -1.05580933e-01
1.02004237e-01 -2.64211357e-01 6.06436133e-01 -6.50976598e-02
9.52079356e-01 -5.69923878e-01 -2.03379080e-01 1.42327517e-01
8.87452781e-01 4.26626742e-01 4.55351293e-01 -1.64848398e-02
6.70613825e-01 8.67287636e-01 3.85669112e-01 2.34131277e-01
8.44606340e-01 3.47545117e-01 3.04825693e-01 3.13830733e-01
-7.84695819e-02 -4.61675018e-01 2.00716138e-01 1.50951540e+00
1.34083837e-01 -7.19714701e-01 -1.01689804e+00 4.29736614e-01
-2.24480891e+00 -9.48174298e-01 -4.39522743e-01 1.42113996e+00
6.82308137e-01 1.79594889e-01 1.29249230e-01 3.45689543e-02
7.18434036e-01 4.97350901e-01 -4.84295070e-01 -7.02342927e-01
1.37446504e-02 1.53609186e-01 -2.96338618e-01 1.00246894e+00
-6.01579785e-01 1.06251192e+00 6.37492466e+00 4.72013801e-01
-6.56963646e-01 7.83586595e-03 2.27960944e-01 2.05965310e-01
-4.89506960e-01 1.05673328e-01 -2.83759892e-01 8.87105465e-02
8.39380980e-01 -4.97680962e-01 5.88804841e-01 5.67876637e-01
-1.98116284e-02 -1.39165238e-01 -9.79553819e-01 8.64570856e-01
3.14732105e-01 -1.48350501e+00 2.22591341e-01 -5.31564116e-01
2.52218753e-01 -4.91031319e-01 -6.46430552e-01 5.73645890e-01
4.27303106e-01 -1.02324498e+00 -1.35711553e-02 3.90935630e-01
6.01425916e-02 -7.78029203e-01 8.41081738e-01 5.60505033e-01
-1.23737907e+00 -2.62334466e-01 -2.84845561e-01 -4.51993495e-01
3.90910566e-01 2.64240615e-02 -1.01771259e+00 9.85421419e-01
3.28035355e-01 5.68248391e-01 -4.71799314e-01 6.50679231e-01
-6.71072602e-01 8.24003160e-01 1.23150073e-01 -7.00928330e-01
3.12018514e-01 -3.45873624e-01 5.20196080e-01 1.20959401e+00
-1.83997080e-01 6.85634613e-01 3.45796466e-01 4.93128926e-01
-2.20174015e-01 3.17835540e-01 -4.11498278e-01 -1.68333516e-01
5.03987849e-01 1.12259638e+00 -5.87435424e-01 -3.85722458e-01
-4.91484791e-01 1.02076125e+00 6.17617428e-01 3.80100727e-01
-6.53428495e-01 -3.68759006e-01 3.94865960e-01 -2.90389150e-01
4.38974202e-02 -2.76264846e-01 8.39954987e-02 -1.27943718e+00
7.45520517e-02 -1.08520210e+00 1.01863551e+00 -7.00305998e-01
-1.33232212e+00 8.12283874e-01 2.62303132e-04 -7.17742860e-01
-5.46120763e-01 -3.73758942e-01 -1.07999146e+00 6.92796528e-01
-1.41628408e+00 -8.95541191e-01 -6.46391928e-01 7.78389335e-01
9.52063739e-01 -8.82029161e-03 8.05370867e-01 -9.56542343e-02
-5.11696339e-01 3.17594886e-01 -4.43463236e-01 3.27579141e-01
4.12639529e-01 -1.34213865e+00 4.00207341e-01 6.13327920e-01
1.95313528e-01 7.87776113e-01 6.30702078e-01 -5.91283798e-01
-1.58808196e+00 -5.17160594e-01 9.46292102e-01 -3.24941307e-01
8.26262653e-01 -3.67130220e-01 -1.35412443e+00 7.23196089e-01
9.30897832e-01 -6.74966335e-01 6.93784475e-01 4.70932335e-01
-2.86103219e-01 -4.46953345e-03 -6.87778831e-01 7.52504468e-01
1.20511854e+00 -6.94419920e-01 -1.36384511e+00 4.98180062e-01
1.24070537e+00 -5.28644562e-01 -5.81847131e-01 7.46187866e-02
2.11914107e-01 -1.03692210e+00 6.08121276e-01 -8.85364115e-01
4.73161906e-01 -9.40953270e-02 1.99759871e-01 -1.25989723e+00
-1.23118959e-01 -1.10251594e+00 -4.08028036e-01 1.22855926e+00
3.39032382e-01 -4.86997962e-01 7.16860175e-01 2.66443700e-01
-1.93290591e-01 -5.27139187e-01 -7.18652666e-01 -3.45828921e-01
2.87707597e-02 -3.05008411e-01 6.01853669e-01 1.03797567e+00
7.28018463e-01 1.15896678e+00 -2.51382262e-01 -8.50439668e-02
2.28502333e-01 5.03916860e-01 8.40323508e-01 -1.14003813e+00
-2.86550641e-01 -5.81148386e-01 2.62565501e-02 -1.47308719e+00
4.44710940e-01 -8.31708789e-01 6.24006651e-02 -2.10754871e+00
6.87630987e-03 1.37567490e-01 1.69047732e-02 2.57747807e-02
-5.64769447e-01 -6.56690538e-01 2.25019082e-01 9.17047411e-02
-9.15069401e-01 6.67074382e-01 1.61185074e+00 -3.26830238e-01
-3.73874754e-01 -9.07198414e-02 -6.21848702e-01 6.47584736e-01
7.51453102e-01 -3.72209959e-02 -9.00596201e-01 -2.94957191e-01
2.50211120e-01 7.18021512e-01 -2.64765713e-02 -5.44026017e-01
7.03480899e-01 -3.67047280e-01 -3.74365211e-01 -6.34031892e-01
1.79125100e-01 -4.66104954e-01 -2.91322023e-01 3.97604883e-01
-5.46866059e-01 2.58758724e-01 1.03246525e-01 8.27562928e-01
-5.35320520e-01 -2.02296585e-01 2.44635865e-01 -1.78170338e-01
-7.87123859e-01 5.46328425e-02 -5.96571326e-01 3.74231249e-01
7.69154549e-01 3.21594328e-01 -6.28520429e-01 -1.07553637e+00
-6.47588909e-01 8.81271720e-01 -1.14923410e-01 6.76515162e-01
6.25889063e-01 -9.76347208e-01 -5.44999242e-01 -4.43052232e-01
5.09583391e-02 1.96693838e-01 5.59913278e-01 4.99251097e-01
-7.82837749e-01 3.92189950e-01 -4.46373485e-02 -5.43083012e-01
-1.43338966e+00 6.87344670e-01 3.39924365e-01 -6.12134516e-01
-7.74038732e-01 6.15794122e-01 3.31859365e-02 -5.19902527e-01
4.12404537e-01 -4.65682000e-01 -8.54608893e-01 1.39536232e-01
7.25903094e-01 2.54805773e-01 -1.41754210e-01 -2.02548578e-01
-2.13525057e-01 3.58042330e-01 -1.73827648e-01 -7.19667226e-02
9.60682452e-01 -3.56234431e-01 -4.49417412e-01 6.48801267e-01
8.78433526e-01 -2.09802706e-02 -8.45789909e-01 -4.17456865e-01
3.26558232e-01 -1.09739341e-01 -4.90022719e-01 -7.98866212e-01
-6.55176997e-01 1.02022266e+00 -1.11557297e-01 7.74562418e-01
1.05963230e+00 1.33025423e-02 1.17787147e+00 8.18391860e-01
6.47987425e-02 -8.84005666e-01 4.07445043e-01 9.32156980e-01
1.02817166e+00 -1.03754163e+00 -5.56083210e-02 -8.54285240e-01
-8.25062335e-01 1.45625937e+00 8.13971460e-01 8.31322297e-02
3.29835057e-01 -2.09853694e-01 3.28949988e-01 -4.78636295e-01
-1.05942333e+00 -7.11063668e-02 8.80682468e-02 3.91544104e-01
4.49198931e-01 -1.52899191e-01 -6.40071273e-01 5.03888726e-01
-3.74562204e-01 -2.39753038e-01 7.52858341e-01 9.27550197e-01
-5.02515197e-01 -1.18189263e+00 -8.73494707e-03 1.14590846e-01
-1.50080651e-01 -6.64607212e-02 -8.66147757e-01 8.30686569e-01
-5.57586372e-01 1.58235526e+00 6.05164953e-02 -2.96035618e-01
5.22209823e-01 3.87170434e-01 3.36133689e-01 -4.99271452e-01
-9.51042235e-01 -2.58433074e-01 6.38299227e-01 -5.45014441e-01
-6.01812184e-01 -3.44914913e-01 -1.61724699e+00 -5.54628134e-01
-2.84512520e-01 6.40362859e-01 2.10834831e-01 1.21306252e+00
2.31657416e-01 8.19748282e-01 5.02575159e-01 5.44122048e-02
-3.97573441e-01 -1.19664097e+00 -4.98374738e-02 6.39333189e-01
2.72135258e-01 -3.70094329e-01 -3.39810789e-01 -1.92750320e-01]
|
[12.156363487243652, 7.919756889343262]
|
cf802f9a-9d2c-4a5a-97a6-a167de3fe0f3
|
coot-cooperative-hierarchical-transformer-for
|
2011.00597
| null |
https://arxiv.org/abs/2011.00597v1
|
https://arxiv.org/pdf/2011.00597v1.pdf
|
COOT: Cooperative Hierarchical Transformer for Video-Text Representation Learning
|
Many real-world video-text tasks involve different levels of granularity, such as frames and words, clip and sentences or videos and paragraphs, each with distinct semantics. In this paper, we propose a Cooperative hierarchical Transformer (COOT) to leverage this hierarchy information and model the interactions between different levels of granularity and different modalities. The method consists of three major components: an attention-aware feature aggregation layer, which leverages the local temporal context (intra-level, e.g., within a clip), a contextual transformer to learn the interactions between low-level and high-level semantics (inter-level, e.g. clip-video, sentence-paragraph), and a cross-modal cycle-consistency loss to connect video and text. The resulting method compares favorably to the state of the art on several benchmarks while having few parameters. All code is available open-source at https://github.com/gingsi/coot-videotext
|
['Thomas Brox', 'Hamed Pirsiavash', 'Mohammadreza Zolfaghari', 'Simon Ging']
|
2020-11-01
| null |
http://proceedings.neurips.cc/paper/2020/hash/ff0abbcc0227c9124a804b084d161a2d-Abstract.html
|
http://proceedings.neurips.cc/paper/2020/file/ff0abbcc0227c9124a804b084d161a2d-Paper.pdf
|
neurips-2020-12
|
['video-text-retrieval']
|
['computer-vision']
|
[ 2.44156737e-02 -3.62423718e-01 -3.43184739e-01 -3.74444962e-01
-8.98996115e-01 -6.25747263e-01 9.06510592e-01 4.14169461e-01
-2.48158917e-01 4.20430303e-01 6.97922349e-01 2.50812888e-01
7.46884048e-02 -4.00692910e-01 -6.83158100e-01 -6.17156029e-01
-8.35518390e-02 1.02363206e-01 7.45043814e-01 1.52216658e-01
2.20978171e-01 -1.52988480e-02 -1.68986928e+00 9.95940804e-01
2.84769416e-01 1.12891483e+00 3.11061829e-01 8.29275966e-01
-1.68721050e-01 1.09214079e+00 -1.10330239e-01 -2.14293942e-01
-7.13872015e-02 -5.75700164e-01 -8.19996476e-01 2.87998259e-01
5.17480791e-01 -1.89946800e-01 -4.68124300e-01 8.23343992e-01
3.28038901e-01 2.47554570e-01 2.13660315e-01 -1.24480438e+00
-2.22020298e-01 6.38925493e-01 -7.98827946e-01 4.44444418e-01
4.75395709e-01 1.50854692e-01 1.27072942e+00 -1.04763961e+00
8.14491212e-01 1.42366397e+00 5.74217856e-01 1.38506085e-01
-1.30828059e+00 -3.94391328e-01 5.17056882e-01 4.71584678e-01
-1.21795309e+00 -5.70186257e-01 7.40635276e-01 -6.73147738e-01
9.27937388e-01 1.08465880e-01 7.08503425e-01 1.20701385e+00
3.85008931e-01 1.04524815e+00 9.96496737e-01 -2.24966273e-01
8.42443779e-02 -2.22689018e-01 9.30289626e-02 7.22079456e-01
-2.44966999e-01 -2.67093986e-01 -1.15903294e+00 -7.39534870e-02
4.86032784e-01 2.01636732e-01 -2.40422860e-01 -3.58080506e-01
-1.40683758e+00 6.15079343e-01 1.88718051e-01 6.01553798e-01
-2.66177505e-01 3.36081535e-01 7.27909386e-01 1.93019375e-01
5.19751608e-01 -1.54629186e-01 -5.30214965e-01 -4.60756719e-01
-1.24409091e+00 3.02442461e-02 6.95867300e-01 1.02410507e+00
7.63655066e-01 -3.18397969e-01 -5.68737566e-01 8.06815803e-01
3.00134927e-01 1.56422526e-01 3.54569703e-01 -9.50792372e-01
8.17156196e-01 3.69232774e-01 -2.21089542e-01 -9.33023214e-01
-3.68107885e-01 -1.65083557e-02 -6.32266521e-01 -2.79049993e-01
8.35473090e-02 5.76562919e-02 -9.02284563e-01 1.81266260e+00
4.08969045e-01 5.43215632e-01 -4.84840095e-01 8.57241452e-01
7.76828229e-01 9.58298802e-01 1.77363694e-01 -3.74300718e-01
1.67996478e+00 -1.37636077e+00 -8.58795106e-01 -3.91777605e-02
3.76823574e-01 -8.52453768e-01 1.20766640e+00 1.50516540e-01
-1.50785613e+00 -4.70337778e-01 -8.35119605e-01 -5.28953016e-01
-4.34636235e-01 6.99315369e-02 6.84728697e-02 -1.40385628e-01
-1.02060437e+00 3.42491448e-01 -1.12868643e+00 -5.42834997e-01
1.83956102e-01 1.27363741e-01 -3.55966866e-01 -1.72448575e-01
-1.29883480e+00 3.16664159e-01 3.14629912e-01 -2.51820713e-01
-9.08208132e-01 -8.03031981e-01 -8.86899173e-01 2.89779305e-01
6.81639254e-01 -8.06964934e-01 1.22947919e+00 -1.16048658e+00
-1.32699323e+00 8.60879362e-01 -3.89568746e-01 -2.70191312e-01
4.58042532e-01 -2.97887832e-01 -2.63752133e-01 5.82824886e-01
1.65624946e-01 7.45373011e-01 1.10215306e+00 -1.05439878e+00
-9.16810691e-01 -4.30764079e-01 2.68990964e-01 4.41882670e-01
-3.09640825e-01 2.45627627e-01 -1.25197268e+00 -9.10305619e-01
-1.92539170e-01 -9.06053305e-01 3.71701211e-01 3.93236279e-02
-3.55716258e-01 -1.42751426e-01 1.03510654e+00 -5.95519900e-01
1.60024488e+00 -2.43748903e+00 5.99336565e-01 -2.87774473e-01
3.03661644e-01 -2.28582710e-01 -5.89518175e-02 8.30606818e-01
8.23284164e-02 6.69938251e-02 -6.43912926e-02 -7.72540450e-01
4.41609770e-02 -1.84513256e-02 2.38672830e-02 3.85158509e-01
-6.38648570e-02 8.49954545e-01 -8.87347400e-01 -7.34348357e-01
3.42211276e-01 7.26013780e-01 -5.16369045e-01 3.96860018e-02
-4.39721018e-01 3.82049859e-01 -5.04408658e-01 5.92448592e-01
2.61592299e-01 -6.10685527e-01 1.58378884e-01 -5.10188282e-01
-2.85379648e-01 2.92711198e-01 -1.01881588e+00 1.97956753e+00
-5.35258114e-01 9.70509768e-01 3.11132342e-01 -8.16011965e-01
2.18774989e-01 5.47013521e-01 7.12615967e-01 -4.64495808e-01
2.08885193e-01 -1.73700288e-01 -4.92407769e-01 -6.01390481e-01
5.90316296e-01 2.51291275e-01 -2.05827847e-01 2.40982935e-01
3.54727298e-01 7.53516033e-02 5.17601907e-01 5.64446867e-01
1.15152574e+00 2.23810047e-01 2.49776646e-01 -2.96977490e-01
5.42856812e-01 -3.18810493e-01 5.06984234e-01 4.67036426e-01
-1.59709707e-01 7.53317058e-01 8.67958784e-01 -1.79591566e-01
-1.05162370e+00 -7.81569600e-01 1.02714539e-01 1.38466823e+00
4.41446453e-01 -1.00606751e+00 -6.39232397e-01 -3.84329677e-01
-3.14857578e-03 2.80332267e-01 -7.85733461e-01 8.62402900e-04
-4.34799284e-01 -1.24285214e-01 6.90530613e-02 5.70725620e-01
4.29850072e-01 -8.82884800e-01 -5.46647727e-01 8.52694139e-02
-5.09659111e-01 -1.56867218e+00 -1.10865140e+00 -1.13875791e-02
-6.98444009e-01 -8.72287154e-01 -4.92570162e-01 -5.03616333e-01
2.39727363e-01 3.70909840e-01 1.32152665e+00 9.78530496e-02
-1.75889209e-01 6.41865015e-01 -6.42859340e-01 1.34841233e-01
7.67500028e-02 5.80507889e-02 -3.30654651e-01 4.18473810e-01
8.55130777e-02 -5.56518614e-01 -7.15256333e-01 2.45912850e-01
-1.17441523e+00 4.12446886e-01 2.61498570e-01 8.45825195e-01
8.32561314e-01 -6.26579225e-02 6.26294594e-03 -7.76820660e-01
2.10294113e-01 -7.08465040e-01 -4.86505955e-01 1.36258513e-01
-8.64210203e-02 -1.76608130e-01 6.05090857e-01 -3.69382411e-01
-9.52794492e-01 2.07758565e-02 2.76929736e-01 -6.89154029e-01
-8.35194811e-02 7.56792128e-01 -2.04316959e-01 5.24641097e-01
1.14083007e-01 1.89411491e-01 -4.77601409e-01 -3.87725860e-01
3.55321914e-01 3.20483178e-01 3.66483092e-01 -6.55823529e-01
5.08006334e-01 6.68148577e-01 -1.30942076e-01 -9.03280497e-01
-1.05302584e+00 -7.72503436e-01 -7.21513391e-01 -3.70374978e-01
1.26113975e+00 -1.19075334e+00 -3.54035944e-01 5.13726413e-01
-9.25041378e-01 -5.91708660e-01 -2.08353430e-01 4.34233814e-01
-5.70716262e-01 3.74136806e-01 -8.92733157e-01 -3.32448989e-01
-9.06703547e-02 -1.11946571e+00 1.57187688e+00 1.54700801e-01
9.66840386e-02 -1.04643595e+00 -1.74570695e-01 5.03553987e-01
1.51364461e-01 3.67632449e-01 6.93756998e-01 -3.23215932e-01
-7.35358000e-01 4.69913445e-02 -2.37865523e-01 3.63666788e-02
2.91604418e-02 3.17067385e-01 -6.56828761e-01 -4.53043818e-01
-2.59982407e-01 -4.38016742e-01 1.06873262e+00 5.19147158e-01
1.20322526e+00 -3.75249833e-01 -3.78088117e-01 5.52310288e-01
1.40430462e+00 -1.41522717e-02 4.99295354e-01 1.16217062e-01
7.86258101e-01 3.48029733e-01 5.75493991e-01 6.49947584e-01
7.47875154e-01 1.06267107e+00 2.52468407e-01 6.34974688e-02
-3.02961469e-01 -8.99651647e-02 6.31824970e-01 9.61130083e-01
6.76587522e-02 -4.96776491e-01 -8.34528744e-01 6.83828294e-01
-2.31878042e+00 -1.22117662e+00 -7.71728382e-02 2.00610018e+00
7.46647954e-01 1.45515800e-01 3.60293001e-01 -9.04751569e-02
8.61400545e-01 5.84782124e-01 -4.47535545e-01 -9.72489715e-02
-6.26572967e-02 -3.09000254e-01 1.18513435e-01 6.39626443e-01
-1.17157781e+00 8.44355762e-01 5.00300026e+00 8.68916094e-01
-1.12493849e+00 4.19529676e-01 4.88731533e-01 -4.98039335e-01
-1.46690607e-01 1.52935907e-02 -7.04627991e-01 8.85046482e-01
8.93915474e-01 -1.16681911e-01 4.64602739e-01 3.21783662e-01
3.98821175e-01 -1.50626764e-01 -1.19264829e+00 8.04260612e-01
1.73031330e-01 -1.44706082e+00 2.52095964e-02 -2.42938250e-01
7.78490067e-01 1.85526446e-01 -1.25138074e-01 1.00432932e-01
2.96349246e-02 -4.66621786e-01 1.18643808e+00 5.00017762e-01
7.68979311e-01 -4.12793577e-01 4.26387548e-01 1.99091081e-02
-1.89986110e+00 -8.14619288e-02 2.26287752e-01 3.44686508e-01
2.89524406e-01 6.30158365e-01 8.11373070e-02 7.44788349e-01
1.07336593e+00 1.15223658e+00 -3.75773937e-01 7.39893198e-01
-2.42098123e-02 5.63839376e-01 -3.09951127e-01 2.95758635e-01
4.26284283e-01 9.34611037e-02 6.18722141e-01 1.62389159e+00
2.12192580e-01 1.98129728e-01 5.58751285e-01 3.21361244e-01
-2.43821830e-01 -1.02422861e-02 -2.08901748e-01 -1.22204803e-01
6.30169690e-01 1.26281250e+00 -7.66154885e-01 -5.66191912e-01
-9.81606424e-01 1.06059456e+00 1.41873062e-01 3.84170443e-01
-1.19170284e+00 -1.14322141e-01 5.80834985e-01 1.58542275e-01
7.51590014e-01 -2.53493071e-01 -4.82963771e-02 -1.57114470e+00
2.59231210e-01 -7.05535889e-01 8.12200367e-01 -8.92403424e-01
-9.99438465e-01 4.44418758e-01 1.08690761e-01 -1.35794187e+00
5.17386943e-02 -2.07129300e-01 -4.69126523e-01 3.65952760e-01
-1.40115201e+00 -1.24390149e+00 -5.17999411e-01 1.02207410e+00
1.03157544e+00 2.21205533e-01 2.20271915e-01 4.15473551e-01
-5.11909425e-01 4.42269206e-01 1.96450688e-02 -2.90936237e-04
6.77666903e-01 -1.12187290e+00 1.13821793e-02 7.67615080e-01
2.35105664e-01 1.72285467e-01 5.17593741e-01 -3.32282096e-01
-1.51832426e+00 -9.50631738e-01 1.01306629e+00 -3.04132074e-01
1.07827234e+00 -6.50320411e-01 -7.54493535e-01 7.43549645e-01
5.19239783e-01 2.64609367e-01 4.47305411e-01 -4.27498370e-02
-5.65748155e-01 -1.67877629e-01 -8.10564101e-01 5.41383982e-01
1.02731717e+00 -8.10894072e-01 -3.42972487e-01 4.06509548e-01
1.04321778e+00 -5.25764942e-01 -1.04133642e+00 1.91028029e-01
5.65022707e-01 -1.10666490e+00 8.18982184e-01 -3.43952894e-01
8.34902346e-01 -3.29173058e-01 -3.79951596e-01 -9.28282559e-01
-3.69977295e-01 -7.13030040e-01 -5.91330767e-01 1.35830748e+00
3.61371011e-01 -1.26734823e-01 3.58563066e-01 3.10056150e-01
-3.01873505e-01 -7.96133399e-01 -1.05777657e+00 -7.30641425e-01
-2.87749946e-01 -3.40690762e-01 2.38512501e-01 1.15944374e+00
2.49278143e-01 6.09171212e-01 -5.35667896e-01 1.02076389e-01
4.40446854e-01 3.28935891e-01 4.73481596e-01 -9.80522931e-01
-3.33834678e-01 -6.25434101e-01 -3.45211327e-01 -1.01529002e+00
8.93095881e-02 -7.87369967e-01 -9.95970611e-03 -1.42060423e+00
4.79662716e-01 2.20535398e-01 -4.48886633e-01 5.23996711e-01
-2.03998372e-01 1.75913587e-01 4.70978320e-01 2.96532989e-01
-1.11776304e+00 5.19317687e-01 1.09646726e+00 -1.35870039e-01
-1.57360911e-01 -4.48873281e-01 -2.41761699e-01 5.85144103e-01
6.33680820e-01 -5.70310295e-01 -4.12208438e-01 -7.19507277e-01
1.45562947e-01 4.40840691e-01 2.95006186e-01 -9.13291335e-01
4.31923687e-01 -1.07561216e-01 7.15751052e-02 -6.18872464e-01
5.37625432e-01 -9.05670345e-01 1.79822266e-01 1.65856034e-01
-6.25070095e-01 2.10111544e-01 1.13381989e-01 7.36428142e-01
-5.79106748e-01 3.03850412e-01 9.36154962e-01 4.71974257e-03
-6.50694430e-01 6.13634825e-01 -2.27832362e-01 2.88709074e-01
9.92008984e-01 4.94303815e-02 -4.53790426e-01 -4.71227914e-01
-8.86413336e-01 5.22955954e-01 6.13812268e-01 6.87436461e-01
2.30546221e-01 -1.50387514e+00 -6.57074451e-01 -1.78814635e-01
2.00695157e-01 -2.75979221e-01 5.13731420e-01 1.36631000e+00
-1.56187832e-01 3.82921815e-01 2.85636801e-02 -8.95316005e-01
-1.41451478e+00 7.20426917e-01 5.55488877e-02 -5.14059782e-01
-8.18590403e-01 6.91273034e-01 4.26086783e-01 2.13717937e-01
3.94039333e-01 -3.50429773e-01 6.57906849e-03 4.01041538e-01
5.43938696e-01 2.15600505e-01 -1.18130274e-01 -8.58587503e-01
-4.38474804e-01 7.82350302e-01 -2.83141565e-02 -1.56498015e-01
1.18432999e+00 -6.07968748e-01 -1.73447967e-01 8.52014244e-01
1.42731845e+00 -1.55182257e-01 -1.58523405e+00 -5.34481466e-01
1.79850794e-02 -6.20628893e-01 1.93488955e-01 -6.23759985e-01
-1.25192463e+00 7.27922261e-01 2.38401070e-01 4.41850126e-01
1.40203166e+00 2.68381774e-01 8.40199888e-01 9.32603143e-03
1.97337002e-01 -1.27129912e+00 4.47303772e-01 5.52230239e-01
8.65999937e-01 -1.08709753e+00 4.76105027e-02 -3.46153021e-01
-7.59958029e-01 8.96078885e-01 3.51269066e-01 3.96913737e-02
8.79899919e-01 4.52291191e-01 -2.08425641e-01 -2.61371583e-01
-1.43193853e+00 -2.60521770e-01 3.88411492e-01 -8.44733715e-02
7.13041246e-01 -1.27791703e-01 -2.71597177e-01 3.86542886e-01
4.18937296e-01 9.93978307e-02 3.64347458e-01 1.05632758e+00
-3.38277310e-01 -9.63517547e-01 -1.44942239e-01 3.12705815e-01
-6.77094340e-01 -2.36495927e-01 -2.92991430e-01 8.13451469e-01
5.45506850e-02 8.78726780e-01 2.38717839e-01 -3.00056964e-01
2.10695475e-01 5.98478094e-02 3.98892432e-01 -4.88156676e-01
-6.57101810e-01 5.60322046e-01 7.62453750e-02 -1.04578984e+00
-8.60716701e-01 -1.06397700e+00 -1.16890788e+00 -4.07295793e-01
2.26021856e-02 -4.18328010e-02 3.38450611e-01 8.29761326e-01
6.65579319e-01 5.05524397e-01 6.71739280e-01 -1.15908694e+00
2.48187080e-01 -7.54226208e-01 -5.36946476e-01 6.29048347e-01
5.97339571e-01 -5.66519856e-01 -3.85784388e-01 6.06442809e-01]
|
[9.943297386169434, 0.5529834032058716]
|
184955c7-baba-4408-a0ab-ff9bc94e777f
|
tuning-models-of-code-with-compiler-generated
|
2305.18341
| null |
https://arxiv.org/abs/2305.18341v1
|
https://arxiv.org/pdf/2305.18341v1.pdf
|
Tuning Models of Code with Compiler-Generated Reinforcement Learning Feedback
|
Large Language Models (LLMs) pre-trained on code have recently emerged as the dominant approach to program synthesis. However, the code that these models produce can violate basic language-level invariants, leading to lower performance in downstream tasks. We address this issue through an approach, called RLCF, that further trains a pre-trained LLM using feedback from a code compiler. RLCF views the LLM as an RL agent that generates code step by step and receives: (i) compiler-derived feedback on whether the code it generates passes a set of correctness checks; and (ii) feedback from a different LLM on whether the generated code is similar to a set of reference programs in the training corpus. Together, these feedback mechanisms help the generated code remain within the target distribution while passing all static correctness checks. RLCF is model- and language-agnostic. We empirically evaluate it on the MBJP and MathQA tasks for Java. Our experiments show that RLCF significantly raises the odds that an LLM-generated program compiles, is executable, and produces the right output on tests, often allowing LLMs to match the performance of 2x-8x larger LLMs.
|
['Chris Jermaine', 'Thomas Reps', 'Swarat Chaudhuri', 'Chima Adiole', 'Abhinav Jain']
|
2023-05-25
| null | null | null | null |
['program-synthesis']
|
['computer-code']
|
[-8.52428749e-02 3.74348611e-01 -5.54360092e-01 -4.35710281e-01
-1.35490620e+00 -7.73357213e-01 5.36755145e-01 3.95777762e-01
-7.27090612e-02 2.74451077e-01 2.71542102e-01 -1.03718841e+00
7.28447139e-01 -9.47415173e-01 -1.38585579e+00 -3.30989272e-03
-9.41253155e-02 2.38916621e-01 2.80295968e-01 -2.01059997e-01
3.97233903e-01 -8.11560377e-02 -1.38941503e+00 7.47546017e-01
1.08872092e+00 3.83325934e-01 1.55627266e-01 1.27826536e+00
-3.15170735e-01 1.43231750e+00 -8.89369667e-01 -4.53616977e-01
1.86925903e-01 -5.41907847e-01 -9.31520104e-01 -4.07400906e-01
4.83749479e-01 -2.22064778e-01 5.25457188e-02 1.18367565e+00
-5.12480326e-02 -1.89927995e-01 4.54867363e-01 -1.41285920e+00
-6.49523020e-01 1.23553538e+00 -4.61631387e-01 -3.70188169e-02
3.18722725e-01 6.92183435e-01 1.13605738e+00 -4.31408912e-01
6.20948493e-01 1.35264730e+00 4.62938279e-01 7.14549899e-01
-1.34842658e+00 -3.03651869e-01 -5.90801686e-02 -5.57542622e-01
-9.23180223e-01 -3.73314559e-01 2.56096959e-01 -7.07241952e-01
1.45301306e+00 2.72062421e-01 1.28340304e-01 8.43798220e-01
5.98167539e-01 6.95491612e-01 7.65631497e-01 -5.05226016e-01
2.97303021e-01 2.02864885e-01 1.24801233e-01 1.09032094e+00
-3.20962779e-02 3.30933511e-01 -1.37015566e-01 -6.25207722e-01
9.18852985e-02 -4.68315065e-01 -1.77496016e-01 -2.28354841e-01
-1.10675538e+00 9.81239676e-01 3.95157218e-01 1.64157778e-01
1.32204741e-01 6.55885160e-01 7.65501678e-01 5.82727969e-01
-6.88680932e-02 9.37899530e-01 -7.45063782e-01 -3.23727280e-01
-1.01946592e+00 2.59967864e-01 1.04083955e+00 1.15832675e+00
9.23376143e-01 1.59494266e-01 -1.60586104e-01 2.85426795e-01
6.39087498e-01 6.21060073e-01 6.44177377e-01 -9.01254535e-01
8.33706021e-01 9.72004831e-01 -2.15519190e-01 -5.35728693e-01
9.68209580e-02 -2.71600634e-01 -2.95140818e-02 5.75662374e-01
3.11098933e-01 -2.00903386e-01 -6.87919736e-01 1.91193593e+00
-4.46795672e-02 -3.53577584e-01 2.28235930e-01 5.91175973e-01
4.52903628e-01 9.68036830e-01 1.22214586e-01 2.39858106e-01
8.22750628e-01 -1.12710929e+00 1.32316098e-01 -6.36493266e-01
1.36980891e+00 -8.22518945e-01 1.36218309e+00 2.04682976e-01
-1.11415589e+00 -6.04342043e-01 -9.95720625e-01 6.23160861e-02
9.97392014e-02 2.21118361e-01 5.33118129e-01 5.08736610e-01
-1.52397573e+00 3.81015718e-01 -8.93204749e-01 8.91071111e-02
7.43395910e-02 1.80876315e-01 -4.87361737e-02 1.36901990e-01
-6.47466540e-01 7.26032913e-01 2.75793195e-01 -3.29792649e-01
-1.30328369e+00 -8.45790744e-01 -1.21628535e+00 3.53255346e-02
1.63779229e-01 -4.16394770e-01 1.76971829e+00 -1.37732303e+00
-1.30278599e+00 8.54569972e-01 -2.74050534e-01 -3.65103304e-01
5.11929929e-01 -2.29622975e-01 -2.07711786e-01 -4.60369617e-01
3.11961293e-01 4.49528068e-01 8.75936329e-01 -1.20567930e+00
-6.47053003e-01 1.90025494e-02 3.16212237e-01 -3.90733600e-01
2.86467016e-01 2.79138058e-01 -3.48842621e-01 -2.39180937e-01
-6.24852598e-01 -1.08246267e+00 -2.10633755e-01 -4.23038602e-01
-5.28711259e-01 -2.36933142e-01 4.21237558e-01 -7.26775348e-01
1.30916417e+00 -2.41356707e+00 3.12827438e-01 3.22253346e-01
2.71904916e-01 1.36964098e-01 -5.59355676e-01 3.60047907e-01
-1.86614953e-02 4.74419564e-01 -2.71906435e-01 1.79176241e-01
1.88501149e-01 -1.79015264e-01 -8.04482698e-01 2.96469480e-01
4.73828822e-01 9.71622467e-01 -1.07653809e+00 -3.10071528e-01
-2.98674583e-01 -3.35839018e-02 -1.16479826e+00 5.96522689e-01
-1.09355497e+00 1.03177339e-01 -2.93101311e-01 3.60719055e-01
2.74057239e-01 -3.35538805e-01 3.44650261e-02 3.03621024e-01
-2.33150259e-01 6.72866404e-01 -5.86910188e-01 1.46633554e+00
-9.46273088e-01 6.81395411e-01 -2.61197379e-03 -3.73244166e-01
8.22675705e-01 1.47287576e-02 -3.50385517e-01 -5.80722570e-01
-9.81654301e-02 4.23078567e-01 5.23690939e-01 -5.67896664e-01
3.69449943e-01 1.83266610e-01 -5.95703542e-01 9.20842469e-01
-2.04784330e-02 -3.85204732e-01 3.98863912e-01 5.05827665e-01
1.46633828e+00 4.69388783e-01 1.21210888e-01 -2.67209113e-01
5.76052964e-01 1.84530735e-01 4.53187287e-01 1.05687439e+00
1.24629401e-01 1.85673177e-01 9.92174506e-01 -2.20390350e-01
-1.16080070e+00 -8.90597522e-01 4.13171470e-01 1.59115136e+00
-3.28719974e-01 -8.24817121e-01 -8.80557775e-01 -1.08954418e+00
1.28628314e-01 1.15052414e+00 -5.50126314e-01 -5.25905609e-01
-8.48532498e-01 -2.65217811e-01 9.09042537e-01 4.45210963e-01
7.66037032e-02 -1.14109695e+00 -7.80638099e-01 1.17636345e-01
-3.63454334e-02 -4.30153996e-01 -8.85802925e-01 2.61191070e-01
-6.93469882e-01 -1.14459336e+00 -3.43198776e-02 -6.84438705e-01
9.19847727e-01 -3.41051668e-01 1.62579501e+00 6.48332298e-01
-5.94016165e-02 -5.15507981e-02 -2.32692942e-01 -1.09935328e-01
-1.78178418e+00 2.17781290e-01 -4.70615566e-01 -4.57772583e-01
4.31676134e-02 -8.17823261e-02 3.73653583e-02 -6.14430122e-02
-7.98609555e-01 1.16918884e-01 6.38329268e-01 6.73349738e-01
6.38099853e-03 -1.76509544e-01 3.29249322e-01 -1.22469842e+00
6.68268859e-01 -5.44266522e-01 -1.23681080e+00 3.23149562e-01
-3.96371514e-01 7.68577754e-01 1.10320103e+00 -3.48874509e-01
-9.31702018e-01 -4.91771437e-02 -2.28890017e-01 -1.42363310e-01
5.03332056e-02 8.08031619e-01 -1.54607440e-03 1.67001411e-01
1.15766203e+00 3.71187776e-01 8.49922281e-03 -9.33866501e-02
4.52445984e-01 3.65097970e-01 4.95480686e-01 -1.11412418e+00
1.00433087e+00 -2.68539906e-01 -4.78425384e-01 -2.01141611e-01
-5.34049034e-01 9.91050228e-02 -9.22828391e-02 4.11843285e-02
6.07496142e-01 -7.84547031e-01 -5.07894397e-01 3.40450823e-01
-1.24194670e+00 -1.09019589e+00 -9.24162120e-02 7.23957643e-02
-4.71271455e-01 3.51824686e-02 -8.12434435e-01 -6.65075719e-01
-2.86071688e-01 -1.75726092e+00 1.05945957e+00 7.14158043e-02
-5.94288945e-01 -8.11295211e-01 2.66070276e-01 -6.59042299e-02
5.43657541e-01 1.22889522e-02 1.77638924e+00 -6.17527604e-01
-8.34213138e-01 -9.43013281e-02 7.75623694e-02 4.43092793e-01
-1.24342836e-01 6.58762097e-01 -7.64867544e-01 -3.19253147e-01
-1.92807809e-01 -5.86392283e-01 5.74103653e-01 -6.72659203e-02
8.36278677e-01 -5.87313831e-01 -2.11008817e-01 5.72155893e-01
1.39668226e+00 -1.23272687e-01 2.89387196e-01 2.36467153e-01
5.52277863e-01 3.52526098e-01 3.44432861e-01 7.62908459e-02
3.19654137e-01 1.39237389e-01 4.62057650e-01 5.56313768e-02
-1.01298325e-01 -4.51735854e-01 1.09994304e+00 9.47469473e-01
5.78267634e-01 2.27975082e-02 -1.20183003e+00 5.36225259e-01
-1.62689292e+00 -6.53440237e-01 -3.04229170e-01 2.23827028e+00
1.26804221e+00 3.75194818e-01 1.14710711e-01 -3.66988808e-01
3.18887085e-01 -1.12404145e-01 -6.32484972e-01 -9.53095138e-01
2.76918143e-01 2.37052411e-01 3.96006703e-01 7.28068292e-01
-7.62408674e-01 1.00157440e+00 6.11127520e+00 3.51852804e-01
-1.40206397e+00 -6.71945736e-02 5.41208029e-01 2.51635492e-01
-8.10102403e-01 3.64237159e-01 -7.78880417e-01 4.72449392e-01
1.47708142e+00 -5.48562050e-01 5.91327965e-01 1.32530642e+00
-1.19371526e-01 -3.25217783e-01 -1.67652583e+00 1.78400740e-01
3.62887532e-02 -1.27788293e+00 -8.50867014e-03 -2.13700552e-02
8.86537671e-01 5.27611315e-01 -5.60296997e-02 1.07232463e+00
9.90602136e-01 -1.00556016e+00 1.21607590e+00 1.89755663e-01
7.30453193e-01 -7.35876679e-01 7.08746850e-01 6.93238854e-01
-8.42297375e-01 -7.09828511e-02 -4.50410321e-02 -6.17186315e-02
-3.26923370e-01 3.87483686e-01 -1.18224382e+00 8.84204172e-03
2.01892912e-01 1.52436823e-01 -1.08502889e+00 7.29428470e-01
-5.80863953e-01 8.39967489e-01 1.69384137e-01 -2.37864897e-01
3.18640530e-01 2.60313779e-01 1.86152622e-01 1.46584427e+00
1.06270991e-01 -5.92730284e-01 3.32721680e-01 1.44355083e+00
-3.03237468e-01 -3.78704593e-02 -6.24671757e-01 -4.76351738e-01
2.74705768e-01 1.03947604e+00 -2.86006033e-01 -6.02788031e-01
-6.82743430e-01 5.95943749e-01 5.74444115e-01 2.46914610e-01
-8.96800756e-01 -4.76567537e-01 5.71390986e-01 -5.55209629e-02
5.08546978e-02 -9.08726677e-02 -5.36680296e-02 -1.20136821e+00
3.30745056e-02 -1.52299881e+00 2.36867875e-01 -7.00172246e-01
-8.20598483e-01 6.67826414e-01 -2.72104174e-01 -6.81538522e-01
-7.51505494e-01 -4.73853916e-01 -7.71854460e-01 1.12680840e+00
-1.30495632e+00 -6.06832325e-01 8.57572109e-02 -1.06720977e-01
5.52188814e-01 -3.03061664e-01 8.08554530e-01 6.33175448e-02
-3.64623159e-01 7.06004977e-01 -2.46648088e-01 4.46134627e-01
5.70824921e-01 -1.42423737e+00 9.13086474e-01 1.21596479e+00
-6.58324882e-02 1.10269046e+00 7.75934339e-01 -8.25115204e-01
-1.82550001e+00 -1.53183782e+00 7.41311371e-01 -8.24443460e-01
9.36050236e-01 -3.60578984e-01 -1.14872694e+00 1.02340281e+00
1.33839309e-01 -1.22011073e-01 4.81258661e-01 -4.95354459e-02
-8.80582690e-01 2.18401715e-01 -8.06050181e-01 6.14642859e-01
4.92564768e-01 -7.59924531e-01 -5.71965933e-01 1.96792677e-01
9.12802637e-01 -6.88118100e-01 -7.24536002e-01 7.65760019e-02
2.80950367e-01 -8.75365198e-01 2.65109479e-01 -9.10214961e-01
1.11854303e+00 -5.54558098e-01 -2.41200060e-01 -1.44670129e+00
-4.92704622e-02 -5.75848401e-01 -8.43681321e-02 1.17639697e+00
8.35117638e-01 -4.46550518e-01 4.02610272e-01 6.69792950e-01
-2.40331307e-01 -5.74727416e-01 -2.37273291e-01 -7.33660758e-01
3.89232814e-01 -5.47844648e-01 6.54680789e-01 6.58323228e-01
1.29666969e-01 4.53297317e-01 1.61237404e-01 2.17265353e-01
2.93241173e-01 4.82257456e-01 1.11804402e+00 -6.38249218e-01
-9.71589684e-01 -7.58184969e-01 -3.29809524e-02 -8.50796103e-01
6.54042900e-01 -1.54348445e+00 6.15150630e-01 -1.03835857e+00
3.91967654e-01 -4.84462917e-01 2.75446475e-01 6.74553692e-01
-3.03612441e-01 -3.79757226e-01 1.68856010e-01 1.12732396e-01
-5.54802537e-01 1.46068275e-01 7.08414137e-01 -2.12717310e-01
-9.09718275e-02 -4.52470146e-02 -8.49528491e-01 6.96697593e-01
6.82907999e-01 -6.31787419e-01 -2.81223983e-01 -6.55501246e-01
7.13136792e-01 3.95538598e-01 1.21710159e-01 -8.58061492e-01
-3.26867662e-02 -3.14858377e-01 -6.41012192e-02 -5.41370139e-02
-6.60211563e-01 -1.96386904e-01 -2.20613964e-02 8.63199234e-01
-8.74715269e-01 4.35762465e-01 3.80152643e-01 2.28516102e-01
-3.20247747e-02 -5.64751565e-01 7.68048108e-01 -3.39144409e-01
-6.68428183e-01 -6.47557452e-02 -5.81405520e-01 5.13246834e-01
6.88355029e-01 4.80633199e-01 -5.89013517e-01 -8.63892585e-02
-8.36427435e-02 3.04543555e-01 8.91882181e-01 7.11009741e-01
1.84568495e-01 -8.45079184e-01 -8.04736733e-01 5.09233534e-01
4.62242007e-01 -1.49525315e-01 -4.79809701e-01 5.57765841e-01
-6.19826376e-01 3.74380946e-01 1.95322260e-01 -6.65485024e-01
-1.02909303e+00 5.71612656e-01 3.80223900e-01 -2.94506192e-01
-3.28883260e-01 9.16115046e-01 3.70996654e-01 -1.13909006e+00
-1.05687663e-01 -7.30211377e-01 5.56476057e-01 -7.32695758e-01
5.03997445e-01 -1.33711427e-01 1.88414097e-01 -3.93976599e-01
-3.37782502e-01 -3.01740561e-02 1.04825245e-03 -1.48107573e-01
1.09896851e+00 7.44049132e-01 -5.92878878e-01 6.62598789e-01
1.30138421e+00 6.65026248e-01 -1.11660659e+00 -1.39742298e-02
3.12302262e-01 -2.87897736e-01 -1.98937088e-01 -1.02084482e+00
-8.58595133e-01 7.51718402e-01 -1.01898864e-01 7.28221685e-02
6.00027323e-01 1.70359164e-01 4.87009317e-01 3.59517902e-01
6.31244242e-01 -6.76070511e-01 2.31903568e-01 7.40227103e-01
7.92728961e-01 -1.19672716e+00 -5.01233578e-01 8.16443004e-03
-3.91709954e-01 1.13784432e+00 1.11559463e+00 -2.40924209e-01
4.35079522e-02 7.86945701e-01 6.93268627e-02 1.42227650e-01
-1.23161364e+00 3.83727819e-01 1.34688348e-01 3.07308525e-01
9.32625353e-01 1.48265943e-01 1.84864383e-02 4.34770197e-01
-4.96949017e-01 -1.11627869e-01 7.82121658e-01 1.12274194e+00
-5.13690233e-01 -1.21778989e+00 -3.77359122e-01 7.10537314e-01
-3.67180884e-01 -3.43948990e-01 -3.53842556e-01 7.35087991e-01
-1.10173352e-01 7.26894379e-01 -2.33078431e-02 -3.36233586e-01
6.67911395e-02 1.67599842e-01 5.03640354e-01 -1.28910708e+00
-9.86578166e-01 -2.58210123e-01 2.13835046e-01 -7.36974180e-01
4.77434337e-01 -5.57803094e-01 -1.67236221e+00 -2.19767168e-01
-2.09993303e-01 3.29048365e-01 5.37773490e-01 7.15406537e-01
5.04264295e-01 6.43905640e-01 4.75543797e-01 -2.70794183e-01
-9.81893241e-01 -6.76639676e-01 1.59254596e-01 3.77896726e-01
6.71867728e-01 1.69109762e-01 -3.14812660e-01 2.95570791e-01]
|
[7.865427494049072, 7.696400165557861]
|
adb45170-e369-4615-954b-63c5d93cad98
|
an-improved-air-light-estimation-scheme-for
| null | null |
https://ieeexplore.ieee.org/document/9201388
|
https://ieeexplore.ieee.org/document/9201388
|
An Improved Air-Light Estimation Scheme for Single Haze Images Using Color Constancy Prior
|
Hazy environment attenuates the scene radiance and
causes difficulty in distinguishing the color and texture of the scene.
A crucial step in dehazing is the recovery of the global air-light
vector. Traditional methods usually interpret the RGB value of
the brightest region in haze images as the air-light. In this letter,
a new prior called ‘color constancy prior’ has been proposed to
improve the robustness of air-light estimation when varicolored
illumination exists.The prior utilizes the statistical observation that
distant scenery objects become the most haze-opaque due to the
pixel escalation towards the higher intensity side. The comparative
evaluation on a variety of haze images manifests that the proposed
prior perform better than existing air-light recovery methods and
can be used for subsequent dehazing applications.
|
['B.K. Panigrahi', 'Tapan Kumar Gandhi', 'Sidharth Gautam']
|
2020-09-21
| null | null | null | null |
['color-constancy']
|
['computer-vision']
|
[ 4.06640947e-01 -7.09319949e-01 5.42312086e-01 -1.58126250e-01
-8.53767768e-02 -2.56439596e-01 3.69302034e-01 -4.50561255e-01
-6.21300936e-02 7.23984480e-01 -6.12248182e-02 -1.38995886e-01
1.87120542e-01 -7.69727409e-01 -2.26244837e-01 -1.49835289e+00
4.30420756e-01 -3.68395329e-01 3.20621312e-01 -5.01883984e-01
3.02313536e-01 1.51478276e-01 -1.62782812e+00 1.74420401e-01
8.18609178e-01 7.87576437e-01 1.83958545e-01 8.18035722e-01
-1.19243991e-02 8.17341745e-01 -6.40794516e-01 1.17637496e-02
7.11944938e-01 -4.94978875e-01 -5.06259240e-02 2.40718275e-01
6.76349878e-01 -5.32136858e-01 -2.63450891e-01 1.33231378e+00
1.78526968e-01 3.32989872e-01 4.94412512e-01 -9.83424544e-01
-8.67845237e-01 -5.28802872e-01 -8.09278905e-01 4.85515922e-01
1.26306161e-01 1.49900675e-01 5.19758344e-01 -1.17181325e+00
4.18870933e-02 9.17550385e-01 2.31630936e-01 2.20765218e-01
-6.52306914e-01 -6.00486636e-01 -1.96837038e-01 4.26656008e-01
-1.55216789e+00 -2.87789434e-01 9.55327511e-01 -2.25460991e-01
4.10264015e-01 6.90062463e-01 6.73702955e-01 3.75356406e-01
6.99237406e-01 3.64834577e-01 1.74222815e+00 -5.43673217e-01
2.49254197e-01 6.31542504e-02 -9.94396433e-02 5.34188449e-01
7.73634076e-01 4.79397476e-01 -5.92314839e-01 1.25001401e-01
4.38350916e-01 3.65680188e-01 -5.81283152e-01 1.07806027e-01
-7.39915013e-01 3.39668304e-01 4.22703296e-01 3.29031684e-02
-3.78381014e-01 1.07587606e-01 -6.52643442e-01 2.19294086e-01
6.81223154e-01 2.52245814e-01 -4.31076884e-02 1.94590375e-01
-8.72891486e-01 1.28126293e-02 2.53669262e-01 5.50648928e-01
1.00788999e+00 3.68013889e-01 2.34231308e-01 5.75773835e-01
7.02870190e-01 1.12775731e+00 -1.73043862e-01 -8.50272477e-01
1.38105229e-01 3.02665085e-01 6.31094635e-01 -9.16216075e-01
1.51383489e-01 -2.56697744e-01 -7.37673461e-01 9.08754706e-01
2.77723342e-01 -1.69672340e-01 -1.22347486e+00 9.30630624e-01
5.19238412e-01 2.57000804e-01 2.54926473e-01 1.24056292e+00
7.58912742e-01 1.08813477e+00 -5.66282511e-01 -1.28595114e-01
9.79162157e-01 -7.64804304e-01 -1.17805994e+00 -2.82446265e-01
-2.65034765e-01 -1.22687268e+00 7.43844032e-01 7.06401110e-01
-9.48805094e-01 -4.49319243e-01 -1.21489704e+00 9.44180712e-02
-3.43479425e-01 -2.26714075e-01 4.56073910e-01 7.92898417e-01
-1.10102093e+00 -3.95433120e-02 -5.94454467e-01 1.47211298e-01
-6.77563101e-02 -1.81133971e-01 -1.58209279e-02 -7.35324204e-01
-1.00477743e+00 9.80805933e-01 -1.74896214e-02 8.22122097e-01
-8.94938886e-01 -5.47330618e-01 -5.31354427e-01 -3.16000432e-01
1.97315291e-01 -3.17070425e-01 6.30453229e-01 -1.11245632e+00
-1.55977130e+00 5.28874636e-01 -2.93200046e-01 2.78538018e-01
3.93557474e-02 -4.14193809e-01 -7.40931094e-01 2.72973806e-01
-4.48393136e-01 -2.61323806e-02 1.53946078e+00 -1.97408187e+00
-6.37852311e-01 -1.54147834e-01 9.72286612e-02 3.79036635e-01
2.96410412e-01 -1.12773776e-01 -4.04984713e-01 -6.05617702e-01
2.66602367e-01 -8.57055962e-01 -1.30062224e-02 2.04962417e-02
-3.03723633e-01 3.41335922e-01 9.35575247e-01 -7.31794238e-01
1.02221107e+00 -2.32028151e+00 -2.51810819e-01 2.96459436e-01
3.23377401e-01 2.37261057e-01 1.54763162e-01 3.53466630e-01
-2.16461018e-01 -4.39678282e-01 -3.49751592e-01 8.66482258e-02
-3.61302286e-01 3.34878743e-01 -1.52393937e-01 1.10027266e+00
6.89332411e-02 2.87887752e-01 -8.87246490e-01 -1.58276707e-01
3.52146298e-01 8.69845569e-01 -1.48366794e-01 3.16937864e-01
2.20902786e-01 6.12294197e-01 -8.42507184e-02 7.10217953e-01
1.47686160e+00 2.76547462e-01 -5.14048696e-01 -1.68773562e-01
-6.42160058e-01 -1.91847324e-01 -1.09518492e+00 7.96980977e-01
-3.65511805e-01 1.12976265e+00 4.30616587e-01 -1.42813221e-01
8.87076855e-01 2.02168196e-01 9.71401855e-02 -7.86973655e-01
1.39087930e-01 -3.33367698e-02 -1.54291272e-01 -6.79332614e-01
7.35034883e-01 -5.32536745e-01 7.35169113e-01 2.05389708e-01
-6.89020157e-01 -6.09159648e-01 -3.08019459e-01 -8.97545964e-02
5.25279224e-01 -1.72439497e-02 -1.31086290e-01 -2.72432268e-01
5.20826280e-01 -1.87446922e-01 4.30165976e-01 4.53596801e-01
-2.13659152e-01 7.83700049e-01 -3.83165061e-01 -3.76617938e-01
-6.87908530e-01 -1.12883127e+00 -1.91402569e-01 9.89084125e-01
6.21261179e-01 1.82883903e-01 -5.33270538e-01 1.47555605e-01
-3.26845616e-01 1.10104334e+00 -4.94972706e-01 -1.36603609e-01
-3.74689788e-01 -9.76794839e-01 -3.06863457e-01 -2.59073466e-01
8.37184548e-01 -3.95417869e-01 -7.23659515e-01 -3.35801691e-01
-2.70791829e-01 -8.68864477e-01 -1.47592098e-01 -9.73299593e-02
-6.40572369e-01 -9.75337863e-01 -6.55151725e-01 -3.52299362e-01
8.23020637e-01 1.18258953e+00 1.00082159e+00 4.38339084e-01
-3.41316998e-01 4.20784831e-01 -5.74100733e-01 -6.13143444e-01
-2.52215236e-01 -1.00320387e+00 -2.89736539e-01 6.49942398e-01
2.30659887e-01 -1.32067859e-01 -1.08906448e+00 2.76098490e-01
-1.04748809e+00 2.41523862e-01 4.31551576e-01 3.74617815e-01
3.32321882e-01 8.58292878e-01 -5.65978944e-01 -5.70817530e-01
1.17252067e-01 -4.07946259e-01 -9.59751666e-01 2.22816721e-01
-6.59345686e-01 -2.91517287e-01 3.25419992e-01 1.96758330e-01
-1.63417935e+00 -5.31273246e-01 4.23737764e-01 -2.64550090e-01
-2.20720485e-01 8.83752555e-02 -2.73670331e-02 -4.85702902e-01
3.79448891e-01 4.23465312e-01 -4.56760466e-01 -3.59179556e-01
1.79620445e-01 4.63938028e-01 5.41160285e-01 -2.72248060e-01
1.50129986e+00 1.12310338e+00 4.22256202e-01 -1.31285536e+00
-7.19350755e-01 -8.94879639e-01 -2.36649290e-01 -6.39340997e-01
1.10232389e+00 -8.86304855e-01 -5.05528092e-01 7.69517422e-01
-8.91899943e-01 -4.02792484e-01 1.35048151e-01 5.69088995e-01
3.41785669e-01 5.09151757e-01 -3.89375985e-01 -1.24703109e+00
-9.42279324e-02 -8.88520896e-01 8.51690769e-01 4.89384770e-01
6.26529932e-01 -1.13743389e+00 1.92717001e-01 5.62256992e-01
7.09692717e-01 2.30697066e-01 6.48097456e-01 7.33304620e-01
-1.05666041e+00 -2.49435995e-02 -3.53908062e-01 5.38048565e-01
5.18418550e-01 4.76365179e-01 -1.26134348e+00 -2.39702269e-01
3.93135965e-01 4.12898302e-01 7.93157220e-01 3.50806475e-01
5.79864085e-01 -4.52268459e-02 3.29636872e-01 9.02102053e-01
1.82054722e+00 2.32238889e-01 9.36252236e-01 5.41456044e-01
7.54390121e-01 6.56605005e-01 9.34147000e-01 4.69694436e-01
3.40165608e-02 4.65305567e-01 8.73673618e-01 -8.18547308e-01
-3.43664885e-01 4.82416451e-01 4.47850674e-01 7.87685513e-01
-5.78241050e-01 -5.66183984e-01 -7.37022042e-01 3.53558987e-01
-1.01716757e+00 -6.83879435e-01 -7.96853721e-01 2.28364825e+00
6.57006383e-01 -3.72494876e-01 -6.05871856e-01 6.79149628e-02
6.04103744e-01 2.04760239e-01 -1.26263753e-01 -1.76059991e-01
-3.75328332e-01 3.33560675e-01 8.20209861e-01 1.03097117e+00
-7.57832170e-01 7.55736172e-01 6.60919571e+00 2.55549788e-01
-1.17121506e+00 5.11467010e-02 2.56693661e-01 1.08343922e-01
-5.59939742e-01 6.88641146e-02 -5.08318365e-01 5.49076200e-01
6.41313314e-01 2.81105280e-01 6.67432785e-01 -7.67585039e-02
5.46553552e-01 -8.57698202e-01 -4.41640794e-01 1.15005517e+00
1.63351461e-01 -5.74782073e-01 4.60917838e-02 -5.37310466e-02
9.71317053e-01 -1.41396910e-01 6.12698793e-01 -5.36483228e-01
2.89102763e-01 -6.89339221e-01 6.74586892e-01 8.25644493e-01
7.61992574e-01 -5.23994029e-01 7.62758017e-01 -9.35698375e-02
-9.88121390e-01 2.99002081e-01 -5.53913772e-01 -2.14004070e-01
1.40032172e-01 8.24702680e-01 -7.34156668e-01 4.86168623e-01
8.61633480e-01 5.15554726e-01 -6.91067100e-01 1.00896406e+00
-7.11869240e-01 7.57477939e-01 -2.36587241e-01 4.14534390e-01
3.08254182e-01 -9.45098877e-01 6.32183015e-01 8.71671975e-01
2.52464712e-01 5.85023165e-01 -2.13930085e-01 8.12478065e-01
5.39011896e-01 -4.00923729e-01 -2.55810529e-01 2.83626497e-01
-1.10333689e-01 9.55181718e-01 -5.76541662e-01 -1.65808871e-01
-7.21068799e-01 1.14062250e+00 -7.00557351e-01 1.17335486e+00
-7.08261073e-01 -3.20674747e-01 6.18828297e-01 -3.74863334e-02
2.38690123e-01 -5.43693662e-01 -3.47276479e-01 -8.86180282e-01
-3.59735370e-01 -5.96111417e-01 -7.27010658e-03 -1.54475760e+00
-9.61538970e-01 3.57460856e-01 1.20216385e-02 -1.18457639e+00
5.10586500e-01 -8.82230759e-01 -5.92940450e-01 1.15573180e+00
-2.19299197e+00 -8.27092946e-01 -9.13045228e-01 1.02946377e+00
4.25758421e-01 3.14900428e-01 4.40865219e-01 1.93687662e-01
-6.45827949e-01 -7.60378242e-02 6.20290518e-01 -2.28494197e-01
6.66671932e-01 -1.32315934e+00 -1.17950402e-01 1.69704819e+00
-3.08075815e-01 7.54708171e-01 1.30132914e+00 -5.27308345e-01
-1.61424422e+00 -9.12144482e-01 2.73430824e-01 -3.40004444e-01
3.91783476e-01 -6.96627609e-03 -8.25258791e-01 3.57736498e-01
6.42743289e-01 1.63763575e-02 7.58786619e-01 -5.29619217e-01
-1.98174283e-01 -4.34238017e-01 -9.63050485e-01 5.73411465e-01
9.23819393e-02 -5.06494999e-01 -6.19513094e-01 1.35260835e-01
1.88710004e-01 -2.22732872e-01 -4.83402580e-01 1.76678792e-01
3.86753738e-01 -1.31018984e+00 1.05821061e+00 8.83495286e-02
1.33819312e-01 -8.98441374e-01 -4.69661444e-01 -1.14807689e+00
-4.46205318e-01 -6.36850595e-01 6.73703179e-02 7.61428535e-01
1.05177918e-02 -8.18905473e-01 4.43571538e-01 5.75433433e-01
-2.29518130e-01 -1.06820822e-01 -5.79697609e-01 -4.06956196e-01
-3.79849434e-01 -7.62974173e-02 4.17817891e-01 9.12398517e-01
-7.60444582e-01 4.56918776e-02 -6.80251122e-01 9.30536509e-01
1.19946921e+00 3.07177782e-01 7.40235627e-01 -1.04009032e+00
-3.61688174e-02 1.97947328e-03 -6.42689690e-02 -7.45584548e-01
-4.25521582e-01 -2.54618406e-01 5.82683742e-01 -1.52486706e+00
1.61113799e-01 -2.11509630e-01 -5.61655879e-01 1.59016982e-01
-6.57847822e-01 6.53980315e-01 4.86677960e-02 2.78358608e-01
-3.59731615e-01 4.41167623e-01 1.50332403e+00 -2.58247316e-01
-1.60960719e-01 -4.20826189e-02 -2.61920094e-01 5.84797978e-01
8.87605548e-01 -4.04909670e-01 -3.69486123e-01 -6.89090788e-01
6.26054704e-01 -4.72238719e-01 3.81562561e-01 -9.36683953e-01
8.92769992e-02 -5.74646831e-01 3.78063858e-01 -7.60199845e-01
4.59661275e-01 -1.22991323e+00 4.10336018e-01 2.80106634e-01
3.85287434e-01 -1.32235363e-01 2.57786619e-03 6.63741767e-01
-2.69828796e-01 -1.70423046e-01 1.07831872e+00 -2.34802768e-01
-8.90914798e-01 1.79422230e-01 -7.75101602e-01 -1.27636999e-01
7.93209136e-01 -6.11851275e-01 -5.30165911e-01 -4.79541034e-01
-3.59367440e-03 -4.54450399e-01 8.17984581e-01 4.37003589e-04
9.97857392e-01 -8.19798231e-01 -7.27413774e-01 4.56995636e-01
1.03975579e-01 -4.28563319e-02 3.53951007e-01 8.83501768e-01
-1.27741539e+00 4.64974791e-02 4.30541448e-02 -6.04361713e-01
-1.38904727e+00 3.50371897e-01 3.14334363e-01 4.84637171e-01
-6.65049076e-01 9.66732264e-01 7.29658127e-01 5.00480950e-01
-3.22213233e-01 -3.60174865e-01 -9.35231671e-02 -5.78844547e-01
7.71704674e-01 8.58384967e-01 8.69683772e-02 -9.34100211e-01
-1.23400442e-01 7.25424945e-01 3.26818615e-01 -4.99973297e-02
1.25976038e+00 -6.86477840e-01 -7.07213759e-01 6.84845328e-01
9.01959479e-01 5.10544777e-01 -1.41911793e+00 -2.40580924e-02
-6.46976829e-01 -1.39542532e+00 7.99702764e-01 -6.49083853e-01
-1.23546314e+00 8.71344268e-01 8.68080795e-01 1.61208600e-01
1.45703387e+00 -3.33493650e-01 6.53261423e-01 1.86534524e-01
2.23910600e-01 -9.33967292e-01 5.76011278e-02 3.30935657e-01
6.52028382e-01 -1.08456802e+00 3.80028248e-01 -4.22447264e-01
-3.02256346e-01 1.17011666e+00 3.65179628e-01 1.12487853e-01
6.86432660e-01 1.44484937e-01 6.12347722e-01 -5.09315252e-01
-2.35628411e-01 -2.55980372e-01 4.25327301e-01 6.25753403e-01
1.74816787e-01 -5.57970069e-02 2.57590979e-01 -4.25871342e-01
1.10485218e-01 -5.14089227e-01 7.48622179e-01 8.48806381e-01
-6.86663449e-01 -5.63799620e-01 -1.24382353e+00 -1.89313725e-01
-5.96277714e-01 -3.92027646e-01 -1.96517825e-01 4.50995833e-01
4.75370437e-01 1.50754654e+00 -3.09610050e-02 3.68131883e-02
-7.42290318e-02 -3.42303038e-01 5.66995859e-01 -3.86547744e-01
5.74613214e-02 4.08161968e-01 -3.13754082e-01 -4.98568863e-01
-9.57722306e-01 -2.82455951e-01 -9.23446476e-01 -4.41520512e-01
-5.69842219e-01 1.72979310e-01 7.73392379e-01 6.09931111e-01
-3.50800335e-01 4.15826738e-01 9.10534680e-01 -7.81495988e-01
3.93492818e-01 -5.93632221e-01 -1.25204861e+00 4.15577352e-01
1.02279007e+00 -7.82492757e-01 -1.00530434e+00 1.73170984e-01]
|
[10.833403587341309, -3.1369264125823975]
|
d7ac21ab-f68f-4b8b-9a16-256bd3e133b6
|
self-training-for-class-incremental-semantic
|
2012.03362
| null |
https://arxiv.org/abs/2012.03362v3
|
https://arxiv.org/pdf/2012.03362v3.pdf
|
Self-Training for Class-Incremental Semantic Segmentation
|
In class-incremental semantic segmentation, we have no access to the labeled data of previous tasks. Therefore, when incrementally learning new classes, deep neural networks suffer from catastrophic forgetting of previously learned knowledge. To address this problem, we propose to apply a self-training approach that leverages unlabeled data, which is used for the rehearsal of previous knowledge. Specifically, we first learn a temporary model for the current task, and then pseudo labels for the unlabeled data are computed by fusing information from the old model of the previous task and the current temporary model. Additionally, conflict reduction is proposed to resolve the conflicts of pseudo labels generated from both the old and temporary models. We show that maximizing self-entropy can further improve results by smoothing the overconfident predictions. Interestingly, in the experiments we show that the auxiliary data can be different from the training data and that even general-purpose but diverse auxiliary data can lead to large performance gains. The experiments demonstrate state-of-the-art results: obtaining a relative gain of up to 114% on Pascal-VOC 2012 and 8.5% on the more challenging ADE20K compared to previous state-of-the-art methods.
|
['Joost Van de Weijer', 'Xialei Liu', 'Lu Yu']
|
2020-12-06
| null | null | null | null |
['class-incremental-semantic-segmentation']
|
['computer-vision']
|
[ 5.42454779e-01 4.12482649e-01 -2.89791107e-01 -5.23129463e-01
-8.30335677e-01 -6.05376184e-01 3.83114994e-01 5.04680425e-02
-7.21205771e-01 1.11563122e+00 -1.35416379e-02 -1.46939635e-01
2.53369063e-01 -3.65811080e-01 -9.52496290e-01 -7.12971747e-01
1.94057554e-01 5.99841893e-01 4.99417245e-01 1.22994147e-01
1.58657841e-02 3.96922342e-02 -1.70800805e+00 3.82830381e-01
1.12023997e+00 1.20621979e+00 4.40256119e-01 3.98172021e-01
-1.32019103e-01 7.98898339e-01 -5.61524332e-01 -4.41395253e-01
1.97928011e-01 -1.98054954e-01 -1.08655608e+00 1.47824883e-01
6.11687839e-01 -2.69124001e-01 -1.02268584e-01 1.02149534e+00
2.18396947e-01 3.23038340e-01 4.02094662e-01 -1.12666678e+00
-4.74619210e-01 7.22722650e-01 -4.37349916e-01 -2.35749758e-03
-3.44184905e-01 1.42760605e-01 8.45430970e-01 -9.44009483e-01
7.82975793e-01 1.00175869e+00 6.61040366e-01 9.22438145e-01
-1.24790728e+00 -7.38949537e-01 6.99832797e-01 3.64756227e-01
-1.09941375e+00 -5.52954555e-01 5.34262359e-01 -2.84763634e-01
8.04129124e-01 -3.05447485e-02 5.91054678e-01 1.08876073e+00
-7.07127005e-02 1.14449573e+00 1.12586558e+00 -3.32099944e-01
3.57057333e-01 2.98521072e-01 4.60047126e-01 6.68166339e-01
1.95920363e-01 -5.70662618e-02 -5.51485956e-01 3.96238454e-02
9.49268639e-02 -2.01916769e-02 -1.63456962e-01 -4.09156024e-01
-8.75910282e-01 5.27705550e-01 4.21446055e-01 1.57797799e-01
-2.82947659e-01 -6.61507389e-03 3.11782271e-01 1.53033074e-03
5.30092776e-01 5.04905283e-01 -1.02808487e+00 -3.42953652e-02
-1.00165236e+00 -2.65012234e-02 6.91719234e-01 9.13682401e-01
1.06053817e+00 -6.84181601e-02 -3.30490053e-01 7.65438259e-01
2.26449774e-04 1.98049903e-01 6.16333246e-01 -1.24440157e+00
4.68179733e-01 5.13428807e-01 4.35507111e-02 -1.40925020e-01
-2.12593526e-01 -7.37548947e-01 -5.41027904e-01 -7.82372206e-02
4.82435733e-01 -3.34588617e-01 -1.61565924e+00 2.11160421e+00
4.32808369e-01 5.17587185e-01 2.91525692e-01 4.99029756e-01
4.93677020e-01 3.83607954e-01 2.32699007e-01 -3.73432070e-01
9.53575134e-01 -1.36196673e+00 -5.96258581e-01 -7.40390837e-01
6.30902410e-01 -4.66087341e-01 9.20415044e-01 4.74083960e-01
-8.95074785e-01 -7.37739086e-01 -1.12972498e+00 -1.92796681e-02
-1.64870054e-01 1.13066562e-01 5.32613277e-01 3.74633610e-01
-1.01265383e+00 8.20806444e-01 -1.01969492e+00 7.61515601e-03
9.38732982e-01 3.23372066e-01 -5.13111912e-02 -2.72377253e-01
-1.04434788e+00 6.82178080e-01 7.26949632e-01 2.95585413e-02
-1.08298028e+00 -1.02605200e+00 -5.17358005e-01 6.24100864e-02
9.83424723e-01 -5.04785895e-01 1.49351275e+00 -1.13462937e+00
-1.20832944e+00 8.10252964e-01 -2.99083501e-01 -6.26453459e-01
5.45326591e-01 -5.50982475e-01 -2.95281298e-02 -7.87187368e-02
2.69970328e-01 1.06582606e+00 8.08380365e-01 -1.49693203e+00
-9.89646316e-01 -5.13513863e-01 -7.53006712e-02 2.33285770e-01
-3.25581521e-01 -7.53156424e-01 -6.98073447e-01 -5.29518127e-01
2.42642164e-01 -1.30304980e+00 -2.31741399e-01 -2.49052703e-01
-4.62698132e-01 -1.64815083e-01 7.81937182e-01 -6.00167215e-01
1.01536763e+00 -2.18274522e+00 1.30579039e-01 -2.64150053e-01
1.24194026e-01 5.92890739e-01 -9.77406725e-02 -4.71976668e-01
-1.76147167e-02 1.77650794e-01 -4.25559551e-01 -7.48841107e-01
-2.84608930e-01 5.25557518e-01 -4.29854274e-01 -1.07307630e-02
2.51393020e-01 8.74813914e-01 -9.30073619e-01 -4.15594667e-01
-1.15654789e-01 1.76344916e-01 -5.06341755e-01 1.08887799e-01
-5.13295472e-01 5.65655589e-01 -3.13553989e-01 5.13579309e-01
6.65201962e-01 -3.94825429e-01 3.48954856e-01 -7.97002167e-02
2.87201226e-01 3.45005721e-01 -9.00726795e-01 1.95423150e+00
-3.54334354e-01 4.10841495e-01 -7.84123093e-02 -1.02588534e+00
5.16606152e-01 1.19039893e-01 1.92021728e-01 -5.81352532e-01
-2.18788553e-02 1.80740669e-01 -4.87531349e-02 -5.65844104e-02
5.60396433e-01 -1.92752793e-01 7.67919123e-02 3.67919177e-01
4.19732213e-01 -7.58159310e-02 1.59334555e-01 2.31543273e-01
9.86104846e-01 2.76237279e-01 -9.56122298e-03 -1.59422390e-03
1.91379085e-01 1.07250102e-01 1.02158606e+00 9.06992912e-01
-3.58324707e-01 4.00400966e-01 4.14478123e-01 -4.06794131e-01
-7.90281713e-01 -1.04194546e+00 5.59513690e-03 1.12218368e+00
1.05636641e-01 -1.71872750e-01 -8.69063973e-01 -1.21403098e+00
-4.41808142e-02 1.13494337e+00 -6.89072430e-01 -4.39418793e-01
-6.10868752e-01 -7.75366426e-01 2.10568413e-01 6.74353123e-01
6.80917263e-01 -9.79148924e-01 -5.92384696e-01 3.47858995e-01
-2.10925162e-01 -1.37173474e+00 -3.21120054e-01 5.32237589e-01
-1.18214417e+00 -9.29434359e-01 -4.58292276e-01 -6.50164127e-01
7.30665743e-01 1.98261857e-01 1.16912806e+00 6.48474842e-02
-1.00952767e-01 1.91294372e-01 -1.79318532e-01 -3.56846392e-01
-4.63906646e-01 3.64901870e-01 1.33324089e-02 -1.02681145e-02
2.47696489e-01 -4.40632254e-01 -4.40713316e-01 1.64701745e-01
-7.47468531e-01 3.18754613e-01 5.44928074e-01 1.06068671e+00
8.64096880e-01 4.95422482e-02 7.16479242e-01 -1.29652202e+00
-2.45115347e-03 -4.54167426e-01 -4.33215320e-01 4.59990710e-01
-9.82889533e-01 4.22083676e-01 5.76191366e-01 -7.16460407e-01
-1.49812233e+00 2.66375989e-01 1.34625494e-01 -6.05559587e-01
-1.27845064e-01 3.37025136e-01 -1.40558645e-01 4.18266237e-01
4.55574274e-01 1.31138703e-02 -2.69850075e-01 -6.13486230e-01
5.02270460e-01 4.42039311e-01 7.15823948e-01 -7.32181489e-01
3.64588678e-01 4.57401961e-01 -2.26650983e-01 -3.58082891e-01
-1.58918762e+00 -1.41484752e-01 -7.65577734e-01 -2.89010443e-02
8.13993096e-01 -9.98652995e-01 -2.68525481e-01 8.92202139e-01
-1.07547927e+00 -7.20895588e-01 -6.72499478e-01 2.31195569e-01
-2.80589163e-01 1.57489091e-01 -5.04180312e-01 -6.21404052e-01
-1.93236321e-01 -1.23998201e+00 7.56124318e-01 4.77911592e-01
3.18008922e-02 -8.30201507e-01 -1.79046974e-01 6.01041973e-01
1.80142909e-01 -7.89527595e-02 1.08400154e+00 -9.50213253e-01
-6.93523705e-01 7.08836541e-02 -1.21924788e-01 6.44594908e-01
1.67858303e-01 -3.88916165e-01 -1.34694922e+00 -4.23847735e-01
1.52878657e-01 -6.44954741e-01 1.52628720e+00 1.23534232e-01
1.24064434e+00 -2.72748888e-01 -5.90809882e-01 4.45651501e-01
1.11082304e+00 2.49388516e-01 3.61718357e-01 2.00505435e-01
9.88536417e-01 6.81565344e-01 8.34119976e-01 2.81832844e-01
4.82704580e-01 3.92653495e-01 4.15533543e-01 1.83511481e-01
-3.56099844e-01 -3.70089978e-01 4.43797708e-02 6.20754302e-01
2.50488400e-01 -1.46246955e-01 -9.33297217e-01 7.43496120e-01
-1.97200501e+00 -5.41406095e-01 3.37192237e-01 2.44022608e+00
1.26253521e+00 5.93698084e-01 -4.63938206e-01 -8.47735777e-02
7.82859027e-01 4.74779643e-02 -1.28312254e+00 1.12832272e-02
5.52239716e-02 3.53255361e-01 5.14630914e-01 5.46237350e-01
-1.21645939e+00 1.43038487e+00 6.08313084e+00 9.32815075e-01
-1.03942049e+00 4.07603770e-01 1.06257844e+00 -3.36306602e-01
-2.08990633e-01 2.09112108e-01 -1.06286871e+00 2.91621625e-01
9.00783539e-01 -1.58236697e-02 3.12722445e-01 9.68300998e-01
-3.72530878e-01 -4.52147007e-01 -1.29707313e+00 6.59013033e-01
1.08545944e-01 -1.14801633e+00 2.42560799e-03 -1.12084031e-01
1.12985885e+00 1.53353155e-01 1.77245125e-01 6.12326980e-01
5.19394577e-01 -6.47058189e-01 8.67867768e-01 4.16711956e-01
7.01317430e-01 -6.59099281e-01 6.67457283e-01 5.48629761e-01
-9.10372555e-01 -1.66739851e-01 -2.66288012e-01 8.91804844e-02
3.95160280e-02 6.78074956e-01 -1.00775445e+00 4.18257326e-01
7.09073186e-01 7.06794262e-01 -7.32336342e-01 6.44992888e-01
-4.83446598e-01 8.55772197e-01 -2.00465694e-01 3.27450961e-01
1.89081505e-01 1.30807668e-01 3.58429253e-01 7.70518064e-01
1.22864567e-01 3.18907723e-02 2.13056505e-01 8.06172609e-01
-4.38037127e-01 -2.46494964e-01 -1.38465822e-01 5.32255210e-02
5.53839028e-01 1.04552794e+00 -8.19145143e-01 -7.74425507e-01
-1.98018998e-01 1.20575190e+00 7.07279742e-01 4.81207341e-01
-7.45193124e-01 8.28189999e-02 5.08619964e-01 -2.16575325e-01
6.63557231e-01 -3.12424712e-02 -3.81285369e-01 -1.20051122e+00
1.12437829e-01 -5.88057458e-01 4.48427737e-01 -6.34709835e-01
-1.14436913e+00 5.62785268e-01 6.35558590e-02 -7.02636719e-01
-3.34933668e-01 -4.55007017e-01 -3.30879331e-01 7.18902528e-01
-1.57980072e+00 -8.79482567e-01 -1.78235903e-01 1.82324991e-01
8.14389884e-01 -1.33917928e-01 6.82007611e-01 5.30538522e-02
-5.98127067e-01 6.35972440e-01 1.26027867e-01 -1.31918281e-01
7.98130751e-01 -1.34012794e+00 5.18759787e-01 8.95446479e-01
2.18222693e-01 2.42637366e-01 4.64981705e-01 -8.15977752e-01
-1.02099144e+00 -1.19905543e+00 7.27764070e-01 -4.00503874e-01
3.38908762e-01 -3.53859514e-01 -1.33953655e+00 7.27193117e-01
-1.10714361e-01 1.37724265e-01 5.73847532e-01 1.72700137e-01
-5.03522813e-01 -1.45923585e-01 -1.06384838e+00 5.87991118e-01
1.16997087e+00 -2.85453558e-01 -7.20568478e-01 1.06850684e-01
1.08815098e+00 -4.91723865e-01 -3.29725385e-01 5.47412932e-01
4.43747371e-01 -7.93782592e-01 7.21013248e-01 -7.36665308e-01
2.42577538e-01 -3.73314582e-02 -1.69052064e-01 -1.40335858e+00
-4.86434735e-02 -3.96263331e-01 -1.62145913e-01 1.27969515e+00
6.61857128e-01 -5.36220968e-01 1.18094826e+00 8.94054234e-01
-3.81282568e-01 -8.41701090e-01 -1.08327949e+00 -7.85546124e-01
1.85973182e-01 -5.57468534e-01 3.29262763e-01 9.42869067e-01
-4.39241678e-01 6.06308937e-01 -2.12945879e-01 1.51281722e-03
6.78071916e-01 6.13756180e-02 4.71601486e-01 -1.23932588e+00
-3.10516596e-01 -6.05068244e-02 1.15676902e-01 -1.12228072e+00
5.58250189e-01 -9.18187022e-01 3.73342037e-01 -1.34128690e+00
3.71474087e-01 -7.19786167e-01 -6.34562373e-01 9.66354251e-01
-6.13457143e-01 8.21672156e-02 3.92404109e-01 1.29208550e-01
-8.85454237e-01 6.31166279e-01 1.26031697e+00 -2.34277412e-01
-4.70315397e-01 -6.02227338e-02 -8.83469820e-01 8.44197452e-01
6.88182294e-01 -7.82341778e-01 -6.00286365e-01 -5.37649632e-01
-1.10886700e-01 -2.14730233e-01 1.79136515e-01 -1.08856094e+00
2.40289241e-01 -3.02195288e-02 3.99677724e-01 -6.51621222e-01
4.15930569e-01 -6.58729911e-01 -4.78270799e-02 5.49777091e-01
-4.08844620e-01 -5.21183848e-01 4.34713930e-01 8.21524322e-01
-2.47303620e-02 -4.83275235e-01 8.59749317e-01 -2.49608025e-01
-8.50958765e-01 2.68600583e-01 -1.79378204e-02 4.63241398e-01
9.16370392e-01 -4.12426470e-03 -3.70063543e-01 -2.28323415e-02
-9.23603773e-01 4.65886295e-01 4.13134426e-01 4.57990378e-01
4.24297214e-01 -1.07675540e+00 -4.58133280e-01 7.29700923e-02
1.15329854e-03 5.21384120e-01 4.44973260e-01 5.50402462e-01
1.01148374e-01 1.87410325e-01 -8.08609352e-02 -6.69419706e-01
-1.11751020e+00 6.49499834e-01 2.09227920e-01 -5.64479649e-01
-2.19586000e-01 1.04941583e+00 3.56759489e-01 -3.30249965e-01
3.67301255e-01 -3.31834704e-01 9.58190393e-03 2.77071416e-01
3.01988125e-01 3.45602304e-01 2.14139909e-01 -3.13532025e-01
-3.17834496e-01 3.01890582e-01 -7.68262565e-01 -2.42477953e-01
1.29816902e+00 -1.82928056e-01 1.73045292e-01 6.99343503e-01
1.02185965e+00 -3.08901578e-01 -1.83083332e+00 -7.29467273e-01
9.56184976e-03 -2.45285287e-01 -3.01973950e-02 -1.12381661e+00
-1.12152684e+00 9.15169120e-01 5.95410287e-01 -4.00847256e-01
1.15792727e+00 1.32448435e-01 8.54361176e-01 5.58535099e-01
4.40830410e-01 -1.36510170e+00 3.26995343e-01 6.49800181e-01
4.05345052e-01 -1.40366054e+00 -1.24919936e-01 -3.39840680e-01
-8.59740078e-01 6.81221306e-01 8.64599764e-01 2.81204760e-01
5.65184534e-01 1.95761230e-02 -6.38283491e-02 1.35761678e-01
-9.93826210e-01 -1.51037514e-01 1.84518620e-01 3.54244530e-01
1.84530411e-02 5.58806099e-02 1.92756318e-02 9.42584515e-01
-1.63916759e-02 2.40552239e-02 3.73039961e-01 1.10042942e+00
-5.72091281e-01 -1.14021254e+00 3.54454946e-03 5.70501983e-01
-4.82845843e-01 -2.91736782e-01 -3.28597248e-01 5.61461270e-01
4.58850592e-01 8.70011449e-01 7.20472708e-02 -3.52481425e-01
1.89362511e-01 5.31766951e-01 4.74938840e-01 -8.99450779e-01
-2.13998541e-01 -1.85897313e-02 5.73177710e-02 -4.39815134e-01
-3.62859011e-01 -6.50449038e-01 -1.46308780e+00 1.84639871e-01
-4.04295415e-01 1.53327748e-01 5.10771036e-01 1.23891687e+00
4.65309769e-01 6.70248866e-01 4.19872403e-01 -6.90145373e-01
-5.96818626e-01 -9.41550910e-01 -3.08929831e-01 4.17671174e-01
3.06641102e-01 -8.08509409e-01 -3.89520496e-01 2.09438786e-01]
|
[9.450080871582031, 2.2746169567108154]
|
d8f9918a-efbc-4317-b307-774f1dee114e
|
sgbanet-semantic-gan-and-balanced-attention
|
2207.10256
| null |
https://arxiv.org/abs/2207.10256v1
|
https://arxiv.org/pdf/2207.10256v1.pdf
|
SGBANet: Semantic GAN and Balanced Attention Network for Arbitrarily Oriented Scene Text Recognition
|
Scene text recognition is a challenging task due to the complex backgrounds and diverse variations of text instances. In this paper, we propose a novel Semantic GAN and Balanced Attention Network (SGBANet) to recognize the texts in scene images. The proposed method first generates the simple semantic feature using Semantic GAN and then recognizes the scene text with the Balanced Attention Module. The Semantic GAN aims to align the semantic feature distribution between the support domain and target domain. Different from the conventional image-to-image translation methods that perform at the image level, the Semantic GAN performs the generation and discrimination on the semantic level with the Semantic Generator Module (SGM) and Semantic Discriminator Module (SDM). For target images (scene text images), the Semantic Generator Module generates simple semantic features that share the same feature distribution with support images (clear text images). The Semantic Discriminator Module is used to distinguish the semantic features between the support domain and target domain. In addition, a Balanced Attention Module is designed to alleviate the problem of attention drift. The Balanced Attention Module first learns a balancing parameter based on the visual glimpse vector and semantic glimpse vector, and then performs the balancing operation for obtaining a balanced glimpse vector. Experiments on six benchmarks, including regular datasets, i.e., IIIT5K, SVT, ICDAR2013, and irregular datasets, i.e., ICDAR2015, SVTP, CUTE80, validate the effectiveness of our proposed method.
|
['Yue Lu', 'Umapada Pal', 'Jiajia Wu', 'Bing Yin', 'Palaiahnakote Shivakumara', 'Shujing Lyu', 'Dajian Zhong']
|
2022-07-21
| null | null | null | null |
['scene-text-recognition']
|
['computer-vision']
|
[ 6.08381689e-01 -2.89397240e-01 7.52999410e-02 -4.31949615e-01
-5.67662179e-01 -4.05039281e-01 7.56510735e-01 -5.87645292e-01
-5.56072965e-02 3.39745611e-01 2.34626889e-01 1.19312651e-01
2.63323277e-01 -7.51813352e-01 -6.72660112e-01 -1.06402278e+00
1.10942459e+00 6.41315162e-01 -1.32885734e-02 -4.57089320e-02
2.44107336e-01 -6.17192537e-02 -1.33468091e+00 5.90369284e-01
1.02735424e+00 1.22808075e+00 6.18740559e-01 4.65646386e-01
-7.07458019e-01 8.86830568e-01 -8.97120297e-01 -2.94440091e-01
2.19143912e-01 -1.11911619e+00 -4.86336797e-01 4.12959665e-01
4.16758835e-01 -3.66649069e-02 -3.28736812e-01 1.27565694e+00
5.68876266e-01 5.20358086e-02 7.88422644e-01 -1.59003699e+00
-9.48574126e-01 4.57732439e-01 -7.25134969e-01 -4.71241213e-02
2.10404545e-01 3.22193325e-01 6.64137244e-01 -9.35976326e-01
4.12880778e-01 1.32303846e+00 3.34050477e-01 6.80125952e-01
-9.29519832e-01 -8.63141239e-01 2.58728445e-01 3.47504944e-01
-1.34526479e+00 -1.60678908e-01 9.85887170e-01 -4.99842584e-01
4.15547639e-01 2.16602340e-01 4.91672218e-01 1.55806756e+00
2.13443696e-01 1.04884887e+00 1.01420593e+00 -2.27664486e-01
2.52481133e-01 1.63665697e-01 -2.31919643e-02 4.26151037e-01
2.46543199e-01 -3.57763499e-01 -5.25812984e-01 3.43771040e-01
6.75757766e-01 4.03376400e-01 -4.41280454e-01 -2.31857151e-01
-1.20161152e+00 8.91674995e-01 3.46358091e-01 3.72747093e-01
-2.61657894e-01 2.31745541e-02 4.61903036e-01 1.04662828e-01
2.64472067e-01 -8.29604343e-02 -2.04940587e-01 8.39457586e-02
-8.35238695e-01 -1.23395018e-01 5.02805531e-01 1.23138666e+00
6.73848569e-01 4.24400091e-01 -6.17374063e-01 8.46040249e-01
3.41798037e-01 1.04005289e+00 1.10463405e+00 -3.31936836e-01
7.68335223e-01 8.93011034e-01 -2.59573877e-01 -1.15045333e+00
1.66850649e-02 -3.85271043e-01 -1.25118387e+00 -1.43907174e-01
1.26504853e-01 4.33082283e-02 -1.47281802e+00 1.56733429e+00
6.39485940e-02 1.00646079e-01 5.45170665e-01 1.01890743e+00
1.22530127e+00 9.48718369e-01 4.95085828e-02 1.42870232e-01
1.31712413e+00 -1.24794710e+00 -8.26997340e-01 -6.32420599e-01
4.83000100e-01 -6.80586815e-01 1.45481205e+00 -7.80430436e-02
-6.98245287e-01 -6.84499681e-01 -1.04851484e+00 -5.45755513e-02
-3.83510888e-01 4.90951091e-01 8.48260596e-02 3.81403059e-01
-6.86764657e-01 -1.41461864e-01 -4.44595277e-01 -4.81430203e-01
5.69843352e-01 -4.43878882e-02 -1.55138418e-01 -2.33044788e-01
-1.06064355e+00 5.33407569e-01 5.44793487e-01 -3.59047041e-03
-1.01671815e+00 -5.03242254e-01 -1.07636070e+00 2.11974725e-01
1.36048943e-01 -8.38500500e-01 9.58669722e-01 -1.95556235e+00
-1.49906504e+00 1.04124403e+00 -4.19331715e-02 -2.23021284e-01
6.69352949e-01 1.90936118e-01 -4.44490731e-01 9.38993022e-02
4.91490483e-01 5.57297885e-01 1.30021358e+00 -1.26065898e+00
-6.22657776e-01 -5.04172266e-01 -4.42488462e-01 6.04185820e-01
-1.60854310e-01 -1.84552491e-01 -4.34568703e-01 -9.82981443e-01
1.11444630e-01 -6.09105051e-01 1.90651640e-01 -4.64760363e-01
-6.27677083e-01 6.78910390e-02 1.26069248e+00 -8.05890858e-01
7.65905082e-01 -2.50881648e+00 2.81798571e-01 5.29446676e-02
-1.59448832e-02 1.44919977e-01 -2.45818079e-01 6.32771403e-02
-1.94898367e-01 -1.43068388e-01 -4.26679045e-01 -2.08245575e-01
-5.03392182e-02 1.59319580e-01 -4.60597187e-01 2.18316555e-01
4.58799638e-02 1.24741220e+00 -7.11890101e-01 -4.40338850e-01
3.28973264e-01 3.26747805e-01 -1.33304760e-01 4.32597131e-01
-3.29595357e-01 4.41217810e-01 -6.64167404e-01 4.68834192e-01
6.61414623e-01 -3.89195055e-01 -9.70031172e-02 -2.26645142e-01
3.59703809e-01 -6.52129203e-03 -1.06137776e+00 1.66405582e+00
-3.46762985e-01 5.44871390e-01 -1.96906492e-01 -1.30880725e+00
1.28945816e+00 5.49055748e-02 9.22431946e-02 -1.06616688e+00
4.37821895e-01 1.00600794e-01 -2.43442014e-01 -4.49192166e-01
3.42576623e-01 -1.63759068e-01 -3.26831490e-01 4.06583935e-01
7.90111348e-02 -3.24590951e-01 -1.74326658e-01 1.69590369e-01
5.81805229e-01 1.57927945e-01 3.78807522e-02 -2.80066431e-01
7.02068567e-01 1.69899076e-01 4.77932960e-01 5.12388647e-01
1.14269599e-01 9.95953977e-01 4.08162713e-01 -4.28756058e-01
-9.38682318e-01 -1.09847343e+00 2.68503666e-01 9.35358346e-01
5.75158834e-01 -6.25118688e-02 -1.12877154e+00 -1.00578821e+00
-2.04431012e-01 9.31167126e-01 -7.80643702e-01 -5.91574669e-01
-1.80153906e-01 -6.54189765e-01 3.00226510e-01 7.16508687e-01
1.25501776e+00 -1.30189407e+00 -4.21540171e-01 -1.54308200e-01
-3.09582800e-01 -1.08808446e+00 -8.41604531e-01 -2.58181617e-02
-3.92048508e-01 -9.38144624e-01 -8.05328369e-01 -1.07995105e+00
9.31472003e-01 4.83410984e-01 8.46855462e-01 -1.73412964e-01
-1.47458315e-01 3.52950960e-01 -3.89121115e-01 -3.63269657e-01
-4.04361755e-01 7.08384858e-03 -4.34084743e-01 5.26796043e-01
4.47239786e-01 -1.23911291e-01 -5.49699783e-01 4.07625496e-01
-1.07382643e+00 7.31771052e-01 5.01888216e-01 1.09159672e+00
6.70274913e-01 1.30921617e-01 4.37172920e-01 -9.03985262e-01
3.31989616e-01 -4.74441290e-01 -4.30040330e-01 3.11302930e-01
-3.75586092e-01 -1.91804558e-01 9.86256838e-01 -3.61696243e-01
-1.23315787e+00 5.45390286e-02 1.18017733e-01 -5.71089804e-01
-1.86519206e-01 1.79787561e-01 -8.56855810e-01 3.31238866e-01
3.87687773e-01 9.44423974e-01 -1.12130485e-01 -1.29326299e-01
2.47204363e-01 9.04217482e-01 6.73503101e-01 -3.95119041e-01
7.51596570e-01 4.73984867e-01 -4.53084648e-01 -6.75293982e-01
-9.72079515e-01 -5.28933145e-02 -2.92830855e-01 -3.54584306e-02
1.28289568e+00 -1.06281769e+00 -1.42852902e-01 1.08344734e+00
-9.87705946e-01 -5.74364901e-01 -5.02923727e-01 1.74414396e-01
-7.47081637e-01 1.38037518e-01 -2.46729761e-01 -2.98766345e-01
-6.20349050e-01 -1.13215768e+00 1.46312571e+00 5.62998950e-01
2.02575639e-01 -9.51027751e-01 -3.69776219e-01 4.46323812e-01
2.86720306e-01 2.58810580e-01 9.52482343e-01 -7.86535859e-01
-5.04145980e-01 -2.88643260e-02 -5.30395687e-01 4.55338389e-01
3.19258273e-01 -2.27240577e-01 -8.71067941e-01 -2.29192242e-01
2.85426885e-01 -2.44311348e-01 8.40073884e-01 3.57631654e-01
1.24053633e+00 -3.25869262e-01 -1.45274013e-01 9.73640561e-01
1.38730979e+00 5.21299362e-01 8.74866426e-01 4.25633758e-01
1.11768568e+00 3.38593751e-01 5.49822092e-01 -1.50635755e-02
3.87910396e-01 5.18360198e-01 2.59777755e-01 -3.95035833e-01
-4.00502443e-01 -5.94874024e-01 4.68484104e-01 6.97730362e-01
5.99815845e-01 -4.81321365e-01 -9.01031971e-01 3.50698024e-01
-1.83787894e+00 -8.52050960e-01 8.08204189e-02 1.93580508e+00
5.64429402e-01 -2.76267491e-02 -3.05555016e-01 -9.64169670e-03
1.15520978e+00 1.90187275e-01 -7.53392577e-01 -1.87413126e-01
-5.83851755e-01 -1.15531176e-01 3.60545725e-01 1.18951626e-01
-8.28620911e-01 1.16502559e+00 4.74353647e+00 1.09499872e+00
-1.41609037e+00 1.30108923e-01 8.98285031e-01 1.84108287e-01
-2.94704467e-01 -2.49857858e-01 -7.77700543e-01 1.02695966e+00
3.29941183e-01 -4.07328248e-01 4.13933843e-01 9.72504675e-01
-2.61280835e-02 2.44199216e-01 -9.13723052e-01 1.31006968e+00
6.68571591e-01 -1.14025128e+00 5.85512340e-01 -3.75662386e-01
1.00894463e+00 -1.82317942e-01 2.63029784e-01 2.37638772e-01
3.17392498e-01 -1.10946226e+00 8.48982930e-01 3.79187196e-01
1.15176880e+00 -6.40541732e-01 8.66785645e-01 2.56872058e-01
-1.15496492e+00 -4.94510233e-02 -3.82298321e-01 4.42080498e-01
-1.78898126e-01 3.84704858e-01 -6.63460433e-01 6.98040128e-01
7.27066398e-01 1.13673985e+00 -5.89041531e-01 3.84483308e-01
-4.89689887e-01 4.34605151e-01 1.07729092e-01 6.16595475e-03
3.77447098e-01 -4.51229185e-01 4.04331058e-01 1.12978494e+00
4.04581070e-01 -3.18835109e-01 1.28994003e-01 9.97785091e-01
-1.74505934e-01 2.65637994e-01 -5.47482252e-01 -6.90205842e-02
2.55125254e-01 1.13556516e+00 -7.25862265e-01 -6.60453856e-01
-3.17698717e-01 1.53598356e+00 -5.48253506e-02 6.03217006e-01
-1.01302516e+00 -4.58595455e-01 4.33704346e-01 -1.55864686e-01
4.74863499e-01 2.82143950e-01 -4.99049574e-01 -1.39697242e+00
1.45313382e-01 -1.05294144e+00 5.11625290e-01 -1.24635887e+00
-1.32936227e+00 5.60082197e-01 -1.93696305e-01 -1.06317425e+00
-2.29321159e-02 -4.44641203e-01 -1.03618741e+00 1.14291918e+00
-1.23439598e+00 -1.42522585e+00 -9.31929469e-01 1.04387331e+00
1.09592426e+00 -4.47286069e-01 4.06870425e-01 -1.62569135e-02
-7.00699329e-01 6.90063119e-01 3.16959888e-01 4.56951737e-01
5.26225746e-01 -1.00230908e+00 4.91836399e-01 7.92455614e-01
-5.52624725e-02 2.37273127e-02 3.31513792e-01 -7.41543233e-01
-1.26220036e+00 -1.67374194e+00 3.44632626e-01 -1.89281732e-01
3.30869257e-01 -5.90369225e-01 -8.86237741e-01 8.35584104e-01
2.26371437e-01 -2.39794642e-01 2.76624680e-01 -7.39467382e-01
-3.91813457e-01 -1.75772488e-01 -1.16383886e+00 5.87868869e-01
9.73501861e-01 -3.71181726e-01 -6.64877295e-01 3.07028979e-01
7.52648711e-01 -5.87558687e-01 -2.44285360e-01 3.21938783e-01
2.35828996e-01 -7.03442276e-01 6.74727321e-01 -3.37423861e-01
6.50282681e-01 -4.56169963e-01 -3.24925393e-01 -1.35779274e+00
-2.24948436e-01 -9.25034881e-02 3.19024324e-01 1.47018981e+00
2.81677302e-02 -9.27690446e-01 7.07943678e-01 4.19153981e-02
-2.57018656e-01 -3.48075807e-01 -8.36622894e-01 -7.86240280e-01
-7.75637627e-02 -6.19220734e-02 8.60227883e-01 1.11235476e+00
-7.74081886e-01 6.34329319e-01 -2.40565330e-01 -1.10866964e-01
4.74791974e-01 4.54623491e-01 8.86029005e-01 -8.67255867e-01
-1.30435720e-01 -5.89987338e-01 -5.92988551e-01 -9.96159017e-01
3.54821622e-01 -1.13795519e+00 9.75546613e-02 -1.59904695e+00
5.17324090e-01 -1.85132980e-01 -1.47586048e-01 3.82304907e-01
-2.98325598e-01 2.33688712e-01 2.66041219e-01 1.48320466e-01
-5.71256638e-01 8.68739545e-01 1.33517337e+00 -6.29265666e-01
4.89519835e-02 -3.44032675e-01 -9.35593426e-01 6.01696670e-01
9.33374405e-01 -2.93463379e-01 -5.90729773e-01 -6.72447383e-01
-1.14571586e-01 -2.96805352e-01 5.77183843e-01 -8.79329681e-01
2.03393716e-02 -2.64796227e-01 8.15736651e-01 -6.43029332e-01
5.36774732e-02 -8.25794816e-01 1.86755985e-01 4.64482069e-01
-3.15898836e-01 -1.55187428e-01 7.54876286e-02 5.19999325e-01
-2.83304036e-01 -9.09686610e-02 9.84889090e-01 -5.66835366e-02
-7.47934818e-01 4.16172028e-01 -3.60070430e-02 4.62260336e-01
1.14818931e+00 -5.24894297e-01 -5.16396344e-01 -3.83238941e-01
-3.32893938e-01 4.29013193e-01 7.28313804e-01 7.99789250e-01
8.37899446e-01 -1.57882512e+00 -8.51049662e-01 6.48110867e-01
3.42522889e-01 3.32697153e-01 6.04628861e-01 5.23635805e-01
-3.96317780e-01 1.48054048e-01 -2.69550115e-01 -9.52395082e-01
-1.09772491e+00 5.64823568e-01 6.00219548e-01 -9.72392634e-02
-7.79437304e-01 6.07660651e-01 1.02215517e+00 -2.29098856e-01
1.89698227e-02 -4.96168546e-02 -1.00171812e-01 -1.56090453e-01
5.14448643e-01 -2.97989044e-03 -7.35424682e-02 -8.63080442e-01
-1.80608317e-01 8.51336718e-01 -2.75324639e-02 2.28149667e-01
9.86617982e-01 -1.82717130e-01 -2.36778352e-02 3.64937872e-01
1.16106689e+00 -3.02302510e-01 -1.22600651e+00 -2.91194409e-01
-3.87387782e-01 -5.45176983e-01 -2.28979781e-01 -8.69364202e-01
-1.35698700e+00 9.61878598e-01 6.05060756e-01 1.12636052e-02
1.45540929e+00 1.30773544e-01 8.84784758e-01 -7.34454468e-02
1.48955397e-02 -1.04737854e+00 3.93284500e-01 4.68627214e-01
9.75507200e-01 -1.10608554e+00 -5.13573170e-01 -2.66603351e-01
-1.10073388e+00 8.34125757e-01 1.04460728e+00 -9.67643186e-02
3.54374908e-02 7.83145204e-02 2.22016543e-01 -1.76948369e-01
-3.62053901e-01 -2.92930417e-02 1.69290617e-01 6.94917262e-01
2.93611083e-02 1.46429718e-01 6.31285682e-02 8.96463096e-01
-2.56128788e-01 -3.77504200e-01 2.99942762e-01 5.86267531e-01
-3.18652987e-01 -6.73922718e-01 -4.12811369e-01 4.73430872e-01
-1.81216985e-01 -2.59197176e-01 -6.12437427e-01 6.05077088e-01
9.31596532e-02 7.63532281e-01 4.04871047e-01 -4.65823025e-01
2.46763527e-01 1.20732270e-01 2.07639083e-01 -5.95363021e-01
-4.30566877e-01 -8.68009031e-02 -4.09079015e-01 -3.39510739e-01
-1.10717647e-01 -4.60781842e-01 -1.41752219e+00 -3.73446532e-02
-4.30372953e-02 1.51723251e-01 6.70379519e-01 9.19078350e-01
3.45763713e-01 7.29703009e-01 9.40547526e-01 -4.70020562e-01
-2.99528152e-01 -8.82411957e-01 -5.28532922e-01 8.85009766e-01
4.43796888e-02 -4.60958660e-01 -4.00935948e-01 3.14683676e-01]
|
[11.391117095947266, -0.038199156522750854]
|
f45b0f73-36cf-47d5-a2f4-cb7c296661c9
|
dudonet-encoding-mask-projection-to-reduce-ct
|
2001.00340
| null |
https://arxiv.org/abs/2001.00340v3
|
https://arxiv.org/pdf/2001.00340v3.pdf
|
Encoding Metal Mask Projection for Metal Artifact Reduction in Computed Tomography
|
Metal artifact reduction (MAR) in computed tomography (CT) is a notoriously challenging task because the artifacts are structured and non-local in the image domain. However, they are inherently local in the sinogram domain. Thus, one possible approach to MAR is to exploit the latter characteristic by learning to reduce artifacts in the sinogram. However, if we directly treat the metal-affected regions in sinogram as missing and replace them with the surrogate data generated by a neural network, the artifact-reduced CT images tend to be over-smoothed and distorted since fine-grained details within the metal-affected regions are completely ignored. In this work, we provide analytical investigation to the issue and propose to address the problem by (1) retaining the metal-affected regions in sinogram and (2) replacing the binarized metal trace with the metal mask projection such that the geometry information of metal implants is encoded. Extensive experiments on simulated datasets and expert evaluations on clinical images demonstrate that our novel network yields anatomically more precise artifact-reduced images than the state-of-the-art approaches, especially when metallic objects are large.
|
['Jing-Jing Lu', 'S. Kevin Zhou', 'Wei-An Lin', 'Yuanyuan Lyu', 'Haofu Liao']
|
2020-01-02
| null | null | null | null |
['metal-artifact-reduction']
|
['medical']
|
[ 5.04159808e-01 1.99001372e-01 2.73595512e-01 -1.06066957e-01
-7.28507757e-01 -1.46164328e-01 6.56606853e-02 -1.27121478e-01
-2.67282158e-01 8.94115090e-01 2.73990154e-01 3.18283811e-02
-4.02283520e-01 -6.00242376e-01 -6.66444600e-01 -1.02032113e+00
-6.33987263e-02 2.76833326e-01 4.52370137e-01 6.92768991e-02
2.34463029e-02 5.81351578e-01 -8.07976663e-01 3.23513508e-01
1.18368042e+00 9.99649107e-01 3.43326122e-01 2.46961461e-03
1.68318674e-01 8.53578508e-01 -3.76517951e-01 6.80642128e-02
4.87583548e-01 -7.02055275e-01 -6.05918825e-01 9.31885391e-02
2.65716255e-01 -1.45590708e-01 -4.12896574e-01 1.29493642e+00
4.10021454e-01 -1.42666847e-01 6.35632455e-01 -4.66677040e-01
-5.05510211e-01 5.67474484e-01 -7.65308619e-01 3.33380818e-01
-2.33378842e-01 -2.16083117e-02 4.30615097e-02 -9.51113343e-01
6.39995754e-01 6.49983883e-01 8.23399723e-01 4.73450363e-01
-1.21665943e+00 -5.24643600e-01 -1.35940686e-01 1.00406833e-01
-1.44400024e+00 -1.56204402e-01 1.11165679e+00 -4.59908307e-01
3.17477286e-01 5.92079222e-01 6.57238007e-01 5.65180659e-01
8.65802526e-01 3.90829176e-01 1.12602663e+00 -4.34797198e-01
2.93032587e-01 1.03061542e-01 -6.12077415e-02 6.30531132e-01
4.58539814e-01 1.94687366e-01 -6.96657076e-02 -3.87138575e-01
1.23849750e+00 3.77790213e-01 -8.20340514e-01 -6.28901184e-01
-1.13257730e+00 6.00314498e-01 6.60745084e-01 6.79331362e-01
-8.17799449e-01 1.14784583e-01 2.32389018e-01 -2.19487980e-01
5.90602398e-01 5.72950244e-01 -1.78622361e-02 3.66300434e-01
-1.06456029e+00 1.54281363e-01 3.44093233e-01 5.65420628e-01
3.49120557e-01 3.38367581e-01 -3.52525145e-01 8.04633796e-01
1.64175913e-01 2.64106929e-01 7.89318383e-01 -6.34755552e-01
2.89439708e-01 4.02217895e-01 1.81264043e-01 -8.86774242e-01
-4.62631822e-01 -6.90142930e-01 -1.20115829e+00 4.62605149e-01
5.31502426e-01 -2.06597343e-01 -1.19145608e+00 1.38164163e+00
4.60724741e-01 1.65424854e-01 -3.69324803e-01 1.23347569e+00
5.45823038e-01 3.56595278e-01 9.48841125e-02 -6.24464989e-01
1.28490353e+00 -8.28477204e-01 -1.08701313e+00 -3.10767531e-01
3.89232129e-01 -7.31749296e-01 6.17970765e-01 4.41711932e-01
-1.47579408e+00 -3.87816131e-01 -1.16047144e+00 2.95036018e-01
2.18783915e-01 2.36071721e-02 4.06685501e-01 4.26400393e-01
-8.06989729e-01 9.70814288e-01 -1.12148571e+00 2.22215459e-01
5.19348562e-01 2.99301207e-01 -1.28318563e-01 1.50470650e-02
-9.86524820e-01 1.06603336e+00 3.19602415e-02 2.76603401e-01
-6.48085177e-01 -9.76258099e-01 -7.25800633e-01 -1.55600905e-01
5.21603346e-01 -5.15179753e-01 1.09381771e+00 -1.11626077e+00
-1.22876632e+00 3.31454158e-01 -1.13290958e-01 -1.92723408e-01
8.25948775e-01 -9.12398994e-02 -3.72915447e-01 3.11153263e-01
8.62055644e-02 -5.83491623e-02 9.26447093e-01 -1.56778932e+00
-1.17108256e-01 -5.62249064e-01 -4.82015729e-01 5.39458469e-02
-2.44127084e-02 -1.47711709e-01 -2.02740103e-01 -1.15517175e+00
8.44139040e-01 -9.07641232e-01 -6.77214622e-01 4.16371107e-01
-3.14944148e-01 4.21860874e-01 6.09312356e-01 -1.11950469e+00
1.22068954e+00 -2.26498318e+00 -1.71969384e-01 3.79782945e-01
4.83022690e-01 1.05555139e-01 2.81545579e-01 -1.25298962e-01
-4.47019696e-01 -2.42347851e-01 -8.02909195e-01 -1.32725425e-02
-7.69865990e-01 5.21000028e-02 1.44577622e-01 8.34269702e-01
-1.28638878e-01 7.62794197e-01 -8.65526080e-01 -5.63949823e-01
3.62508655e-01 5.73530555e-01 -3.15447539e-01 -6.15502223e-02
2.28416204e-01 9.32175636e-01 -7.92808771e-01 4.11056042e-01
1.13877940e+00 -3.16373140e-01 1.59122169e-01 -4.20463771e-01
-1.90832749e-01 2.82025393e-02 -1.13721037e+00 1.72285450e+00
-5.63816786e-01 2.07677618e-01 2.00689420e-01 -7.86149621e-01
5.11049747e-01 6.33370221e-01 1.09276414e+00 -8.30356896e-01
6.55709580e-02 4.52420503e-01 2.38630310e-01 -7.45396256e-01
6.44669756e-02 -6.38336241e-01 4.74725544e-01 3.47227663e-01
-5.17359495e-01 -2.10870862e-01 -3.96545291e-01 -2.45065793e-01
1.05081439e+00 -1.55470431e-01 3.06912482e-01 -5.26681244e-01
4.21601593e-01 7.98098394e-04 6.69689059e-01 6.07381999e-01
-2.21253052e-01 1.08257723e+00 -2.58983094e-02 -6.80141926e-01
-1.03007138e+00 -1.23283672e+00 -5.82651556e-01 -1.18973747e-01
3.25545937e-01 1.43327907e-01 -9.33591664e-01 -6.73714995e-01
-3.74844968e-01 6.39371693e-01 -6.72372460e-01 -2.78867722e-01
-1.07358360e+00 -9.30516958e-01 1.24005834e-02 7.38907993e-01
4.06463295e-01 -1.02637684e+00 -7.41743445e-01 6.00716531e-01
-3.30606788e-01 -7.96245575e-01 -5.54743946e-01 3.52127075e-01
-1.20824981e+00 -9.45690334e-01 -1.15989280e+00 -6.97850108e-01
1.04741633e+00 1.92478299e-01 9.48937416e-01 1.97965160e-01
-4.82840449e-01 -2.71841466e-01 -1.79273918e-01 -1.82594001e-01
-4.01494443e-01 -7.52063453e-01 -2.45980278e-01 4.75018211e-02
-2.08020553e-01 -6.71839237e-01 -9.19463992e-01 3.78746092e-01
-1.08890021e+00 2.30217084e-01 6.82075739e-01 9.68026400e-01
7.64433205e-01 3.59199166e-01 4.54094946e-01 -9.33121741e-01
3.62277478e-01 -2.45678276e-01 -5.04183352e-01 2.00004250e-01
-5.35838127e-01 1.22692391e-01 6.46204770e-01 -4.33942974e-01
-1.32600462e+00 3.72525334e-01 -5.20175584e-02 -3.44365537e-01
1.67236179e-02 2.41909623e-01 9.19828638e-02 -5.37266850e-01
8.45802188e-01 3.94583821e-01 -4.64763381e-02 -4.06668246e-01
-2.30800539e-01 2.78731436e-01 4.44641024e-01 -3.48544687e-01
7.03789175e-01 8.81232500e-01 2.31070295e-01 -6.36157334e-01
-6.14446521e-01 -1.94141984e-01 -5.59848428e-01 -1.78561553e-01
9.22735333e-01 -4.85537738e-01 -2.44889274e-01 2.02879488e-01
-1.15721738e+00 -8.93176347e-02 -5.98219097e-01 6.97845995e-01
-4.34180021e-01 6.64663017e-01 -7.22148240e-01 -7.12235391e-01
-5.13569653e-01 -1.54712415e+00 7.58782268e-01 1.34986788e-01
-1.52093515e-01 -7.07509518e-01 5.43665998e-02 2.58181896e-02
6.11440361e-01 4.19037759e-01 1.15724111e+00 -2.94154942e-01
-6.03710055e-01 -1.90024227e-01 -2.19062492e-01 4.04646307e-01
4.92559582e-01 -6.36110783e-01 -8.38522553e-01 -6.59054294e-02
8.52716863e-01 2.43261248e-01 5.61326444e-01 9.84726131e-01
1.55552244e+00 -1.99118704e-01 -3.71484905e-01 6.93512619e-01
1.55881226e+00 4.90497380e-01 6.91267550e-01 2.04324812e-01
5.16075969e-01 4.19534951e-01 3.68370354e-01 3.62442940e-01
-1.30216733e-01 7.71433890e-01 5.06716073e-01 -3.64405125e-01
-4.86467481e-01 1.02687456e-01 -3.33880007e-01 5.58196127e-01
-3.28187555e-01 6.32008240e-02 -6.94734693e-01 6.02844119e-01
-1.79426634e+00 -5.93814671e-01 -5.58626652e-01 2.26921511e+00
9.43151057e-01 -1.30802587e-01 -4.13142443e-01 5.51590323e-02
6.91368282e-01 -2.48318631e-02 -3.70361656e-01 2.08460316e-01
4.91425879e-02 4.64997500e-01 5.87368667e-01 4.41360772e-01
-9.28529203e-01 2.44179934e-01 6.22577286e+00 8.36397409e-01
-1.20539987e+00 4.40709502e-01 6.20500207e-01 -1.02278376e-02
-2.78773874e-01 -2.01340914e-01 -8.77833068e-02 6.16751254e-01
4.57119763e-01 4.56734784e-02 1.11671813e-01 7.04750478e-01
4.32764709e-01 -3.75336558e-01 -9.21601295e-01 8.86097848e-01
-6.39473870e-02 -1.26051962e+00 -2.47156397e-01 -1.20198481e-01
9.55329180e-01 -1.75679624e-01 -1.16681047e-01 -3.07973266e-01
-1.58091620e-01 -9.35812056e-01 7.60921538e-01 6.69076502e-01
7.72436738e-01 -5.12118876e-01 8.02149355e-01 2.15539917e-01
-8.52193296e-01 1.59143016e-01 -4.16065186e-01 3.27645540e-01
3.10571402e-01 1.04322791e+00 -6.91662848e-01 7.63600588e-01
7.33167350e-01 3.39503139e-01 -3.53793949e-01 1.57468581e+00
-1.26749307e-01 3.15982670e-01 -3.26292753e-01 4.77479696e-01
8.40232897e-05 -2.72142231e-01 5.69370866e-01 8.21939290e-01
4.98935223e-01 3.86045158e-01 -1.14505477e-01 1.16729772e+00
2.01227039e-01 -2.85492893e-02 -4.33292240e-01 6.88010573e-01
-3.43431160e-02 1.08610666e+00 -8.94329607e-01 -4.55574155e-01
-3.99130076e-01 9.96021509e-01 -2.36241430e-01 5.60995996e-01
-7.57460296e-01 1.04302634e-02 9.06547680e-02 6.33661747e-01
-5.28097153e-04 2.63145287e-02 -9.15489256e-01 -9.51762021e-01
3.81158918e-01 -5.15118837e-01 1.94414899e-01 -7.51674116e-01
-1.18433654e+00 9.31124210e-01 -8.37009028e-02 -1.53569698e+00
-6.44536167e-02 -3.34916294e-01 -6.49455547e-01 9.39876556e-01
-1.35978675e+00 -7.78591692e-01 -5.14056563e-01 5.78242779e-01
4.59947944e-01 4.54412103e-01 6.14354372e-01 5.39684296e-01
4.81847227e-02 3.27050507e-01 3.38463515e-01 -3.03694028e-02
5.27475774e-01 -9.66124952e-01 -2.35666126e-01 8.14948440e-01
-2.68235624e-01 4.08686966e-01 6.63874090e-01 -1.01834118e+00
-9.46466506e-01 -1.03518164e+00 5.00500858e-01 -1.80250555e-01
4.48038280e-01 -9.74580422e-02 -1.24546349e+00 5.17395973e-01
1.52976900e-01 5.32083452e-01 2.13380978e-01 -6.73997104e-01
2.99526304e-01 -6.56248815e-03 -1.54909039e+00 5.16307175e-01
6.93284690e-01 -2.31258199e-02 -6.31952345e-01 4.78280604e-01
3.11390698e-01 -6.40299439e-01 -7.74627149e-01 6.82656884e-01
3.36383462e-01 -8.68480504e-01 1.04122663e+00 -1.45418897e-01
4.56972748e-01 -3.04765940e-01 2.77931273e-01 -1.51289225e+00
-4.49422807e-01 -2.44725645e-01 1.65509835e-01 5.48103869e-01
1.98451042e-01 -5.70932686e-01 9.78008330e-01 6.98580146e-01
-6.54523432e-01 -7.34275401e-01 -1.35813189e+00 -7.83614159e-01
2.33226523e-01 -1.00394964e-01 2.86770880e-01 1.03830099e+00
-1.56139389e-01 -4.52326417e-01 -2.81314224e-01 3.25179785e-01
8.75169039e-01 5.41442037e-02 -1.32465258e-01 -1.00690794e+00
-2.52919197e-01 -1.59149781e-01 -9.53102931e-02 -7.20197320e-01
-4.56637532e-01 -6.47619367e-01 5.10909379e-01 -1.47501981e+00
3.73189926e-01 -7.14852035e-01 -3.35617870e-01 3.04149121e-01
-1.99838057e-01 3.21547925e-01 -1.43868655e-01 3.01208407e-01
6.68021366e-02 4.79931295e-01 1.99682486e+00 -1.60947472e-01
-1.44610181e-01 2.30300605e-01 -2.97609568e-01 1.07469904e+00
4.49462682e-01 -7.29682982e-01 -1.60807684e-01 -4.14499551e-01
-2.68779457e-01 3.95233482e-01 5.65352142e-01 -1.14439154e+00
2.46891707e-01 -1.58627868e-01 6.90285087e-01 -5.78321993e-01
3.16044748e-01 -1.34086585e+00 4.56606239e-01 6.99707806e-01
-5.90851307e-02 -2.32971206e-01 1.94976047e-01 2.06323817e-01
-2.14200020e-01 -4.61979210e-01 1.23697877e+00 -4.56398666e-01
-5.51711097e-02 2.23212197e-01 -4.62326318e-01 -2.56726623e-01
9.51809168e-01 -3.22685480e-01 1.08661838e-01 -1.63999632e-01
-1.00255179e+00 -4.30341482e-01 3.03433418e-01 -6.68770596e-02
8.59947681e-01 -1.32467878e+00 -5.72960556e-01 4.06510651e-01
-1.91256568e-01 3.26701611e-01 7.60538280e-01 1.24352133e+00
-7.58351147e-01 3.99512380e-01 -1.47391707e-01 -5.65691113e-01
-8.41049671e-01 6.52684093e-01 6.84815407e-01 -5.30302823e-01
-1.10398889e+00 7.29426026e-01 8.02227318e-01 9.11789685e-02
-5.20717725e-02 -5.25733232e-01 8.92937854e-02 -5.96128881e-01
5.15901268e-01 1.80963367e-01 4.03924435e-01 -3.90994072e-01
-2.45962411e-01 7.10186005e-01 -1.21825576e-01 1.87779844e-01
1.51934898e+00 -8.90031159e-02 -9.45245922e-02 -1.07731164e-01
8.14760029e-01 1.08288191e-01 -1.29112089e+00 -3.08990091e-01
-1.00293398e-01 -6.44658923e-01 3.11349213e-01 -9.35583413e-01
-1.49043214e+00 6.31592155e-01 1.00253141e+00 2.37941574e-02
1.20262325e+00 -8.14372972e-02 7.75832474e-01 -2.94183463e-01
3.81747991e-01 -9.61797535e-01 8.48315135e-02 -5.62882647e-02
1.16064084e+00 -9.46418345e-01 2.76308089e-01 -8.75018120e-01
-4.71622050e-01 1.01018226e+00 5.74050903e-01 -4.14757103e-01
7.51742661e-01 5.11543691e-01 8.56509134e-02 -3.65821272e-01
2.75588557e-02 4.49555695e-01 3.93362552e-01 4.63188499e-01
3.87406796e-01 -6.64102137e-02 -5.64631104e-01 7.35080540e-01
1.55757844e-01 9.72481668e-02 6.26405418e-01 1.21832907e+00
-2.79014051e-01 -7.65149832e-01 -8.63010466e-01 4.13463473e-01
-6.12868011e-01 -2.82999694e-01 1.15642212e-01 6.74000025e-01
2.33453408e-01 6.11368656e-01 -2.46259198e-01 1.74417853e-01
4.47411627e-01 -1.41086414e-01 6.44185185e-01 -5.74731171e-01
-6.36707127e-01 5.77687085e-01 -3.23421359e-01 -5.76044321e-01
-1.32733256e-01 -5.50432801e-01 -1.45506430e+00 2.73442447e-01
-4.11109418e-01 7.50384331e-02 6.99683726e-01 9.17160213e-01
6.89026788e-02 1.08003807e+00 5.08759081e-01 -7.77565300e-01
-5.53067386e-01 -9.36848938e-01 -9.30941939e-01 5.15715122e-01
4.16052401e-01 -7.84197986e-01 -3.26284975e-01 -3.78238969e-02]
|
[13.492403984069824, -2.552842378616333]
|
e6705f8c-3b12-4d5e-b874-cba1006afdc6
|
symphony-generation-with-permutation
|
2205.05448
| null |
https://arxiv.org/abs/2205.05448v2
|
https://arxiv.org/pdf/2205.05448v2.pdf
|
Symphony Generation with Permutation Invariant Language Model
|
In this work, we propose a permutation invariant language model, SymphonyNet, as a solution for symbolic symphony music generation. We propose a novel Multi-track Multi-instrument Repeatable (MMR) representation for symphonic music and model the music sequence using a Transformer-based auto-regressive language model with specific 3-D positional embedding. To overcome length overflow when modeling extra-long symphony tokens, we also propose a modified Byte Pair Encoding algorithm (Music BPE) for music tokens and introduce a novel linear transformer decoder architecture as a backbone. Meanwhile, we train the decoder to learn automatic orchestration as a joint task by masking instrument information from the input. We also introduce a large-scale symbolic symphony dataset for the advance of symphony generation research. Empirical results show that the proposed approach can generate coherent, novel, complex and harmonious symphony as a pioneer solution for multi-track multi-instrument symbolic music generation.
|
['Maosong Sun', 'Feng Yu', 'Xiaobing Li', 'Xinran Zhang', 'Zehua Cheng', 'Yuanliang Dong', 'Jiafeng Liu']
|
2022-05-10
| null | null | null | null |
['audio-generation', 'music-generation', 'music-generation']
|
['audio', 'audio', 'music']
|
[ 3.20558816e-01 -1.40000850e-01 -2.19189599e-02 9.76803973e-02
-1.12861550e+00 -8.54556739e-01 4.52558279e-01 -6.56523824e-01
1.14817962e-01 6.76120400e-01 5.80476582e-01 -3.60702090e-02
-3.29674602e-01 -6.90575123e-01 -7.99419999e-01 -5.25884926e-01
2.30562672e-01 5.08260667e-01 -3.81506115e-01 -5.45439363e-01
2.68544763e-01 1.88626289e-01 -1.37878799e+00 5.20669639e-01
4.62269098e-01 4.72000003e-01 4.18121874e-01 9.88193274e-01
-2.87590865e-02 7.90517569e-01 -8.29959095e-01 -4.77220774e-01
4.85325456e-01 -1.19341731e+00 -4.98567581e-01 -2.64148146e-01
1.29853234e-01 7.94748142e-02 -3.78316581e-01 5.85753500e-01
8.71869981e-01 -9.45247896e-03 6.95053995e-01 -1.02280366e+00
-6.39219999e-01 1.77409792e+00 -2.57731766e-01 -3.42861712e-01
3.11405033e-01 1.25073761e-01 1.49696720e+00 -6.93110347e-01
5.49928188e-01 1.11406672e+00 9.00601566e-01 3.67709339e-01
-1.23143923e+00 -8.58754098e-01 -5.07026911e-01 1.93608016e-01
-1.58406174e+00 -4.36027288e-01 1.28588772e+00 -2.00389057e-01
8.15965295e-01 6.26339972e-01 1.13316011e+00 1.35354388e+00
3.17941427e-01 9.46155846e-01 8.11592162e-01 -4.41435099e-01
-1.11821920e-01 -3.68094802e-01 -5.47829270e-01 5.66184103e-01
-2.38189474e-01 9.35427696e-02 -1.10429239e+00 -1.16599835e-01
1.09996188e+00 -4.31613475e-01 1.44129291e-01 -2.19538137e-02
-1.52519000e+00 7.51298964e-01 9.15753767e-02 6.41188502e-01
-4.02962804e-01 6.38898373e-01 5.60161352e-01 5.07106483e-01
-2.02869341e-01 9.26486135e-01 -2.02171713e-01 -6.32102907e-01
-1.36882341e+00 4.36466068e-01 7.07087874e-01 8.52189660e-01
2.40554541e-01 8.52902770e-01 -5.09403467e-01 9.30128872e-01
9.52916592e-02 5.49159527e-01 9.70061183e-01 -8.19608629e-01
4.74852711e-01 1.84416071e-01 -2.35314786e-01 -8.02490830e-01
2.18474437e-02 -1.03836107e+00 -9.60071564e-01 -3.78497392e-01
-2.42238879e-01 1.96376503e-01 -5.91625512e-01 1.83154309e+00
-4.21496093e-01 6.17620766e-01 5.95244728e-02 6.83698714e-01
6.42279267e-01 9.13924634e-01 -4.73519027e-01 -7.05151781e-02
1.10150063e+00 -1.06965101e+00 -6.32939696e-01 2.21737638e-01
3.62549275e-01 -1.17965424e+00 1.31286085e+00 5.82942247e-01
-1.32555842e+00 -8.07102799e-01 -1.11719286e+00 -1.55122325e-01
6.95606545e-02 5.42564094e-01 4.53159511e-01 4.64505553e-01
-6.32986605e-01 9.02474344e-01 -3.91699642e-01 3.34986240e-01
-2.48589188e-01 1.57475770e-01 1.44907519e-01 5.65464199e-01
-1.23529124e+00 5.49464166e-01 6.27644420e-01 4.73658741e-02
-1.02352929e+00 -7.42873073e-01 -7.32892275e-01 9.50595438e-02
-2.60066211e-01 -9.79072034e-01 1.26846063e+00 -7.48924732e-01
-2.08418465e+00 7.81637132e-01 9.36108530e-02 -6.33538485e-01
2.43095741e-01 -8.13591927e-02 -4.39769328e-01 -1.19239606e-01
5.19820787e-02 4.40240055e-01 1.00299346e+00 -9.49965239e-01
-2.09380686e-01 1.78083420e-01 -5.67827702e-01 2.79131055e-01
-6.98041320e-02 -7.18189403e-02 -2.12641194e-01 -1.50621104e+00
2.21980184e-01 -1.27550924e+00 -1.65085733e-01 -8.82183850e-01
-8.41897905e-01 1.80509955e-01 3.45110536e-01 -7.73737788e-01
1.71298003e+00 -2.10695505e+00 7.16172040e-01 2.77623057e-01
-2.68655688e-01 2.00294014e-02 -4.79743093e-01 7.48797476e-01
-1.06476694e-01 -9.41939577e-02 -4.22587156e-01 -3.76714796e-01
5.25378406e-01 1.55164063e-01 -9.19239759e-01 -1.97462961e-01
-2.56057270e-02 1.35191560e+00 -7.25568831e-01 -1.43106192e-01
-7.41530731e-02 5.54547906e-01 -5.85599363e-01 6.16562180e-02
-2.69559145e-01 7.44053960e-01 -2.05204576e-01 4.14434880e-01
1.04562432e-01 1.89013094e-01 1.30179171e-02 -2.29139596e-01
-2.89023727e-01 6.41969681e-01 -1.19420123e+00 2.48711395e+00
-7.62282431e-01 4.64383394e-01 -5.63565612e-01 -5.63702404e-01
1.34554398e+00 4.43795174e-01 4.84364867e-01 -5.79280853e-01
9.34102088e-02 6.29373312e-01 1.40587285e-01 -2.24752411e-01
1.17357028e+00 -4.94143903e-01 -6.23164117e-01 6.50935650e-01
-4.06925231e-02 -6.22193396e-01 6.53974265e-02 -1.51464343e-01
1.19193709e+00 4.63570893e-01 2.38948137e-01 1.87774539e-01
3.91204685e-01 -2.39695624e-01 4.89707440e-01 6.72868013e-01
6.32749259e-01 9.62547481e-01 1.47748500e-01 -3.29895258e-01
-1.31194651e+00 -1.28430760e+00 3.25361699e-01 1.00422359e+00
-3.22903931e-01 -9.39994633e-01 -3.96246254e-01 3.50326300e-02
-1.77691042e-01 8.59152079e-01 -1.33177325e-01 -3.13633144e-01
-9.78922188e-01 -5.53413153e-01 1.49858463e+00 2.39975259e-01
4.05173659e-01 -1.43604493e+00 -3.57873678e-01 7.92440951e-01
-5.52756429e-01 -7.39325941e-01 -8.19263458e-01 1.15252636e-01
-6.53162718e-01 -5.22530198e-01 -1.01291764e+00 -9.73578036e-01
-2.99891353e-01 -2.70457178e-01 1.13700092e+00 -5.26619256e-01
-5.06578565e-01 5.14989644e-02 -3.55266035e-01 -2.03154743e-01
-7.74766982e-01 3.17523241e-01 1.27492175e-01 1.16431467e-01
-2.03918651e-01 -9.12165463e-01 -2.62438208e-01 2.29458814e-03
-8.83951128e-01 3.78835708e-01 6.95457458e-01 8.34820986e-01
8.87145698e-01 -3.58936675e-02 8.88827562e-01 -4.42312360e-01
9.26036239e-01 -2.39050746e-01 -2.32689291e-01 9.09611732e-02
-1.04699381e-01 3.16443175e-01 8.85738552e-01 -6.43247843e-01
-6.86308682e-01 2.08189338e-02 -2.47030854e-01 -6.47270381e-01
5.26133001e-01 5.51906109e-01 3.05038746e-02 2.49072835e-01
5.20208359e-01 9.97961104e-01 -4.24481243e-01 -7.12632775e-01
6.59290314e-01 7.77098835e-01 8.52715135e-01 -6.44963086e-01
1.11771226e+00 2.42070053e-02 1.44060269e-01 -7.64060080e-01
-5.21205962e-01 -3.27764936e-02 -5.44632494e-01 5.93172945e-02
6.53528452e-01 -1.07348812e+00 -6.77028835e-01 3.82132024e-01
-1.17829013e+00 -3.19512278e-01 -7.63380408e-01 5.24609387e-01
-1.17856252e+00 2.82666892e-01 -6.84994936e-01 -7.18327463e-01
-7.70077825e-01 -9.32213426e-01 1.27909350e+00 -3.08882594e-01
-6.70895219e-01 -5.67052960e-01 6.83616579e-01 1.25729650e-01
3.48954320e-01 3.48219961e-01 8.57080698e-01 -1.42104417e-01
-9.49655592e-01 -1.46764684e-02 4.00126010e-01 3.62606972e-01
9.35383439e-02 -2.03972697e-01 -7.13855803e-01 -1.49489149e-01
-1.89118192e-01 -2.17111051e-01 9.07972693e-01 2.19129831e-01
9.87526774e-01 -3.00735772e-01 3.02699894e-01 1.05562210e+00
1.19174147e+00 -8.79256986e-03 7.46747494e-01 2.78096378e-01
8.45488489e-01 -7.41871353e-03 3.58765781e-01 7.08756924e-01
4.04888302e-01 1.10264730e+00 -1.42698884e-01 2.24156842e-01
-6.65739000e-01 -1.07233119e+00 9.19381082e-01 1.96099842e+00
-2.17395902e-01 -8.09158385e-02 -5.07175028e-01 5.96590936e-01
-1.73416805e+00 -1.33464050e+00 -1.04227759e-01 2.13187194e+00
1.08014333e+00 -1.67150378e-01 1.41521424e-01 6.09471679e-01
3.55046093e-01 1.58311367e-01 -2.50408173e-01 -8.07085097e-01
-5.60525298e-01 9.29842710e-01 3.71757358e-01 1.75388724e-01
-7.90138721e-01 1.34260762e+00 6.12292337e+00 1.27959621e+00
-1.22887063e+00 3.70934866e-02 -1.72162950e-01 -2.27534771e-01
-7.07805932e-01 1.00051120e-01 -6.44424617e-01 6.30742669e-01
1.03204763e+00 -2.56513327e-01 8.94730806e-01 4.79590684e-01
2.39487261e-01 7.75146306e-01 -8.03019464e-01 1.49688828e+00
2.34587252e-01 -1.58516574e+00 5.07033408e-01 -2.02453271e-01
9.05081809e-01 -1.50139421e-01 3.56784284e-01 4.31646734e-01
-6.92929607e-03 -1.19287264e+00 1.20294344e+00 7.27773368e-01
1.10803699e+00 -8.50582898e-01 9.65761021e-02 1.69501722e-01
-1.71702456e+00 -1.05288163e-01 -1.06658161e-01 -1.16247050e-01
4.92688000e-01 2.76126862e-01 -9.05253470e-01 7.98570871e-01
8.45960453e-02 9.85278308e-01 -4.46279705e-01 1.13823247e+00
-2.27011561e-01 9.08136845e-01 -1.90675139e-01 1.13117412e-01
2.11216092e-01 -2.18865886e-01 1.03626621e+00 1.17990899e+00
1.00505757e+00 -3.56408924e-01 -7.45965317e-02 1.03265643e+00
-5.59413619e-02 2.05239713e-01 -8.02529693e-01 -5.44025838e-01
4.93545413e-01 9.72128510e-01 -5.26418686e-01 -1.32134259e-01
3.20116878e-01 1.37013960e+00 -1.99062198e-01 6.22975305e-02
-1.00303638e+00 -5.61647236e-01 2.38642961e-01 -2.86377460e-01
4.49020743e-01 -6.04103625e-01 -3.15422744e-01 -1.22020900e+00
-1.17645055e-01 -1.24567580e+00 6.28387481e-02 -9.44897175e-01
-1.01306844e+00 7.18678713e-01 -4.17919844e-01 -1.78064144e+00
-7.88211465e-01 -1.56137059e-02 -5.96756399e-01 7.70733774e-01
-1.29660487e+00 -1.63287151e+00 3.69799048e-01 4.01229262e-01
6.30446076e-01 -7.37483799e-01 1.27181423e+00 4.17954683e-01
-5.96185625e-02 8.28460395e-01 1.78942412e-01 6.08554622e-03
6.72420859e-01 -1.12491596e+00 6.79476976e-01 3.82963181e-01
1.16023350e+00 6.12806857e-01 5.46802819e-01 -5.17773926e-01
-1.50457823e+00 -1.23495626e+00 1.09627628e+00 -4.27365661e-01
6.02985740e-01 -4.29352522e-01 -3.73736352e-01 7.40802526e-01
1.94791719e-01 -1.01187694e+00 9.22799766e-01 -1.34409100e-01
-3.58925223e-01 -2.80661676e-02 -4.78321791e-01 9.85082567e-01
1.25434053e+00 -7.07251430e-01 -8.41910839e-01 4.15454768e-02
9.78737950e-01 -4.97514129e-01 -8.20651233e-01 4.03027236e-01
8.89265776e-01 -7.81817019e-01 1.10568571e+00 -3.13981384e-01
6.25589252e-01 -5.39208233e-01 -2.83312410e-01 -1.41049612e+00
-4.16940987e-01 -1.50104761e+00 -1.34389281e-01 1.11097407e+00
4.32805330e-01 -3.70555222e-02 5.48872888e-01 -7.40343809e-01
-5.92825413e-01 -3.53117645e-01 -9.83254373e-01 -1.07112193e+00
-1.84043497e-01 -6.49361253e-01 9.34701383e-01 7.73708522e-01
-3.58628154e-01 5.62477589e-01 -1.10087907e+00 -2.46733129e-01
5.14239132e-01 6.00073040e-01 1.01872277e+00 -7.54945695e-01
-1.11740875e+00 -5.35132229e-01 -3.74293834e-01 -9.30459142e-01
1.64817259e-01 -1.67880893e+00 -1.66926116e-01 -1.01781738e+00
-9.29025188e-03 -3.52855176e-01 -5.50854445e-01 2.20869869e-01
4.16460127e-01 8.06844413e-01 6.09551787e-01 1.91522822e-01
-2.15438664e-01 9.86978292e-01 1.39806652e+00 -1.55457973e-01
-5.00513792e-01 2.51072764e-01 -4.69436914e-01 5.00091910e-01
7.25899041e-01 -5.70017219e-01 -3.01053435e-01 -3.97310346e-01
6.84798658e-01 3.13130170e-01 3.95271152e-01 -1.20744944e+00
3.85537371e-02 1.88179344e-01 1.27366215e-01 -6.84741437e-01
7.69227087e-01 -2.66724676e-01 6.88969254e-01 5.19849062e-01
-7.41364419e-01 2.71047086e-01 2.19378114e-01 2.50861615e-01
-3.30115050e-01 -1.90147966e-01 2.66319722e-01 -9.11258906e-02
-2.59168476e-01 1.80596575e-01 -2.57727832e-01 6.75889850e-02
4.69385535e-01 -2.11175010e-01 2.89501965e-01 -3.65332574e-01
-8.09069991e-01 -5.05219579e-01 5.46596572e-02 7.83029199e-01
7.12173283e-01 -1.90010381e+00 -1.02304113e+00 4.37663198e-01
6.45876154e-02 -7.52092719e-01 2.64826208e-01 4.18008387e-01
-4.83301669e-01 4.92014915e-01 -2.91975051e-01 -2.21885905e-01
-1.17481542e+00 -1.62150100e-01 -8.61521885e-02 -4.60242927e-01
-7.17722476e-01 7.85878241e-01 -3.35796714e-01 -7.41804123e-01
1.61600590e-03 -3.14079881e-01 1.17703073e-01 -1.42085537e-01
1.16902374e-01 2.64019310e-01 -3.44088733e-01 -8.16314876e-01
5.25530390e-02 8.81514430e-01 5.49113870e-01 -6.61985457e-01
1.39230692e+00 5.79620153e-02 -3.01930606e-01 1.07072604e+00
1.02993655e+00 5.99257648e-01 -5.34420550e-01 1.03065617e-01
-1.52519375e-01 -1.35128170e-01 -2.86846578e-01 -7.99804389e-01
-5.19924104e-01 7.54314721e-01 2.56111324e-01 -1.44491553e-01
9.69621718e-01 -3.44259769e-01 1.43649662e+00 4.45430756e-01
3.25485677e-01 -9.77947176e-01 2.86249459e-01 8.57555687e-01
1.24204445e+00 -2.77684331e-01 -4.33441937e-01 2.61485487e-01
-8.28545213e-01 1.06755114e+00 -1.08711548e-01 -3.62322628e-01
3.06855798e-01 7.25339055e-02 -2.40419875e-03 5.35113625e-02
-7.08359540e-01 -1.40059009e-01 4.45463866e-01 1.97502136e-01
5.26589811e-01 3.93263847e-01 -4.44834858e-01 1.06961536e+00
-1.52425444e+00 1.11348055e-01 4.71898079e-01 2.92901337e-01
-1.65719360e-01 -1.60249293e+00 -3.38751972e-01 2.62197927e-02
-2.11001247e-01 -5.87620676e-01 -3.92688781e-01 3.25774610e-01
3.49891216e-01 4.90053624e-01 -1.36734784e-01 -8.78503263e-01
1.42438799e-01 3.47639889e-01 7.86427438e-01 -6.06664598e-01
-9.71044004e-01 3.12457085e-01 -2.99472474e-02 -2.76298136e-01
-3.68982293e-02 -1.68887943e-01 -1.14241505e+00 -2.06069686e-02
1.68122038e-01 2.54115075e-01 6.54968560e-01 4.23722535e-01
6.04911864e-01 1.08727741e+00 9.49081123e-01 -9.34004903e-01
-7.20566571e-01 -8.70441914e-01 -9.19097006e-01 2.75243282e-01
2.28630546e-02 -7.56531432e-02 -8.77895132e-02 1.62944183e-01]
|
[16.02560043334961, 5.516234874725342]
|
511f2100-ea71-498a-82ba-66ad14759771
|
robust-lane-detection-through-self-pre
|
2305.17271
| null |
https://arxiv.org/abs/2305.17271v1
|
https://arxiv.org/pdf/2305.17271v1.pdf
|
Robust Lane Detection through Self Pre-training with Masked Sequential Autoencoders and Fine-tuning with Customized PolyLoss
|
Lane detection is crucial for vehicle localization which makes it the foundation for automated driving and many intelligent and advanced driving assistant systems. Available vision-based lane detection methods do not make full use of the valuable features and aggregate contextual information, especially the interrelationships between lane lines and other regions of the images in continuous frames. To fill this research gap and upgrade lane detection performance, this paper proposes a pipeline consisting of self pre-training with masked sequential autoencoders and fine-tuning with customized PolyLoss for the end-to-end neural network models using multi-continuous image frames. The masked sequential autoencoders are adopted to pre-train the neural network models with reconstructing the missing pixels from a random masked image as the objective. Then, in the fine-tuning segmentation phase where lane detection segmentation is performed, the continuous image frames are served as the inputs, and the pre-trained model weights are transferred and further updated using the backpropagation mechanism with customized PolyLoss calculating the weighted errors between the output lane detection results and the labeled ground truth. Extensive experiment results demonstrate that, with the proposed pipeline, the lane detection model performance on both normal and challenging scenes can be advanced beyond the state-of-the-art, delivering the best testing accuracy (98.38%), precision (0.937), and F1-measure (0.924) on the normal scene testing set, together with the best overall accuracy (98.36%) and precision (0.844) in the challenging scene test set, while the training time can be substantially shortened.
|
['Yongqi Dong', 'Ruohan Li']
|
2023-05-26
| null | null | null | null |
['lane-detection']
|
['computer-vision']
|
[-1.6787034e-02 -1.1635632e-01 -2.8684003e-02 -6.2667370e-01
-4.1807675e-01 -8.3366588e-02 3.1105918e-01 -3.1461698e-01
-7.3719049e-01 6.2558568e-01 -4.5137629e-01 -4.0997961e-01
1.8443343e-01 -7.4600816e-01 -6.6496223e-01 -9.5112967e-01
1.0791019e-01 -1.8367307e-02 7.5675595e-01 -6.9419682e-02
3.1843510e-01 5.6228507e-01 -1.8150030e+00 -1.1068242e-01
1.1415566e+00 1.1451063e+00 5.3727239e-01 5.0825369e-01
-1.7720196e-01 6.1217284e-01 -2.7070189e-01 -1.8946806e-01
1.8331289e-01 2.6758712e-02 1.0444739e-01 3.0965525e-01
5.0460440e-01 -7.1555603e-01 -5.4826486e-01 1.0359268e+00
4.3779838e-01 1.6612926e-01 2.0052940e-01 -1.3589450e+00
-1.5898325e-01 4.4615429e-02 -6.9120145e-01 8.1649505e-02
-3.6250198e-01 5.9886223e-01 4.1255045e-01 -1.0661342e+00
2.8475514e-01 1.0044507e+00 7.7302212e-01 1.6842325e-01
-8.0172700e-01 -7.4104893e-01 5.4596156e-02 7.2177392e-01
-1.4441744e+00 -5.1102424e-01 9.6107936e-01 -4.9627563e-01
6.3597739e-01 -1.2873507e-01 5.2643865e-01 4.9348104e-01
4.2958778e-01 7.7577233e-01 1.0768551e+00 -2.2464760e-01
-1.4152800e-01 4.2244479e-01 3.3916456e-01 1.0014074e+00
2.9738954e-01 3.0109316e-01 -7.9699069e-02 6.3104421e-01
4.8301801e-01 -2.2820480e-02 -1.6787916e-01 -1.8322967e-01
-1.1414145e+00 7.5786453e-01 4.8022538e-01 -4.3055434e-02
-4.4874683e-01 -2.2206886e-01 3.3996373e-01 -5.9462674e-02
7.2465345e-02 -1.6814822e-01 -2.3240365e-01 2.1139565e-01
-9.6232343e-01 5.2926607e-02 3.7730694e-01 8.5566366e-01
1.3191700e+00 4.5519263e-01 -1.1990183e-01 8.1223720e-01
4.1341841e-01 7.5887215e-01 3.4584495e-01 -8.4153146e-01
4.3797204e-01 6.3847822e-01 -6.0514171e-02 -1.0973874e+00
-5.5813843e-01 -7.5447047e-01 -6.9268799e-01 4.9768075e-01
3.8313949e-01 -3.6810401e-01 -9.0070415e-01 1.4393885e+00
3.4408608e-01 3.2295585e-01 2.9999453e-01 9.9027014e-01
8.6725992e-01 8.7062365e-01 1.0011797e-01 -2.6644418e-01
1.1863786e+00 -1.2383103e+00 -7.5390774e-01 -7.2741687e-01
5.4061699e-01 -8.3874643e-01 8.9505482e-01 1.1406344e-02
-7.2288829e-01 -1.1472335e+00 -1.4444883e+00 -5.6216087e-02
-4.1731420e-01 7.1512812e-01 5.0991219e-01 3.8083684e-01
-8.9787602e-01 5.3072628e-02 -3.2429135e-01 -8.6258419e-02
3.8721719e-01 1.3918237e-01 -3.1135154e-01 -2.8850529e-01
-1.2213045e+00 8.4129113e-01 7.2373152e-01 6.4290726e-01
-8.3059758e-01 -6.4875263e-01 -8.9672357e-01 -2.6876041e-01
3.5167235e-01 -2.6247498e-01 7.6620722e-01 -7.7459830e-01
-1.3400751e+00 7.2446603e-01 -8.6817741e-02 -5.0898373e-01
7.1404094e-01 -3.9268926e-02 -7.2169024e-01 -1.1938906e-02
2.6672593e-01 1.1639701e+00 7.9547143e-01 -1.3368981e+00
-1.3895708e+00 1.7779663e-02 -3.3604315e-01 1.5700316e-01
5.7467274e-02 -3.9900601e-01 -5.5983889e-01 -1.4392599e-01
2.0454445e-01 -8.7874860e-01 4.7121998e-02 -7.5966924e-02
-3.4288329e-01 -1.4408258e-01 1.3221034e+00 -8.4417003e-01
1.0511092e+00 -2.4731781e+00 -6.8886197e-01 1.3043749e-01
-5.3057256e-03 4.7776523e-01 -1.2109054e-01 -1.2935290e-01
8.2548194e-02 -5.4218435e-01 -2.1971542e-01 -1.5019597e-01
-1.2753047e-01 1.3098635e-01 -2.3889264e-01 4.7794539e-01
3.4812719e-01 9.5227432e-01 -6.4682537e-01 -5.5317098e-01
7.3587501e-01 4.0187526e-01 -1.0614537e-01 3.2151937e-01
2.2077252e-01 2.6127478e-01 -1.5953085e-01 4.3483907e-01
9.6238142e-01 2.8039485e-01 -3.8827145e-01 -4.2299798e-01
-4.8567608e-01 -1.7103674e-01 -1.3025482e+00 1.0483904e+00
-3.5308293e-01 1.2241333e+00 3.9951507e-02 -7.1380246e-01
1.3485764e+00 -2.2286208e-01 1.8699168e-01 -9.7832870e-01
2.3531668e-01 1.9713363e-01 1.9401030e-01 -8.0403268e-01
4.6857351e-01 1.5687597e-01 -6.9953978e-02 -2.0088229e-01
-3.0548650e-01 1.9849545e-01 1.0616186e-01 -1.8808907e-01
5.2255183e-01 -7.4464224e-02 -1.6623066e-01 -3.6233675e-02
9.9865448e-01 2.3685567e-01 6.5376943e-01 5.7779396e-01
-5.0679439e-01 1.7677397e-01 1.4689727e-01 -4.1269350e-01
-1.1134622e+00 -9.5003378e-01 -1.8503973e-01 8.3939445e-01
4.2414066e-01 3.5182863e-01 -8.4901744e-01 -4.8604330e-01
6.5876432e-02 9.6986586e-01 -2.7485457e-01 -3.4664205e-01
-5.4537129e-01 -4.1036806e-01 4.8537335e-01 5.7609630e-01
1.1802164e+00 -1.0988487e+00 -5.9542435e-01 2.4233536e-01
-2.0919188e-05 -1.5107349e+00 -2.8542989e-01 -6.9453232e-02
-5.5923074e-01 -9.0119666e-01 -3.7238505e-01 -1.1400143e+00
7.3783743e-01 7.1591538e-01 4.1910693e-01 3.6256284e-02
-1.5526325e-01 -2.7557278e-01 2.4214943e-01 -4.3292028e-01
-3.2842484e-01 -2.2734302e-01 -1.7713343e-01 4.6569106e-01
4.0700424e-01 -1.1919771e-01 -5.6315267e-01 5.3827620e-01
-4.0558770e-01 5.1424098e-01 1.0333889e+00 8.6421931e-01
5.5191219e-01 2.6366603e-01 7.9389483e-01 -5.3657073e-01
2.6727313e-01 -1.7454162e-01 -1.0139862e+00 -1.4207156e-01
-7.1997058e-01 -3.4691545e-01 6.1593771e-01 -3.0099574e-01
-1.2882574e+00 1.2898043e-01 -2.2691506e-01 -5.1898050e-01
-4.6720645e-01 3.6652508e-01 -3.1178349e-01 -1.4973466e-01
2.6187232e-01 4.4957170e-01 3.2927322e-01 -1.4913520e-03
1.5304722e-01 8.2208616e-01 9.7342801e-01 2.3014665e-01
8.2830584e-01 4.2956373e-01 -4.0923923e-02 -7.7973074e-01
-5.7509273e-01 -5.4942083e-01 -6.2119257e-01 -6.6903782e-01
9.7374529e-01 -1.0501827e+00 -6.6738015e-01 7.7044737e-01
-8.9814049e-01 -3.7948465e-01 1.4514144e-01 6.4644623e-01
-3.7327546e-01 2.9641718e-01 -2.1776341e-01 -7.9030144e-01
-1.7522639e-01 -1.3321207e+00 6.9113219e-01 5.2836335e-01
3.9768380e-01 -7.4947077e-01 -6.8987417e-01 6.0109621e-01
4.1877428e-01 1.2222066e-01 8.0580384e-01 -3.7289613e-01
-8.4514201e-01 -7.2042406e-01 -6.4147151e-01 6.2107444e-01
-1.1170690e-01 6.7756809e-02 -1.1931452e+00 2.8267442e-03
-3.7019300e-01 4.1932352e-02 1.0084316e+00 4.2357990e-01
9.7456896e-01 4.4556942e-02 -3.1861842e-01 4.9343777e-01
1.3822482e+00 4.5516136e-01 6.8941194e-01 5.0251454e-01
9.0854293e-01 7.6907641e-01 9.4745141e-01 -1.6277884e-03
5.5319023e-01 3.0922905e-01 4.4695622e-01 -2.1377496e-01
-3.3917764e-01 -2.5804737e-01 4.0904564e-01 6.1726028e-01
4.1764000e-01 1.3277556e-01 -7.2853363e-01 6.4262164e-01
-1.8294965e+00 -1.0895766e+00 -4.6562243e-01 2.2243595e+00
3.9437333e-01 7.9201370e-01 -1.2736588e-02 2.4890935e-01
8.9720994e-01 1.6259730e-01 -6.6160256e-01 -2.7318618e-01
-1.8582360e-01 -4.9688873e-01 8.8774675e-01 6.3418788e-01
-1.3916438e+00 1.1285713e+00 4.8152957e+00 9.2637438e-01
-1.5479207e+00 -1.0793913e-01 5.9370226e-01 3.8079008e-01
1.4584504e-01 -1.4547290e-01 -1.1626704e+00 6.7744970e-01
9.8886442e-01 1.7930207e-01 3.4284428e-01 1.0054746e+00
6.9130278e-01 -3.1905410e-01 -6.3468015e-01 1.0029463e+00
1.3771465e-01 -1.1988775e+00 -3.6089811e-01 -1.4878397e-01
7.4195069e-01 1.8375751e-01 5.1180590e-03 6.0044670e-01
-2.8716531e-01 -7.2260499e-01 8.3316565e-01 7.5067776e-01
7.0764256e-01 -8.4837401e-01 1.0923686e+00 7.2405481e-01
-1.1259974e+00 -3.2336614e-01 -3.4095415e-01 8.1779808e-02
2.3806472e-01 5.8223701e-01 -1.0961367e+00 2.2802544e-01
4.2438000e-01 6.6842717e-01 -6.3247460e-01 1.2103115e+00
-2.9490200e-01 6.8527305e-01 -2.0697963e-01 -1.1545684e-01
4.9861103e-01 -2.3425174e-01 3.9561000e-01 1.3583544e+00
6.3022308e-02 -2.7859300e-01 3.2635650e-01 6.5375185e-01
2.0953460e-01 -1.1952371e-01 -4.7863579e-01 6.1731488e-01
5.4809719e-01 1.4890689e+00 -4.7576761e-01 -3.4787473e-01
-4.9497345e-01 4.0929106e-01 -1.1548565e-02 5.3850472e-01
-1.2106601e+00 -7.6258105e-01 3.7935188e-01 1.9974639e-01
3.4103739e-01 -1.9818120e-01 -3.4691760e-01 -4.1551778e-01
3.9409779e-02 -3.7566814e-01 -8.4791571e-02 -1.0000275e+00
-8.5746419e-01 3.9917752e-01 -1.0650761e-01 -1.3398926e+00
-1.6091338e-01 -7.2712904e-01 -6.7350811e-01 8.8818443e-01
-2.0121880e+00 -1.1953796e+00 -8.0662256e-01 2.2633366e-01
7.5577354e-01 -3.7244290e-01 1.3167231e-01 5.4334527e-01
-9.9196380e-01 6.7750847e-01 1.4168976e-01 2.2953887e-01
6.3408506e-01 -8.1759632e-01 -5.9902489e-02 1.1401607e+00
-5.8524001e-01 4.7065467e-02 5.1036763e-01 -4.5699808e-01
-1.2126036e+00 -1.6678815e+00 6.8547994e-01 1.0982416e-01
4.9243101e-01 -8.6038515e-02 -9.6078557e-01 4.0118822e-01
-4.4794395e-03 2.2247550e-01 5.7271451e-02 -5.4733163e-01
1.3275437e-01 -7.4588519e-01 -1.0824693e+00 5.6796324e-01
6.8140334e-01 -3.2936385e-01 -3.7562600e-01 -6.9831155e-02
5.0444424e-01 -4.5163003e-01 -4.5748118e-01 5.9132093e-01
6.6426015e-01 -9.2743236e-01 7.7583516e-01 8.2019627e-02
2.3296164e-01 -8.5602909e-01 -6.5700635e-02 -9.4294143e-01
-3.5378107e-01 6.4232075e-03 2.5314108e-01 1.2182575e+00
5.0453246e-01 -9.2023963e-01 7.1497864e-01 3.3663332e-01
-7.2759533e-01 -8.3780003e-01 -8.2580072e-01 -3.9819717e-01
-4.0587685e-01 -6.1497694e-01 3.5807216e-01 6.3303632e-01
-7.4387896e-01 3.2926464e-01 -6.2319074e-02 5.4126996e-01
7.8663439e-01 -3.5956521e-02 9.9336749e-01 -1.0296004e+00
2.1682620e-01 -4.7554868e-01 -6.1114794e-01 -1.1174990e+00
3.9455146e-01 -6.6420412e-01 4.2204621e-01 -1.5666115e+00
-1.5121739e-01 -4.5853317e-01 -2.9881456e-01 4.3223116e-01
-3.2985508e-01 2.3649055e-01 1.0474051e-01 2.0129791e-01
-5.5894554e-01 3.2229093e-01 1.2611010e+00 -3.9446899e-01
-1.5258871e-01 8.6179614e-02 -3.2344684e-01 7.8964090e-01
8.6834890e-01 5.9613921e-02 -5.2421993e-01 -2.6277554e-01
-5.0373304e-01 -3.0219996e-02 6.2394464e-01 -1.4048567e+00
4.4720611e-01 -1.3735907e-01 7.4361771e-01 -1.2355373e+00
2.4971819e-01 -9.9410003e-01 -5.3371787e-02 4.7776157e-01
-6.1964057e-02 -8.3218135e-02 5.1299953e-01 5.4588330e-01
-1.8407385e-01 -1.1979829e-01 8.3906609e-01 3.8915327e-01
-1.7003063e+00 9.8872609e-02 -3.3919561e-01 -2.8569055e-01
1.4444162e+00 -9.2548162e-01 -5.2436185e-01 -1.9130222e-01
-3.4541431e-01 5.3398299e-01 2.2100821e-01 5.0536454e-01
8.5898161e-01 -1.1945301e+00 -7.1503770e-01 7.2767675e-01
2.2928420e-01 7.2444841e-02 5.6029183e-01 1.0278351e+00
-6.0133952e-01 4.9770224e-01 -3.8199958e-01 -8.8202435e-01
-9.5361602e-01 4.5388708e-01 4.8388901e-01 3.0816352e-01
-5.1145214e-01 4.2917758e-01 1.3359943e-01 -1.4240479e-01
2.1681876e-01 -2.2556892e-01 -3.8382056e-01 -6.1991669e-02
4.3842751e-01 4.9530068e-01 1.2849742e-01 -1.0149467e+00
-2.8263837e-01 6.1225045e-01 8.3014900e-03 2.6655483e-01
9.7790509e-01 -3.0862817e-01 1.5666957e-01 4.2341605e-01
1.2413210e+00 7.2563745e-02 -1.7692599e+00 4.1013323e-02
-3.7640384e-01 -2.9470479e-01 4.3905106e-01 -8.3310241e-01
-1.1029488e+00 9.6415019e-01 1.0573747e+00 -2.5296172e-02
1.0318031e+00 -4.3893120e-01 9.2007494e-01 3.2530031e-01
1.5463856e-01 -1.0906465e+00 -3.4805602e-01 4.7711164e-01
5.7192349e-01 -1.4281955e+00 -3.5632995e-01 -4.2157751e-01
-7.7249229e-01 1.0853978e+00 8.5379571e-01 -2.1015127e-01
4.4457337e-01 1.7194788e-01 4.3580234e-01 2.9248986e-01
-4.6027020e-01 -2.5891194e-01 2.6671523e-01 5.0378805e-01
-6.3861728e-02 1.1082249e-01 -1.1137723e-01 3.0339006e-01
-1.7986979e-02 -8.3521098e-02 3.6402065e-01 5.2306187e-01
-1.0287647e+00 -3.0286929e-01 -3.8982409e-01 5.8321339e-01
8.6330630e-02 1.3297388e-01 2.2554600e-01 9.2793453e-01
5.5940866e-01 1.0141023e+00 3.0860484e-01 -6.1270410e-01
4.7065517e-01 1.2700814e-01 -2.0269795e-01 -1.6570291e-03
9.8292850e-02 1.0264697e-02 1.5443395e-01 -2.8394994e-01
-4.8180208e-02 -4.4625166e-01 -1.5525956e+00 -4.5446792e-01
-3.7756199e-01 -5.4825589e-02 7.5664377e-01 1.0150108e+00
4.1293764e-01 6.9651234e-01 9.1854042e-01 -9.1651475e-01
-3.3514374e-01 -8.7026703e-01 -2.4466732e-01 2.3975453e-01
5.1515108e-01 -7.4435407e-01 -2.7034682e-01 8.8525467e-02]
|
[8.069791793823242, -1.292013168334961]
|
c31930bb-0b45-4d88-aae8-14d8e40c2967
|
spatial-temporal-graph-learning-with
|
2306.10683
| null |
https://arxiv.org/abs/2306.10683v1
|
https://arxiv.org/pdf/2306.10683v1.pdf
|
Spatial-Temporal Graph Learning with Adversarial Contrastive Adaptation
|
Spatial-temporal graph learning has emerged as a promising solution for modeling structured spatial-temporal data and learning region representations for various urban sensing tasks such as crime forecasting and traffic flow prediction. However, most existing models are vulnerable to the quality of the generated region graph due to the inaccurate graph-structured information aggregation schema. The ubiquitous spatial-temporal data noise and incompleteness in real-life scenarios pose challenges in generating high-quality region representations. To address this challenge, we propose a new spatial-temporal graph learning model (GraphST) for enabling effective self-supervised learning. Our proposed model is an adversarial contrastive learning paradigm that automates the distillation of crucial multi-view self-supervised information for robust spatial-temporal graph augmentation. We empower GraphST to adaptively identify hard samples for better self-supervision, enhancing the representation discrimination ability and robustness. In addition, we introduce a cross-view contrastive learning paradigm to model the inter-dependencies across view-specific region representations and preserve underlying relation heterogeneity. We demonstrate the superiority of our proposed GraphST method in various spatial-temporal prediction tasks on real-life datasets. We release our model implementation via the link: \url{https://github.com/HKUDS/GraphST}.
|
['Ruihua Han', 'SiuMing Yiu', 'Zheng Wang', 'Lianghao Xia', 'Chao Huang', 'Qianru Zhang']
|
2023-06-19
| null | null | null | null |
['contrastive-learning', 'graph-learning', 'contrastive-learning']
|
['computer-vision', 'graphs', 'methodology']
|
[ 1.75618440e-01 1.13950767e-01 -6.03628933e-01 -4.39449400e-01
-8.83279920e-01 -4.62067664e-01 5.66890657e-01 3.55434448e-01
1.00349016e-01 6.09772861e-01 5.09625912e-01 -4.29798216e-01
-1.73930377e-01 -1.07402480e+00 -7.90687978e-01 -4.57210094e-01
-2.61132061e-01 2.43035018e-01 2.53269315e-01 -3.72875780e-01
-1.55405149e-01 5.64506948e-01 -1.23566675e+00 4.53132570e-01
1.24522698e+00 7.60498822e-01 -4.84686671e-03 5.90794325e-01
6.08628206e-02 8.50315630e-01 -2.57893950e-01 -3.53351921e-01
3.47156346e-01 -5.41075543e-02 -5.20036101e-01 1.27972007e-01
6.13085926e-01 -1.64602548e-01 -8.61416399e-01 9.00172174e-01
4.02261078e-01 3.60708445e-01 6.26492023e-01 -1.60484898e+00
-9.10654783e-01 5.61337531e-01 -9.64346945e-01 6.02161944e-01
2.21456423e-01 2.93613851e-01 8.26548040e-01 -5.31400979e-01
6.69498920e-01 1.19789422e+00 6.55767858e-01 4.52953279e-01
-1.16496277e+00 -9.46528912e-01 6.35953784e-01 3.70585889e-01
-1.57307661e+00 -4.52308089e-01 1.24144673e+00 -6.66279137e-01
6.91766918e-01 2.83342481e-01 6.74777746e-01 1.28088307e+00
1.67622000e-01 8.47192526e-01 1.20416903e+00 2.39957511e-01
2.84536090e-02 -1.46917060e-01 1.02734677e-01 7.51896739e-01
1.41177177e-01 2.67583042e-01 -4.45037246e-01 -8.78126249e-02
8.73642921e-01 3.59010100e-01 1.38308508e-02 -3.52371782e-01
-8.45263362e-01 6.89420164e-01 1.13464546e+00 1.93639286e-02
-2.05946922e-01 1.35239124e-01 4.51201916e-01 6.06505387e-02
9.58216965e-01 -1.75663635e-01 -6.65037632e-02 2.31418684e-01
-8.37251186e-01 1.37348920e-01 -1.27351359e-01 1.06623006e+00
8.07910919e-01 3.59250814e-01 -3.49864155e-01 7.65312254e-01
1.66955337e-01 6.10230744e-01 3.99070270e-02 -5.33691764e-01
1.14587259e+00 9.43747759e-01 -3.13040555e-01 -1.37159467e+00
-4.82347369e-01 -6.81679130e-01 -1.34853160e+00 -7.90412575e-02
2.89414227e-01 4.55255210e-02 -1.06213260e+00 1.83926487e+00
5.51643252e-01 6.02877855e-01 -8.20191205e-02 7.52139688e-01
1.02197683e+00 6.15498602e-01 1.72788635e-01 5.81904724e-02
1.09026432e+00 -7.97253251e-01 -6.27587020e-01 -4.58974332e-01
6.25291884e-01 -1.46068215e-01 1.09714365e+00 -1.70370862e-01
-7.47481823e-01 -6.16317868e-01 -8.23637009e-01 -3.79756242e-02
-6.60870373e-01 -1.42896339e-01 6.21504068e-01 3.13655049e-01
-7.74215043e-01 2.71895707e-01 -8.35497975e-01 -1.01775490e-01
1.00129783e+00 1.04739621e-01 -6.44773781e-01 -4.34438139e-01
-1.28484690e+00 3.14386010e-01 2.94853061e-01 2.11541250e-01
-5.37501991e-01 -8.74852002e-01 -1.29578257e+00 -2.03247577e-01
3.85314763e-01 -4.76707757e-01 4.63560611e-01 -6.69685066e-01
-6.25754654e-01 8.03611577e-01 -1.61369666e-01 -4.53579754e-01
5.34368813e-01 1.88479871e-01 -7.85883844e-01 2.07880158e-02
6.31254911e-01 3.96370769e-01 8.66437495e-01 -1.38834894e+00
-2.41041884e-01 -7.27312386e-01 1.84022840e-02 1.51623383e-01
-1.97351262e-01 -4.49300140e-01 -4.26656395e-01 -1.01730311e+00
1.43967733e-01 -6.22430921e-01 -4.49048996e-01 -1.28651634e-01
-4.95082140e-01 -1.27602685e-02 8.69003534e-01 -8.32088709e-01
1.34141934e+00 -2.21176434e+00 -1.35797456e-01 2.85865188e-01
6.46633148e-01 1.41591340e-01 -2.28854984e-01 4.42220420e-01
-1.52813554e-01 1.45717204e-01 -4.24350858e-01 -3.95701915e-01
-2.54558414e-01 2.75863141e-01 -3.36244106e-01 5.84671140e-01
1.80257574e-01 1.21955347e+00 -1.24398136e+00 -6.26445115e-01
3.06050807e-01 5.23865402e-01 -4.17726964e-01 1.80229664e-01
2.58088484e-02 8.19933534e-01 -6.85943305e-01 6.93938375e-01
8.91168475e-01 -4.89955038e-01 -1.11247450e-01 -1.66694105e-01
2.92304575e-01 5.09364977e-02 -1.21053314e+00 1.56802785e+00
-4.27306652e-01 3.98935974e-01 1.65432179e-03 -1.05740011e+00
9.37600076e-01 -5.59871197e-02 5.81936359e-01 -1.11911237e+00
-2.36819148e-01 -1.34561881e-01 -3.08316618e-01 -3.45422626e-01
5.82279742e-01 7.40268305e-02 -3.05819243e-01 3.53408724e-01
-1.88411593e-01 1.20133981e-01 -9.44613740e-02 5.99835038e-01
1.08806777e+00 2.73822341e-03 3.53711218e-01 -7.78921917e-02
3.14553559e-01 -7.22687915e-02 6.74171507e-01 4.77630556e-01
-3.01959366e-01 7.20567524e-01 3.81354392e-01 -7.50857055e-01
-8.49038661e-01 -1.26020527e+00 -1.89639833e-02 9.91971076e-01
9.96804982e-02 -3.80193561e-01 -3.45328569e-01 -9.83078957e-01
4.26534265e-02 4.28119332e-01 -8.73981237e-01 -3.47994030e-01
-5.62917292e-01 -5.72177231e-01 4.46984887e-01 7.52589047e-01
6.02758229e-01 -7.86528468e-01 1.44227043e-01 -6.67214617e-02
-2.97297090e-01 -1.27106059e+00 -7.39745140e-01 -3.60768348e-01
-7.42127895e-01 -1.05511332e+00 -3.86715502e-01 -3.75975132e-01
8.44362557e-01 5.69871306e-01 1.08098817e+00 8.42660144e-02
-1.30137071e-01 3.09428126e-01 -3.86921644e-01 -1.38661265e-01
-2.92609036e-01 2.22378567e-01 -1.20706707e-01 3.80353481e-01
-1.66616701e-02 -1.04047740e+00 -7.92179942e-01 1.70664817e-01
-8.93616438e-01 3.74892294e-01 2.58911371e-01 6.81046426e-01
8.46154034e-01 -2.25921739e-02 8.12530041e-01 -1.21329534e+00
5.42632043e-01 -9.64381218e-01 -5.44343054e-01 2.00000659e-01
-4.23548669e-01 -6.15929514e-02 8.05918694e-01 -2.65166163e-01
-8.10739279e-01 -6.33027405e-02 -4.78111356e-02 -7.66990244e-01
-1.87149286e-01 4.60830092e-01 -3.10681611e-01 1.76228017e-01
8.92574787e-01 3.77665371e-01 -1.91098213e-01 -1.46882072e-01
6.04930878e-01 3.07967484e-01 5.23281693e-01 -6.06938303e-01
1.07671428e+00 6.19044423e-01 9.97762159e-02 -7.16604054e-01
-9.08355236e-01 -5.90352416e-01 -6.96296155e-01 -4.13758755e-01
8.38886499e-01 -1.22983825e+00 -2.70607293e-01 2.48523533e-01
-7.66433895e-01 -5.99897802e-01 -3.49526018e-01 8.06691796e-02
-4.15334702e-01 3.80785733e-01 -2.83895344e-01 -7.97985077e-01
-3.21290016e-01 -9.23778236e-01 1.37461460e+00 2.85660345e-02
2.53275931e-01 -1.29209208e+00 9.07201990e-02 6.94226027e-01
2.88635850e-01 8.55078459e-01 5.75886905e-01 -4.02176768e-01
-7.19211519e-01 -1.49694383e-01 -4.78547275e-01 8.05033743e-02
3.46267641e-01 -3.53584826e-01 -9.82616425e-01 -4.41740155e-01
-6.67059541e-01 -1.70265794e-01 1.13936090e+00 2.50639051e-01
1.52593315e+00 -6.87886834e-01 -3.77957731e-01 9.30558801e-01
1.34536505e+00 -3.58550489e-01 4.41629171e-01 -8.05187225e-02
1.41739810e+00 6.77354276e-01 4.98988956e-01 3.30848098e-01
8.27632844e-01 7.08797812e-01 4.62792695e-01 -2.75262982e-01
-2.66122431e-01 -8.32573235e-01 9.80524048e-02 5.33170223e-01
-1.86371967e-01 -3.15379053e-01 -1.23931515e+00 7.62222886e-01
-2.15053225e+00 -1.32744563e+00 -2.05557689e-01 1.97517014e+00
2.72075981e-01 1.20848373e-01 4.06650543e-01 5.36021171e-03
5.99119306e-01 7.65549898e-01 -6.39724255e-01 6.13044910e-02
-1.69746414e-01 -9.94376540e-02 6.98917449e-01 6.09719336e-01
-1.25897431e+00 1.10894835e+00 5.11203480e+00 9.94545877e-01
-1.03719938e+00 2.93876648e-01 9.70433950e-01 8.53710622e-02
-6.62972331e-01 -2.80784369e-01 -3.51437718e-01 4.54490542e-01
7.15025961e-01 -8.54765102e-02 3.85954022e-01 7.82523334e-01
3.88156503e-01 3.42316210e-01 -6.39562428e-01 1.00954819e+00
2.01438065e-03 -1.69558775e+00 2.71876752e-01 2.44997606e-01
8.47266495e-01 1.67901292e-01 2.64030814e-01 3.04907531e-01
4.08483684e-01 -1.15820253e+00 3.94212663e-01 5.62546492e-01
9.99581754e-01 -7.39634812e-01 2.74849296e-01 3.64223212e-01
-1.82990491e+00 -4.74644750e-02 -6.60981610e-02 5.20420410e-02
2.17205107e-01 5.01082540e-01 -7.68724442e-01 9.52573776e-01
5.49803972e-01 1.32187653e+00 -9.53794181e-01 6.45373702e-01
-7.99621195e-02 5.64387262e-01 -1.38043255e-01 5.66025972e-01
1.22495987e-01 -2.92367101e-01 6.03249788e-01 1.01595986e+00
1.69609580e-02 2.05744699e-01 4.55324560e-01 8.96462381e-01
-9.57505256e-02 3.67956795e-02 -1.26698470e+00 3.08081090e-01
6.36443436e-01 1.01738501e+00 -6.33933127e-01 -2.42307603e-01
-4.37593699e-01 6.73424304e-01 8.22244346e-01 5.27630448e-01
-8.70779574e-01 2.72722602e-01 6.92941129e-01 6.66543901e-01
2.50303224e-02 -2.22854778e-01 -3.58354956e-01 -1.23744822e+00
2.44709462e-01 -6.81386888e-01 7.53356814e-01 -5.50241649e-01
-1.54983270e+00 6.75853848e-01 4.52551693e-01 -1.47408962e+00
-9.87645537e-02 -2.07800210e-01 -7.90043771e-01 6.10833943e-01
-1.48207510e+00 -1.89279115e+00 -4.02625382e-01 1.06566763e+00
4.38914567e-01 -3.06885242e-01 3.05828899e-01 3.48226517e-01
-7.60128796e-01 7.58875072e-01 -3.03206295e-01 4.13669497e-01
3.74556363e-01 -1.19760883e+00 7.37496197e-01 1.20788503e+00
1.86781585e-01 3.68541509e-01 3.50871950e-01 -9.29905534e-01
-1.08645678e+00 -1.89182806e+00 4.66783464e-01 -4.98635828e-01
7.33298898e-01 -7.13765860e-01 -1.02398777e+00 8.63038540e-01
-1.63178846e-01 8.26004207e-01 5.64710319e-01 8.24036971e-02
-6.11126244e-01 -4.25043970e-01 -1.12598860e+00 6.03279293e-01
1.42987525e+00 -8.07436228e-01 -7.09010512e-02 4.54772145e-01
8.43450725e-01 -4.85265225e-01 -9.48053122e-01 4.47596967e-01
1.69961020e-01 -8.66109192e-01 1.11425269e+00 -5.76399684e-01
2.61821091e-01 -2.34488338e-01 -1.70122549e-01 -1.22814524e+00
-4.84219462e-01 -5.00985444e-01 -3.98401588e-01 1.29489040e+00
3.39174956e-01 -8.46436501e-01 9.70502615e-01 3.93087059e-01
-1.52177885e-01 -8.79466295e-01 -1.22339642e+00 -6.31611645e-01
1.01834044e-01 -6.74739301e-01 9.76482213e-01 1.15461588e+00
-2.73853272e-01 4.74380434e-01 -6.02608383e-01 4.60669994e-01
9.00696278e-01 2.09602416e-01 9.17136908e-01 -1.08249962e+00
-9.34282169e-02 -3.75582933e-01 -6.97398841e-01 -7.88751483e-01
2.69970387e-01 -1.16532898e+00 -5.89829326e-01 -1.66697526e+00
1.21288039e-01 -6.92287385e-01 -2.35097602e-01 3.85412395e-01
-4.16155279e-01 1.29354253e-01 1.05871044e-01 5.45428097e-02
-7.26184011e-01 6.81219518e-01 1.55939519e+00 -3.43639821e-01
-2.10457504e-01 2.42115259e-01 -7.29717731e-01 3.27312917e-01
1.03725135e+00 -2.85624772e-01 -9.70639646e-01 -3.77955347e-01
1.87839329e-01 2.81262964e-01 6.59190595e-01 -9.47511673e-01
1.38059303e-01 -4.78479326e-01 2.64577150e-01 -6.53103650e-01
3.12318325e-01 -8.27898264e-01 1.35885894e-01 6.98568895e-02
-1.96197793e-01 3.83407742e-01 3.27871621e-01 1.03970718e+00
-9.49625671e-02 7.68657148e-01 5.12274444e-01 -8.23007300e-02
-7.42272675e-01 1.06937230e+00 9.81025845e-02 4.23931807e-01
1.10874009e+00 -3.17722231e-01 -5.82436025e-01 -5.16212165e-01
-7.28007138e-01 5.13758898e-01 4.67086047e-01 6.28565133e-01
9.01356399e-01 -1.67785418e+00 -8.89514923e-01 5.14186382e-01
5.92156827e-01 4.15602297e-01 8.50864410e-01 6.36768997e-01
-1.80471510e-01 -1.29520912e-02 -1.83325186e-01 -6.58583879e-01
-1.03836608e+00 7.82409906e-01 3.29058290e-01 -5.98193228e-01
-8.32056284e-01 4.65385765e-01 3.18848759e-01 -6.91102147e-01
-1.66998699e-01 -2.92107821e-01 -3.26878488e-01 -9.72784460e-02
3.81817847e-01 3.30237985e-01 -6.89048469e-02 -1.09731674e+00
-3.50356996e-01 4.23304409e-01 1.53754100e-01 1.77394688e-01
1.46958590e+00 -8.64633098e-02 2.60451347e-01 4.50668037e-01
1.15649319e+00 1.75928231e-02 -1.32711565e+00 -4.37014192e-01
-4.34379280e-01 -6.03926420e-01 -9.04858205e-03 -3.18973988e-01
-1.40665603e+00 7.89715230e-01 4.91019905e-01 1.51998267e-01
1.04197764e+00 1.18394673e-01 7.56375074e-01 -8.68975148e-02
3.79058659e-01 -7.83534169e-01 1.69548303e-01 2.02526525e-01
1.01084173e+00 -1.62062085e+00 1.69928908e-01 -6.35087848e-01
-7.72061050e-01 5.42238832e-01 8.59980285e-01 -2.26619005e-01
9.51527536e-01 2.66800728e-02 -1.09107338e-01 -5.74953139e-01
-6.60687625e-01 -3.95179451e-01 7.59023964e-01 9.68815506e-01
1.13373771e-01 4.99542952e-01 3.68474454e-01 4.20096397e-01
-2.29876667e-01 -4.87866670e-01 1.11497805e-01 6.36681914e-01
8.90052691e-02 -8.70575905e-01 -2.51760304e-01 7.16345847e-01
-4.95512895e-02 6.18260773e-03 -2.40361869e-01 7.83944666e-01
7.37370104e-02 8.50122094e-01 1.21355839e-01 -6.74699128e-01
4.03091013e-01 -3.87489915e-01 1.16910636e-01 -6.41263127e-01
-3.35120976e-01 -2.04136968e-01 -1.83288548e-02 -7.15439558e-01
-3.91749173e-01 -6.46727741e-01 -9.94856954e-01 -4.07000989e-01
2.35449299e-01 -3.46979611e-02 -1.10733159e-01 8.07748497e-01
6.55220747e-01 6.14430070e-01 8.43280196e-01 -7.88242698e-01
5.35739772e-02 -7.25212276e-01 -4.64170575e-01 7.43436158e-01
6.77230239e-01 -8.11220646e-01 -1.38222044e-02 -2.46689081e-01]
|
[6.531975746154785, 2.1012609004974365]
|
5d2aa744-a108-4fc1-acc5-75ed8554e190
|
knowing-the-distance-understanding-the-gap
|
2303.15219
| null |
https://arxiv.org/abs/2303.15219v1
|
https://arxiv.org/pdf/2303.15219v1.pdf
|
Knowing the Distance: Understanding the Gap Between Synthetic and Real Data For Face Parsing
|
The use of synthetic data for training computer vision algorithms has become increasingly popular due to its cost-effectiveness, scalability, and ability to provide accurate multi-modality labels. Although recent studies have demonstrated impressive results when training networks solely on synthetic data, there remains a performance gap between synthetic and real data that is commonly attributed to lack of photorealism. The aim of this study is to investigate the gap in greater detail for the face parsing task. We differentiate between three types of gaps: distribution gap, label gap, and photorealism gap. Our findings show that the distribution gap is the largest contributor to the performance gap, accounting for over 50% of the gap. By addressing this gap and accounting for the labels gap, we demonstrate that a model trained on synthetic data achieves comparable results to one trained on a similar amount of real data. This suggests that synthetic data is a viable alternative to real data, especially when real data is limited or difficult to obtain. Our study highlights the importance of content diversity in synthetic datasets and challenges the notion that the photorealism gap is the most critical factor affecting the performance of computer vision models trained on synthetic data.
|
['Orly Zvitia', 'Moran Rubin', 'Max Kogan', 'Vladimir Loginov', 'Alexey Gruzdev', 'Assaf Lehr', 'Eli Friedman']
|
2023-03-27
| null | null | null | null |
['face-parsing']
|
['computer-vision']
|
[ 4.08585399e-01 1.48307696e-01 -1.11418463e-01 -3.64586502e-01
-8.71433258e-01 -6.67194664e-01 7.67021060e-01 3.86916250e-02
-5.53224564e-01 5.32830000e-01 8.51857886e-02 -2.30158582e-01
2.55698472e-01 -4.21339005e-01 -8.75796318e-01 -2.91870177e-01
6.15830123e-01 4.53146428e-01 -2.48326715e-02 3.44508812e-02
9.33727846e-02 5.07523417e-01 -1.86573279e+00 2.17595428e-01
6.50484443e-01 7.79276550e-01 1.98612344e-02 2.96719909e-01
-2.94760108e-01 8.21726918e-01 -7.25613773e-01 -3.74778777e-01
4.18974429e-01 -4.68258619e-01 -6.54314399e-01 2.74974376e-01
1.02079976e+00 -4.69313592e-01 -3.49164866e-02 7.37374067e-01
4.85140443e-01 -2.88550317e-01 6.82699084e-01 -1.70882225e+00
-5.78950286e-01 2.36663356e-01 -5.27340770e-01 2.31008783e-01
1.73762202e-01 3.78776729e-01 9.43190932e-01 -8.24211478e-01
9.42446768e-01 1.30084908e+00 8.14263940e-01 7.41330504e-01
-1.54911184e+00 -6.59252107e-01 -1.91773251e-01 -9.73059610e-02
-1.13946044e+00 -6.97476208e-01 6.80160463e-01 -7.19332159e-01
5.92583776e-01 4.53851698e-03 5.45440555e-01 1.45568430e+00
-9.99478027e-02 4.21051562e-01 1.44981062e+00 -7.36776352e-01
1.62906319e-01 3.81114930e-01 -1.19879685e-01 5.32924533e-01
5.39139807e-01 2.33242810e-01 -4.09014285e-01 -8.06103498e-02
5.18163264e-01 -3.99334669e-01 -1.57969117e-01 -3.56627434e-01
-1.04332721e+00 8.98336411e-01 2.26608783e-01 1.30629644e-01
-1.23753749e-01 4.06195689e-03 4.74237174e-01 2.01349288e-01
4.86658216e-01 6.53938711e-01 -1.58741936e-01 -2.13623449e-01
-9.81319547e-01 1.98025092e-01 6.58969939e-01 5.99761367e-01
4.42011237e-01 2.73118526e-01 1.58302635e-01 1.10734665e+00
2.62068897e-01 3.95840764e-01 4.36463237e-01 -1.27415669e+00
4.68167722e-01 6.18401229e-01 7.53876716e-02 -1.02266037e+00
-3.69880348e-01 -2.53007710e-01 -3.17375213e-01 4.59585041e-01
1.02924550e+00 -2.49523157e-03 -8.93516660e-01 2.20266414e+00
2.86375552e-01 -2.00089991e-01 -5.73650636e-02 7.23844469e-01
9.78021622e-01 2.51473576e-01 4.85856563e-01 1.03984587e-01
1.20042074e+00 -8.64287794e-01 -2.54763067e-01 -4.91543531e-01
6.52178407e-01 -9.88567770e-01 1.42877781e+00 -6.68640137e-02
-8.76210928e-01 -4.71136719e-01 -1.15679252e+00 -2.19666539e-03
-3.60951930e-01 1.38349727e-01 4.05365318e-01 7.96800792e-01
-1.17090499e+00 3.11426461e-01 -4.08373535e-01 -7.22576201e-01
6.25914812e-01 -1.02122530e-01 -6.30482018e-01 -4.19368923e-01
-8.20933342e-01 1.08057606e+00 2.37185329e-01 -1.83350861e-01
-5.35912812e-01 -9.08241689e-01 -8.40009212e-01 -2.08446592e-01
4.41613980e-02 -3.60628396e-01 1.31841290e+00 -1.29532433e+00
-7.47196317e-01 1.23174334e+00 1.91340148e-01 -1.76557109e-01
6.92366719e-01 4.11509991e-01 -2.28263393e-01 2.85859853e-01
3.57984394e-01 1.05703270e+00 6.22985184e-01 -1.61404669e+00
-4.05393094e-01 -4.70672280e-01 4.81008738e-03 7.51598701e-02
-2.28924677e-01 5.61585762e-02 -1.85140476e-01 -5.34312546e-01
5.91756031e-02 -1.17989779e+00 2.28781059e-01 1.38013139e-01
-1.15022682e-01 -2.13220328e-01 7.96571553e-01 -6.54917717e-01
6.59867585e-01 -2.27804351e+00 -4.10473764e-01 -2.82430619e-01
2.92563170e-01 2.80134588e-01 -3.52900922e-01 5.13545275e-01
-1.13292076e-01 3.85670990e-01 -1.35582477e-01 -4.42221403e-01
-2.15098083e-01 9.89964306e-02 -1.64719880e-01 3.17444742e-01
4.36310738e-01 7.93520451e-01 -6.48485184e-01 -5.91796517e-01
7.94744343e-02 5.61128438e-01 -1.87559813e-01 -9.06396359e-02
-2.76251435e-01 4.78234202e-01 4.87416564e-03 6.03641510e-01
8.15179646e-01 -3.37552160e-01 4.83375117e-02 -2.59124428e-01
-1.06141157e-01 5.80132045e-02 -7.96841919e-01 1.27868867e+00
-3.44611019e-01 1.05061758e+00 -8.54546502e-02 -5.21932900e-01
7.16520429e-01 2.33215794e-01 3.91245186e-01 -1.02304649e+00
7.45147169e-02 3.07136834e-01 7.68139660e-02 -5.35861135e-01
4.79419738e-01 -4.22516346e-01 3.22874725e-01 6.85914278e-01
-5.45113869e-02 -3.84528488e-01 3.16191107e-01 3.06475252e-01
1.04162931e+00 2.00705856e-01 -8.79842564e-02 -1.29453495e-01
-7.01751709e-02 5.61657369e-01 4.88769531e-01 5.47327101e-01
-3.87727231e-01 9.56690907e-01 5.61680675e-01 -3.10829282e-01
-1.51935780e+00 -9.49654698e-01 -8.95000696e-02 8.36674273e-01
-2.24124864e-02 -7.37389699e-02 -9.18787360e-01 -5.91491699e-01
-6.52595684e-02 8.05243015e-01 -7.95834899e-01 -1.55380562e-01
-2.83184677e-01 -6.51946962e-01 8.18881631e-01 4.23490793e-01
4.80221301e-01 -9.13668811e-01 -9.88901436e-01 -2.83810161e-02
-2.98588783e-01 -1.60468698e+00 -5.18283546e-02 -3.64726603e-01
-7.72595584e-01 -1.21642530e+00 -5.54718792e-01 -7.20410824e-01
7.94832408e-01 6.23730481e-01 1.29493988e+00 2.51378119e-01
-4.38054800e-01 6.86821938e-01 -3.58833343e-01 -5.02019405e-01
-9.68834281e-01 -1.82184398e-01 -3.35891485e-01 -2.71038145e-01
2.93012887e-01 -2.50743926e-01 -5.72794795e-01 3.62134725e-01
-1.16971445e+00 2.91937619e-01 5.20929635e-01 7.64296949e-01
3.59849215e-01 -2.67279714e-01 6.78209186e-01 -8.62500727e-01
6.03871405e-01 -4.23035830e-01 -2.62191057e-01 2.69003093e-01
-7.61189103e-01 -9.45002139e-02 4.29186881e-01 -5.20112693e-01
-1.03995752e+00 -1.05372213e-01 9.14487615e-02 -3.85343671e-01
-2.74346501e-01 2.05450475e-01 2.06470326e-01 -1.97955474e-01
9.13180888e-01 -3.04656595e-01 4.99939233e-01 -2.83358365e-01
1.30034104e-01 6.20906949e-01 2.35036641e-01 -5.04686832e-01
5.26851058e-01 6.27724051e-01 -5.34446025e-03 -9.35681403e-01
-8.52323472e-01 4.44311611e-02 -2.91564435e-01 -3.86884481e-01
6.78644776e-01 -9.10171688e-01 -1.56744286e-01 4.37062085e-01
-9.10171270e-01 -3.62971097e-01 -4.32567179e-01 5.43304026e-01
-3.46527368e-01 2.51414388e-01 -4.89733964e-01 -5.20066619e-01
-6.17288537e-02 -1.28753507e+00 9.29032981e-01 2.53426373e-01
-2.72060156e-01 -7.90934086e-01 1.04293479e-02 8.91814947e-01
5.30082047e-01 7.89811194e-01 1.11453056e+00 -5.36358416e-01
-3.29521149e-01 -2.59334832e-01 -6.66800976e-01 3.47400814e-01
1.09112389e-01 3.02992016e-01 -1.17826545e+00 -2.48688966e-01
-2.29240768e-02 -9.50580895e-01 5.90307415e-01 1.29326522e-01
6.39032662e-01 1.07710041e-01 3.65522541e-02 4.38227616e-02
1.54209709e+00 1.21094577e-01 4.19261903e-01 4.77720380e-01
6.16566539e-01 9.96792614e-01 4.69733775e-01 -1.50524884e-01
7.31301665e-01 6.14739239e-01 4.69216794e-01 -1.43904433e-01
-6.27609313e-01 -3.26586366e-01 7.69089013e-02 4.29357052e-01
3.60298991e-01 -1.41057521e-01 -1.18106377e+00 6.36895001e-01
-1.37846506e+00 -7.35002279e-01 -2.66338121e-02 2.30984163e+00
5.77086091e-01 2.04536796e-01 2.04065263e-01 7.78890252e-02
7.10633934e-01 6.28513172e-02 -5.39421558e-01 -4.62728649e-01
-2.76246518e-01 -2.60740310e-01 2.99101830e-01 1.86110675e-01
-7.16168821e-01 7.81345844e-01 7.02821684e+00 6.77323222e-01
-1.38345456e+00 1.74943537e-01 7.88967490e-01 -1.13140471e-01
-3.57586652e-01 -1.18595615e-01 -5.12645543e-01 5.14298201e-01
1.06098056e+00 2.62182385e-01 1.42581195e-01 5.85333467e-01
2.02701360e-01 -3.75137657e-01 -1.21055269e+00 9.61535573e-01
4.72463369e-01 -1.16191328e+00 9.98188183e-02 1.51068211e-01
6.83798075e-01 2.41854653e-01 1.43225506e-01 -6.14053104e-03
1.38424799e-01 -1.26181519e+00 9.54680622e-01 -3.04035842e-02
9.23813105e-01 -3.52754265e-01 3.79323304e-01 1.40396863e-01
-7.33651459e-01 6.82539046e-02 -1.17393970e-01 1.01887740e-01
1.58255436e-02 3.48293364e-01 -1.02248549e+00 5.85245341e-03
4.85593855e-01 1.27145290e-01 -9.63202178e-01 8.82455528e-01
7.44292662e-02 6.88060701e-01 -2.82985270e-01 2.72796839e-01
2.08760872e-01 -1.02838770e-01 1.77247018e-01 9.19239700e-01
1.77390933e-01 -4.57592577e-01 -1.06663570e-01 6.65632665e-01
-1.84913233e-01 1.69930011e-01 -9.22645092e-01 -3.85453105e-01
6.56672239e-01 1.03370702e+00 -1.07588828e+00 4.05124854e-03
-7.55865574e-01 5.89034855e-01 4.47854787e-01 2.55656004e-01
-5.63319266e-01 1.35295317e-01 2.03129366e-01 2.13319197e-01
2.52844784e-02 -9.92344618e-02 -5.84560752e-01 -8.38796556e-01
1.87749162e-01 -1.21663892e+00 2.96282321e-01 -9.30220306e-01
-1.15751243e+00 6.73214912e-01 -5.70115559e-02 -1.16464114e+00
-2.84991086e-01 -5.27248383e-01 -1.57209903e-01 7.79844701e-01
-1.32499564e+00 -1.42612374e+00 -5.59161246e-01 1.07975625e-01
6.70624435e-01 -1.46680012e-01 7.52080798e-01 1.84481934e-01
-3.73399884e-01 7.07541347e-01 8.59691799e-02 -4.98365909e-02
7.05683351e-01 -8.84020388e-01 4.82936591e-01 5.44089437e-01
1.94114283e-01 3.80351335e-01 7.94854820e-01 -4.26329583e-01
-1.07534981e+00 -6.98612750e-01 6.87740743e-01 -6.03724241e-01
3.37961435e-01 -1.79581434e-01 -7.64676392e-01 5.53684235e-01
1.05329730e-01 -1.88458309e-01 8.48407090e-01 -2.16219321e-01
-9.81009543e-01 6.08769283e-02 -1.57350588e+00 6.06308460e-01
8.76024544e-01 -4.93590325e-01 -3.76675427e-01 2.26346850e-01
6.59442365e-01 -2.60862529e-01 -8.21876884e-01 5.42655408e-01
7.47922182e-01 -1.20272124e+00 7.60569632e-01 -4.35142636e-01
9.33913708e-01 2.69195368e-03 -2.63085991e-01 -1.35998440e+00
9.91963446e-02 1.30355312e-03 4.07864183e-01 1.41228807e+00
7.31206298e-01 -6.32378101e-01 9.14654493e-01 9.90000725e-01
3.30089629e-01 -6.25084460e-01 -8.10868979e-01 -6.28669322e-01
2.28358015e-01 -3.91306758e-01 4.29583788e-01 1.15100217e+00
-6.57875240e-01 3.48320872e-01 -1.35047913e-01 -4.18099552e-01
7.35700309e-01 2.72272490e-02 7.47214258e-01 -1.34664965e+00
-4.89258282e-02 -2.83435643e-01 -4.39451009e-01 -3.24920565e-01
2.51177490e-01 -7.08638728e-01 -1.64025396e-01 -1.56894171e+00
4.22034174e-01 -7.23548174e-01 2.27527902e-01 2.82871604e-01
6.28975928e-02 8.45126927e-01 6.16128385e-01 3.48791897e-01
-1.22497067e-01 1.99400276e-01 1.30388212e+00 3.83119620e-02
7.56522045e-02 -3.11245203e-01 -1.06894302e+00 6.71088159e-01
9.28863645e-01 -3.35999846e-01 -6.15823507e-01 -7.50260651e-01
1.34702101e-01 -2.96467185e-01 3.76419365e-01 -9.77707744e-01
-7.23018199e-02 -7.22733214e-02 2.77541786e-01 -1.16166070e-01
6.47484839e-01 -8.02352846e-01 1.48781031e-01 2.96366632e-01
-3.82054955e-01 1.58906952e-01 4.68742132e-01 3.98748845e-01
-2.40961704e-02 -3.46021950e-01 8.85955215e-01 -1.22517757e-01
-6.28387809e-01 -1.72201514e-01 -1.80501461e-01 5.28536916e-01
1.05029368e+00 -4.50541764e-01 -7.53069878e-01 -4.26302522e-01
-3.39870125e-01 -1.64493665e-01 9.53520775e-01 6.88098490e-01
3.60632092e-01 -1.23770261e+00 -6.98446929e-01 1.16244875e-01
4.01524276e-01 -1.98867798e-01 2.26287439e-01 5.96979022e-01
-7.56316066e-01 1.99667484e-01 -4.64629740e-01 -5.76794088e-01
-1.43151867e+00 1.45186484e-01 2.10010156e-01 8.15230012e-02
-4.94912446e-01 5.96198499e-01 -5.42703979e-02 -4.22056407e-01
2.23418534e-01 9.54207033e-02 2.43024132e-03 2.36153081e-01
3.67445678e-01 3.89023185e-01 1.05205968e-01 -1.12939501e+00
-2.07748353e-01 4.84316379e-01 -4.13018391e-02 -4.45489794e-01
1.10777247e+00 4.80750650e-02 -4.30431440e-02 5.37089944e-01
1.05477202e+00 -7.60557652e-02 -1.30108523e+00 -1.42899752e-01
-1.09254666e-01 -6.19829118e-01 -2.05740497e-01 -8.88207614e-01
-1.00651085e+00 6.45510495e-01 7.77979076e-01 2.87722707e-01
9.17355359e-01 7.69687667e-02 5.17132640e-01 -5.41526824e-02
3.73038590e-01 -1.10940850e+00 2.32108727e-01 7.46726394e-02
8.69426906e-01 -1.60351562e+00 -5.01849980e-04 -4.92578208e-01
-8.67960632e-01 7.05194592e-01 8.27998996e-01 1.96982443e-01
3.49984497e-01 6.47154301e-02 4.48796928e-01 -1.56216159e-01
-5.29782951e-01 2.23642737e-02 -1.83910102e-01 7.27284908e-01
5.51699460e-01 -4.43959571e-02 -5.21943092e-01 -2.30740860e-01
-3.68046343e-01 -5.64110093e-02 6.74847960e-01 1.00793111e+00
-1.51025712e-01 -1.27143228e+00 -4.71668363e-01 5.49580157e-01
-5.83650768e-01 1.39009446e-01 -6.83441818e-01 1.14802849e+00
1.27724588e-01 1.28357065e+00 2.13967264e-01 -4.19396371e-01
2.45118663e-01 2.53022850e-01 6.31643057e-01 -5.34087420e-01
-4.16196555e-01 -3.81096125e-01 3.41519803e-01 -3.26959789e-01
-6.07479334e-01 -6.39131844e-01 -8.65670145e-01 -2.76783735e-01
-7.71263018e-02 -1.30744241e-02 1.19816434e+00 9.24874306e-01
4.75982338e-01 2.40242660e-01 4.74351317e-01 -6.10105336e-01
-6.38425827e-01 -1.06743884e+00 -2.58462131e-01 8.14546406e-01
2.10404530e-01 -7.51485169e-01 -4.77490038e-01 5.46456613e-02]
|
[11.22565746307373, 1.3554472923278809]
|
26fa6a0c-1fa9-44f8-a29d-356f530fed92
|
aspect-category-opinion-sentiment-extraction
| null | null |
https://ieeexplore.ieee.org/document/10013820
|
https://ieeexplore.ieee.org/document/10013820
|
Aspect-Category-Opinion-Sentiment Extraction Using Generative Transformer Model
|
Sentiment analysis is one of Natural Language Processing's applications that aims to process and extract sentiment information quickly and effectively. To expand upon the previous triplet extraction, that being aspect-opinion-sentiment triplets, Aspect-Category-Opinion-Sentiment (ACOS) quadruple extraction was created. There are several methods to extract quadruples, albeit with several limitations, such as their effectiveness towards implicit information and their overall low-performance score. This paper proposes a method of using BART-Aspect-Based-Sentiment-Analysis (BARTABSA), a sentiment analysis model that aims to unify the previous Aspect Based Sentiment Analysis subtask - namely Aspect-Opinion pair extraction and Aspect-Opinion-Sentiment triplet extraction, and solve them without changing the core algorithm or adding other models to it - to solve the ACOS subtask. After some modification to the data and the model's outer layer, the result shows significant and promising improvements over previous results.
|
['Ngoc Hong Tran', 'Quang Vinh Dinh', 'Cao Duy Hoang']
|
2023-01-18
| null | null | null |
rifv-2023-1
|
['aspect-based-sentiment-analysis', 'aspect-category-opinion-sentiment-quadruple']
|
['natural-language-processing', 'natural-language-processing']
|
[ 1.70540616e-01 2.43019268e-01 1.09145187e-01 -8.08203518e-01
-7.24730909e-01 -8.02599728e-01 6.84137166e-01 5.92174053e-01
-2.62637109e-01 5.04382014e-01 3.24452907e-01 -5.03024280e-01
5.67948222e-02 -7.26372898e-01 1.30587325e-01 -5.07311106e-01
2.85184175e-01 4.59979057e-01 1.26242921e-01 -8.83464515e-01
6.88198566e-01 3.67485136e-02 -1.79361975e+00 7.94382751e-01
4.84985262e-01 1.10799444e+00 -4.20633227e-01 9.38930333e-01
-8.87335956e-01 1.06697166e+00 -7.72782505e-01 -9.15062249e-01
2.35059559e-01 -3.91517669e-01 -1.02974522e+00 -6.82574138e-02
-1.51361540e-01 3.47652584e-01 1.02768087e+00 8.51651013e-01
1.94939092e-01 -1.55621842e-01 6.29731119e-01 -1.54935515e+00
-4.68030125e-01 3.44677240e-01 -7.16914952e-01 -2.82680601e-01
9.06232178e-01 -3.54525805e-01 1.35484946e+00 -1.00568938e+00
5.23285568e-01 1.05647624e+00 9.51528847e-01 1.66033313e-01
-5.58940828e-01 -2.35239148e-01 3.95147234e-01 -3.99050452e-02
-6.37658179e-01 -1.13219380e-01 6.66456521e-01 -2.74670333e-01
1.61891520e+00 6.50431931e-01 1.16603518e+00 5.07029831e-01
4.60119814e-01 1.16473293e+00 1.60589111e+00 -6.66752636e-01
2.48052835e-01 6.96074784e-01 8.54227364e-01 2.13449761e-01
3.66701156e-01 -5.04878461e-01 -7.30339587e-01 -1.68345764e-01
-2.08596408e-01 -1.74082994e-01 5.20309150e-01 -1.70963764e-01
-9.20229912e-01 8.99474919e-01 -2.78179377e-01 5.28733909e-01
-3.40538144e-01 -5.44324636e-01 7.59504497e-01 8.78388464e-01
9.05224323e-01 7.78249204e-01 -1.41954565e+00 -4.43410307e-01
-8.94861400e-01 3.59646112e-01 1.39360619e+00 1.02899349e+00
8.73685837e-01 -7.95389935e-02 -8.47093686e-02 5.67556798e-01
4.41813320e-01 7.22906053e-01 5.69721937e-01 -3.09402019e-01
3.27266067e-01 1.45724058e+00 -4.32946719e-02 -1.11121535e+00
-5.65288424e-01 -1.03103660e-01 -5.08954227e-01 1.87239930e-01
-2.01261759e-01 -3.84423733e-01 -1.09495652e+00 8.62175405e-01
4.67865944e-01 -7.62985885e-01 3.43216628e-01 5.69735587e-01
1.12432623e+00 3.95651251e-01 1.62330672e-01 -2.58566469e-01
2.06172490e+00 -1.24513781e+00 -8.36389959e-01 -5.35208642e-01
9.75599468e-01 -1.35918820e+00 1.08558071e+00 5.80628395e-01
-8.57794106e-01 -1.09700657e-01 -1.06340897e+00 -8.56093764e-02
-1.37702262e+00 -1.79737449e-01 1.19227910e+00 9.55573976e-01
-1.41498280e+00 -1.18667744e-02 -3.12853664e-01 -3.43487620e-01
3.04419603e-02 6.36832654e-01 -4.39099818e-01 4.12424237e-01
-1.06951189e+00 9.54132080e-01 8.61400291e-02 -1.33138552e-01
2.56345004e-01 -6.53660536e-01 -1.14261401e+00 -1.22743636e-01
4.51312691e-01 -1.04395783e+00 1.52019787e+00 -1.57082796e+00
-1.54648876e+00 9.22436655e-01 -6.95758641e-01 3.46287005e-02
-2.35923067e-01 -2.39726856e-01 -5.85188925e-01 -1.21513449e-01
5.15549839e-01 2.98799902e-01 8.58925641e-01 -1.26730859e+00
-9.50562179e-01 -6.09641016e-01 3.35661292e-01 3.69240612e-01
-4.70911413e-01 6.16405785e-01 -2.95193225e-01 -6.51581049e-01
1.13595903e-01 -8.30662549e-01 -5.62969625e-01 -8.07030380e-01
-3.17783773e-01 -1.68328047e-01 8.41033876e-01 -2.46411517e-01
1.20939171e+00 -1.66535950e+00 -2.27856830e-01 4.05925840e-01
-4.58491929e-02 3.67037177e-01 -3.08629125e-01 6.81642354e-01
-5.00492990e-01 2.33856082e-01 -3.17464709e-01 -6.38870835e-01
-2.77105365e-02 1.03746749e-01 -2.25647926e-01 -2.82029599e-01
5.37309110e-01 1.09656942e+00 -1.00816691e+00 -4.75934744e-01
4.48022410e-03 4.01982039e-01 -3.66580278e-01 -4.98777144e-02
-1.71571806e-01 -5.91409318e-02 -7.01021850e-01 1.03088415e+00
7.58913577e-01 1.99398503e-01 5.95150106e-02 -2.55942017e-01
-3.35143238e-01 4.04356986e-01 -1.22925401e+00 1.15615416e+00
-6.53818786e-01 4.88750488e-01 2.39308421e-02 -7.77622640e-01
9.12482917e-01 2.64870077e-01 8.10147762e-01 -7.23667324e-01
4.27009672e-01 2.54687816e-01 -3.41606230e-01 -6.52037382e-01
1.29085863e+00 -5.33504248e-01 -5.41816950e-01 7.35668361e-01
1.61832973e-01 -8.97927225e-01 7.97532856e-01 4.19737250e-01
7.91335642e-01 1.40794232e-01 8.03384304e-01 -4.50459570e-01
1.12554514e+00 5.77152967e-01 1.36173889e-01 1.29705533e-01
-4.91895080e-02 7.20438898e-01 9.28599834e-01 -5.80625474e-01
-6.95861578e-01 -2.43423834e-01 2.49524787e-01 1.09889352e+00
-2.75375694e-01 -1.20326793e+00 -5.86262167e-01 -1.05775344e+00
-2.58192211e-01 6.27571344e-01 -7.74979174e-01 3.52426529e-01
8.30102786e-02 -1.14809620e+00 -1.45520978e-02 3.02301049e-01
1.02527849e-01 -1.18651712e+00 -2.28940412e-01 1.13244049e-01
-2.13442683e-01 -1.07886446e+00 3.16662103e-04 4.08815593e-01
-6.76747501e-01 -1.20065415e+00 -3.36551994e-01 -8.24879825e-01
7.89108455e-01 3.08232874e-01 1.64152038e+00 -3.87302116e-02
1.03065230e-01 5.27250290e-01 -1.09390092e+00 -1.28937638e+00
-1.09538756e-01 2.39198983e-01 -2.59168476e-01 -6.34391978e-02
1.16906548e+00 -3.08659822e-01 -4.77771610e-01 -7.22209439e-02
-1.14376259e+00 -2.02643022e-01 5.94233036e-01 4.09339696e-01
5.13282835e-01 1.63714886e-01 3.53494048e-01 -1.44781721e+00
1.14883852e+00 -3.07471365e-01 -1.49022579e-01 1.84082370e-02
-9.52583313e-01 -2.56615132e-01 6.80872262e-01 2.38256574e-01
-8.70861471e-01 -1.16752945e-01 -4.85730141e-01 5.43367922e-01
-1.13752447e-01 8.63442123e-01 2.53338609e-02 5.21407090e-02
1.31511390e-01 -4.16165963e-03 -3.95133980e-02 4.75782268e-02
3.94820571e-01 8.84365499e-01 -1.85121223e-01 -1.99539348e-01
5.62652051e-01 6.37240887e-01 -2.57544890e-02 -7.58955002e-01
-1.30784905e+00 -1.23696756e+00 -6.64417267e-01 -1.85539439e-01
5.67214191e-01 -1.00447738e+00 -6.02914274e-01 7.15680897e-01
-9.66770709e-01 2.65095800e-01 -5.68291426e-01 1.74825624e-01
-3.02584201e-01 3.42735678e-01 -3.25536460e-01 -1.04559386e+00
-1.00342083e+00 -9.62749004e-01 1.49232769e+00 2.68883795e-01
-6.60140336e-01 -1.22921491e+00 3.44053179e-01 6.32938802e-01
6.01666749e-01 8.73472691e-02 4.06347036e-01 -8.68948519e-01
2.41605327e-01 -6.14151955e-01 -7.54161105e-02 6.85960829e-01
2.67764986e-01 3.05114061e-01 -1.00879395e+00 2.12095097e-01
4.00719732e-01 -6.02651127e-02 4.56867784e-01 9.62449238e-02
2.51927853e-01 -2.56051183e-01 1.58545330e-01 1.32459819e-01
1.48106349e+00 1.77295581e-01 6.03335619e-01 1.01421618e+00
3.22928935e-01 8.56481850e-01 1.24084437e+00 3.27369660e-01
8.11972618e-01 9.53457728e-02 1.88117549e-01 -1.98853195e-01
2.76389927e-01 3.68790537e-01 7.49281824e-01 1.49098635e+00
-7.10505769e-02 -6.92031672e-03 -6.30283177e-01 7.61993587e-01
-1.83526611e+00 -7.12538660e-01 -7.26041377e-01 1.47436082e+00
6.44922793e-01 2.86851317e-01 1.69264242e-01 4.48662519e-01
-1.46919176e-01 2.72302866e-01 -8.10182616e-02 -1.47901070e+00
-4.72519487e-01 5.88986099e-01 -5.64208068e-03 4.79212552e-01
-1.14961278e+00 1.12656176e+00 6.70216227e+00 5.01358688e-01
-7.49702156e-01 -2.37034652e-02 5.31828344e-01 1.06259085e-01
-7.57112026e-01 3.85697931e-01 -8.16026151e-01 -3.11270636e-02
7.04455554e-01 2.77319774e-02 -1.48630396e-01 9.41806614e-01
1.10848270e-01 -5.78178108e-01 -5.79880714e-01 4.83242244e-01
6.03335083e-01 -7.57988036e-01 3.50424141e-01 -4.21012402e-01
8.70969594e-01 -2.54559606e-01 -1.11111924e-02 4.84616786e-01
1.93543717e-01 -5.53215683e-01 5.98083913e-01 3.65495056e-01
3.20420235e-01 -8.79366457e-01 1.70520055e+00 -2.19945788e-01
-1.49078369e+00 1.35978207e-01 -1.67827886e-02 -4.45922464e-01
2.25004911e-01 1.12287152e+00 -5.38814425e-01 1.10455287e+00
1.03612125e+00 9.20600951e-01 -7.27081597e-01 4.61564273e-01
-4.12446141e-01 2.03503385e-01 -9.11277905e-02 -5.76937973e-01
5.03184915e-01 -7.02304065e-01 5.33264935e-01 1.46729982e+00
1.73273042e-01 -9.57645401e-02 -3.02097082e-01 4.56500091e-02
6.45643651e-01 6.91469014e-01 -5.82529247e-01 -4.17544991e-02
-3.64964575e-01 1.99516571e+00 -1.04857075e+00 -7.22731829e-01
-7.85525024e-01 7.02498734e-01 -1.56743661e-01 1.12341642e-01
-2.80330092e-01 -8.28357935e-01 6.44718826e-01 -2.28313729e-01
4.42308277e-01 1.94622707e-02 -8.90216410e-01 -1.35297906e+00
3.41922283e-01 -1.46394289e+00 2.94019282e-01 -9.87590730e-01
-1.20572686e+00 9.92951393e-01 -9.81043950e-02 -1.26064980e+00
-4.65656608e-01 -1.03653860e+00 -6.09717309e-01 6.97027981e-01
-1.74593699e+00 -1.62914598e+00 1.53176934e-02 6.67610168e-01
6.06387436e-01 -3.55900884e-01 1.04023015e+00 9.69599281e-03
-6.70913234e-02 3.09115320e-01 -4.43056107e-01 -1.53852314e-01
7.01331317e-01 -1.62411404e+00 4.34088737e-01 7.51674354e-01
-1.19779401e-01 7.31664121e-01 8.33602607e-01 -5.01201212e-01
-1.43699300e+00 -6.25588119e-01 1.80830491e+00 -9.83651936e-01
1.23528516e+00 -3.88491303e-01 -3.53150219e-01 4.71058875e-01
6.43575609e-01 -7.06914127e-01 1.20684218e+00 5.49783885e-01
-3.04512769e-01 -1.62643835e-01 -1.04782057e+00 8.03208172e-01
3.77602249e-01 -5.95743895e-01 -9.42639530e-01 2.11660847e-01
6.17918313e-01 2.28550080e-02 -8.93223822e-01 5.80286443e-01
7.36337662e-01 -1.23656499e+00 7.50229359e-01 -5.13792217e-01
6.25884473e-01 -4.92168814e-01 -1.50509596e-01 -1.28302634e+00
2.65433967e-01 -5.47673047e-01 1.29446670e-01 1.39403582e+00
1.00377667e+00 -7.98209369e-01 7.55087912e-01 7.06288099e-01
-4.82902713e-02 -9.71512496e-01 -3.53754848e-01 -2.29803398e-01
-1.56508788e-01 -7.94277430e-01 9.09585714e-01 8.75870526e-01
5.19996107e-01 1.06750011e+00 1.33869752e-01 -2.18584731e-01
3.78084667e-02 4.13768381e-01 1.03447413e+00 -9.94007170e-01
7.35455453e-02 -5.37373126e-01 -3.59578937e-01 -3.56530517e-01
-1.45521790e-01 -4.28153604e-01 -1.02232665e-01 -1.90185773e+00
1.79202244e-01 -7.73492977e-02 -2.01399565e-01 7.25532949e-01
-4.01679605e-01 5.08753002e-01 2.33296886e-01 -2.65754044e-01
-9.39913750e-01 3.78032804e-01 1.19424057e+00 -1.63579792e-01
-2.81741649e-01 3.74173492e-01 -1.53140771e+00 9.36357200e-01
7.72457600e-01 -6.12677574e-01 -5.23902953e-01 -5.10349311e-02
1.35391092e+00 -6.35681927e-01 -4.25146788e-01 -5.88078678e-01
2.35827491e-01 1.26409709e-01 -6.01261891e-02 -1.05142164e+00
3.05746078e-01 -1.02371240e+00 -2.66191244e-01 9.18362215e-02
2.50400126e-01 6.51549578e-01 2.63799459e-01 -4.21642549e-02
-9.61568177e-01 -4.69931751e-01 1.05109783e-02 -2.00264052e-01
-7.23249912e-01 -8.90144035e-02 -8.85128498e-01 -1.36541262e-01
1.00177467e+00 -3.55892181e-01 -2.09319293e-01 -3.19378197e-01
-6.10123038e-01 1.78670183e-01 4.47166413e-01 5.64158678e-01
3.78750324e-01 -8.74321938e-01 -4.65813518e-01 3.82364511e-01
3.82494360e-01 3.24674062e-02 -1.00964271e-02 9.64008391e-01
-4.13462520e-01 6.01782382e-01 6.18225783e-02 -1.56321973e-01
-1.46747911e+00 3.70686918e-01 -1.50852859e-01 -9.74623263e-01
1.57824442e-01 6.55145347e-01 -3.55536371e-01 -1.08240879e+00
-5.24130702e-01 -3.77545476e-01 -1.07105231e+00 7.91171551e-01
4.67788666e-01 -9.24441144e-02 5.45658231e-01 -9.71250296e-01
-4.87273932e-01 1.04174590e+00 -1.74882069e-01 -1.40818164e-01
1.44972229e+00 -3.20227146e-01 -9.27801192e-01 5.27750015e-01
1.15621603e+00 4.75896120e-01 1.69665329e-02 2.73727477e-01
2.63145536e-01 -7.37291053e-02 -2.53467351e-01 -9.13281679e-01
-1.11187351e+00 3.67086083e-01 -7.42982849e-02 9.03024614e-01
1.49034441e+00 -2.74194568e-01 7.29738057e-01 4.58317757e-01
9.77616683e-02 -1.40621293e+00 -1.65207669e-01 1.13267910e+00
6.50078773e-01 -1.51030445e+00 5.59021831e-01 -5.77057421e-01
-1.12213242e+00 1.09215796e+00 3.96337390e-01 -3.40673774e-02
1.30095291e+00 5.44571459e-01 8.37852478e-01 -7.42047012e-01
-8.25491130e-01 -5.06204188e-01 3.91377300e-01 6.05015099e-01
7.44710386e-01 8.68070498e-02 -5.83960056e-01 9.08626080e-01
-7.98560202e-01 -1.72864333e-01 2.87657529e-01 1.36616766e+00
-1.23305701e-01 -1.51562500e+00 -3.16248387e-01 7.21029282e-01
-7.86310494e-01 -5.34161568e-01 -9.69824255e-01 7.55754411e-01
3.06115393e-02 1.57040775e+00 -1.07310556e-01 -6.12138629e-01
7.90600955e-01 -4.70752306e-02 -2.47951582e-01 -6.17666245e-01
-1.46236074e+00 -3.53357606e-02 4.93477821e-01 -6.19955957e-01
-1.28376245e+00 -7.54442513e-01 -8.99229467e-01 -2.11549982e-01
-5.15709341e-01 6.17036343e-01 1.18995702e+00 1.19209778e+00
4.70467359e-01 4.66705173e-01 9.15078104e-01 -6.40423834e-01
1.54125929e-01 -8.95167828e-01 -5.62484324e-01 3.67857814e-01
2.93959826e-01 9.88540798e-03 -4.02352095e-01 4.71917801e-02]
|
[11.296218872070312, 6.78853702545166]
|
ea513108-d2b1-4344-bcb6-5dd62980604e
|
gesture-recognition-with-mmwave-wi-fi-access
|
2306.17062
| null |
https://arxiv.org/abs/2306.17062v1
|
https://arxiv.org/pdf/2306.17062v1.pdf
|
Gesture Recognition with mmWave Wi-Fi Access Points: Lessons Learned
|
In recent years, channel state information (CSI) at sub-6 GHz has been widely exploited for Wi-Fi sensing, particularly for activity and gesture recognition. In this work, we instead explore mmWave (60 GHz) Wi-Fi signals for gesture recognition/pose estimation. Our focus is on the mmWave Wi-Fi signals so that they can be used not only for high data rate communication but also for improved sensing e.g., for extended reality (XR) applications. For this reason, we extract spatial beam signal-to-noise ratios (SNRs) from the periodic beam training employed by IEEE 802.11ad devices. We consider a set of 10 gestures/poses motivated by XR applications. We conduct experiments in two environments and with three people.As a comparison, we also collect CSI from IEEE 802.11ac devices. To extract features from the CSI and the beam SNR, we leverage a deep neural network (DNN). The DNN classifier achieves promising results on the beam SNR task with state-of-the-art 96.7% accuracy in a single environment, even with a limited dataset. We also investigate the robustness of the beam SNR against CSI across different environments. Our experiments reveal that features from the CSI generalize without additional re-training, while those from beam SNRs do not. Therefore, re-training is required in the latter case.
|
['Jeroen Famaey', 'Rafael Berkvens', 'Nabeel Nisar Bhat']
|
2023-06-29
| null | null | null | null |
['pose-estimation', 'gesture-recognition']
|
['computer-vision', 'computer-vision']
|
[ 3.65179837e-01 -4.84012067e-02 -3.31635058e-01 -3.99521202e-01
-9.82640684e-01 -2.18005627e-01 2.54325867e-01 -6.02870941e-01
-5.36317468e-01 6.83508277e-01 4.08321798e-01 -3.81851047e-01
-2.60297954e-01 -8.14592123e-01 -3.93234253e-01 -1.16874588e+00
-3.51517439e-01 -1.16368815e-01 -4.66161408e-02 9.26074758e-02
-3.90898943e-01 3.17844808e-01 -1.19936919e+00 -5.13337590e-02
3.10611218e-01 1.39015675e+00 1.69166803e-01 8.43531966e-01
2.67433703e-01 3.94725472e-01 -9.09689605e-01 -5.90177514e-02
3.63948286e-01 -1.67157203e-01 1.56038348e-02 -5.11767685e-01
1.40834108e-01 -7.90636182e-01 -1.12188530e+00 5.30238748e-01
1.17401767e+00 -1.23122953e-01 5.03453374e-01 -9.64916885e-01
1.73930585e-01 8.37415636e-01 -3.95821631e-01 1.12264089e-01
5.35896480e-01 2.16896728e-01 5.10033369e-01 -2.67248601e-01
2.99439818e-01 8.28458726e-01 8.61266315e-01 5.48170567e-01
-3.32653165e-01 -1.32638502e+00 -2.47888491e-01 1.60813048e-01
-1.52660871e+00 -4.51648712e-01 6.55580997e-01 4.07583565e-02
4.91290420e-01 4.95498002e-01 5.49032152e-01 1.73165834e+00
1.65005445e-01 6.77486658e-01 8.80014837e-01 -3.54072303e-01
3.06469679e-01 -4.33607101e-01 3.83206680e-02 5.66708088e-01
3.03660750e-01 4.83113974e-01 -9.83976960e-01 4.09792103e-02
7.62899041e-01 9.88655686e-02 -4.76543009e-01 2.86120653e-01
-1.14452541e+00 1.59532249e-01 6.27068281e-01 7.34395564e-01
-3.16338241e-01 6.77932084e-01 -4.22819316e-01 3.57991368e-01
-1.69357918e-02 1.42618120e-01 -3.73634547e-01 -5.93149364e-01
-4.82764482e-01 1.10552579e-01 7.23436058e-01 1.00331593e+00
1.12811334e-01 7.25183114e-02 -3.36150467e-01 4.43898082e-01
6.17908478e-01 1.51762390e+00 2.79603284e-02 -5.56932151e-01
9.58671212e-01 -3.33758861e-01 2.16358334e-01 -7.23914206e-01
-1.11697555e+00 -1.35008252e+00 -1.00831580e+00 -2.72910863e-01
7.66272366e-01 -6.37847245e-01 -9.99629676e-01 1.90748894e+00
1.06359035e-01 5.42298496e-01 8.18772912e-02 9.14137602e-01
8.48133504e-01 1.99963555e-01 -1.00761995e-01 -5.98784722e-02
1.09375370e+00 -1.36394709e-01 -6.67603970e-01 -1.98649555e-01
5.44403613e-01 -3.55743676e-01 5.71775854e-01 7.61346161e-01
-6.10212684e-01 -5.33441067e-01 -1.32284820e+00 7.13746727e-01
7.23769516e-02 1.59074709e-01 7.96829998e-01 1.62990773e+00
-4.12140816e-01 2.55712807e-01 -1.14773738e+00 -1.99787036e-01
4.58918750e-01 5.21249890e-01 2.63408631e-01 -3.23021621e-01
-1.37638128e+00 3.23101938e-01 -2.59453028e-01 4.07478184e-01
-4.54856604e-01 -6.11174881e-01 -4.44694996e-01 7.01761618e-02
1.20332286e-01 -3.90727580e-01 1.21585464e+00 -1.08860172e-01
-1.72439337e+00 -1.27624571e-01 -7.32349902e-02 -2.82983929e-01
3.93754959e-01 -2.07134023e-01 -9.98020053e-01 1.22194916e-01
-3.99174750e-01 -2.92670615e-02 5.87855995e-01 -1.05188382e+00
-7.50176489e-01 -5.62624753e-01 2.10840151e-01 -2.98117753e-03
-4.91056502e-01 -3.42657596e-01 -4.10815537e-01 -5.78264713e-01
8.81549597e-01 -1.11960828e+00 1.59683861e-02 -3.81019503e-01
-4.43920314e-01 3.11778188e-01 3.87776852e-01 -6.13872111e-01
1.22745275e+00 -2.07043266e+00 -4.04930413e-01 7.51536250e-01
8.79942477e-02 -9.61118490e-02 -3.60682048e-02 1.43341929e-01
4.33910400e-01 -1.27794132e-01 1.70729280e-01 -3.23317945e-01
1.10490575e-01 -1.07014664e-02 -2.61379443e-02 6.59037709e-01
-4.85842884e-01 6.68694973e-01 -7.23962605e-01 4.26823497e-02
1.46534726e-01 4.22144711e-01 -3.56556714e-01 4.63237837e-02
4.45813775e-01 1.11094391e+00 -9.77050245e-01 9.07919347e-01
9.83734369e-01 1.58368781e-01 4.58745092e-01 -5.94129741e-01
1.11270368e-01 1.12664893e-01 -1.15039229e+00 1.80381334e+00
-1.03067386e+00 4.28304434e-01 1.56152815e-01 -8.28446746e-01
7.06464946e-01 4.09306407e-01 7.57414997e-01 -1.27181304e+00
4.38361108e-01 2.55270302e-01 2.53873080e-01 -7.79810131e-01
-1.10057138e-01 -1.12831414e-01 -3.94609690e-01 6.41456783e-01
2.81610209e-02 2.40816206e-01 -3.64012897e-01 -6.91610277e-02
1.61362803e+00 -1.66511789e-01 -1.00951061e-01 1.75313726e-01
1.08114645e-01 -6.15427315e-01 2.79701501e-01 1.51585662e+00
-4.97108139e-02 3.53564590e-01 -2.96010435e-01 9.19110626e-02
-2.41479099e-01 -1.40130901e+00 -3.20444345e-01 8.71284306e-01
5.03129661e-01 -8.51774961e-02 -3.02108973e-01 -4.86237437e-01
-1.95663031e-02 4.72614259e-01 -2.11090982e-01 -9.04068872e-02
-7.94815063e-01 -9.39737082e-01 1.12081218e+00 7.10271597e-01
1.05842531e+00 -4.38305557e-01 -8.02452445e-01 2.50703812e-01
-2.46989578e-01 -1.29324186e+00 1.98334634e-01 3.79995346e-01
-3.27870727e-01 -6.28500223e-01 -7.98867822e-01 -9.47975814e-02
1.55689150e-01 4.06745374e-01 6.11481011e-01 -2.14785859e-02
3.39802541e-02 6.64526999e-01 -5.49860835e-01 -5.75092316e-01
1.77721828e-01 2.19438910e-01 4.65040207e-01 -3.71649340e-02
2.53233254e-01 -7.63572991e-01 -7.26755261e-01 6.67837560e-01
-5.11801183e-01 -3.29096466e-01 9.82344687e-01 5.26198924e-01
-1.12841964e-01 1.82389334e-01 5.97678840e-01 -3.20805639e-01
2.69253522e-01 -3.85698617e-01 -2.75868416e-01 5.51888905e-02
-2.97869146e-02 -6.68761581e-02 1.66397959e-01 -3.02954525e-01
-1.28871942e+00 -4.26946282e-02 -7.57178485e-01 3.90930772e-01
-1.86619133e-01 7.50034869e-01 -7.39262640e-01 -3.74052346e-01
7.66246080e-01 -2.04087198e-01 -6.73354626e-01 -4.60611701e-01
1.06262520e-01 1.40985131e+00 6.09093130e-01 -3.63831550e-01
9.98834252e-01 6.12979412e-01 2.98159242e-01 -1.11729026e+00
-7.06878543e-01 -6.29516065e-01 -2.69156277e-01 -3.87863278e-01
4.68174487e-01 -1.01098919e+00 -9.91079986e-01 5.30459166e-01
-8.04352343e-01 -4.49893117e-01 4.60131705e-01 1.10733092e+00
-1.83847651e-01 1.06165990e-01 -5.13702929e-01 -1.12041390e+00
-8.25421661e-02 -8.10850084e-01 9.76960540e-01 2.72505045e-01
-1.07949153e-01 -3.95191789e-01 -1.81139737e-01 5.19240320e-01
9.19111967e-01 2.00830340e-01 7.07704723e-02 -1.05228946e-01
-6.05930328e-01 -5.38182259e-01 7.77989626e-02 -4.60828424e-01
1.01187840e-01 -1.12184942e+00 -1.44549882e+00 -2.28089243e-01
8.36260095e-02 -9.38725546e-02 8.27411711e-01 6.66072965e-01
1.23829710e+00 1.64283253e-02 -7.61577487e-01 1.03118134e+00
8.12759161e-01 3.08597296e-01 7.48418510e-01 6.61256239e-02
6.04808450e-01 -1.31495535e-01 6.59091294e-01 6.68071628e-01
6.58883527e-03 9.43451345e-01 2.48477146e-01 1.56919155e-02
-1.65421486e-01 -5.43192364e-02 2.16651019e-02 1.79786310e-01
-3.19419861e-01 -7.84067869e-01 -7.93747663e-01 -2.38194674e-01
-1.47215760e+00 -7.59844542e-01 -1.55980378e-01 2.17409682e+00
2.25433722e-01 3.56104344e-01 -1.31346300e-01 4.17932272e-01
1.49111405e-01 7.65489265e-02 -6.24745250e-01 5.37279606e-01
-7.15576485e-02 6.28522098e-01 1.10983729e+00 4.02319491e-01
-1.23810446e+00 2.80094445e-01 4.97662497e+00 6.89700305e-01
-1.35086548e+00 4.63815123e-01 2.84909993e-01 -2.46456176e-01
-2.02675819e-01 -7.06683695e-01 -5.56895554e-01 2.18247622e-01
8.12570691e-01 6.97419941e-01 2.30859369e-01 4.14789945e-01
1.12826385e-01 -3.61555666e-01 -8.78442764e-01 1.45932865e+00
-2.16412887e-01 -9.89343107e-01 -8.30777764e-01 2.78554738e-01
2.78259456e-01 8.53022411e-02 3.19538683e-01 3.21690172e-01
-1.23029016e-01 -9.44454968e-01 7.21892297e-01 5.67703128e-01
1.18284488e+00 -5.85201800e-01 1.06694794e+00 4.41320986e-01
-1.32190347e+00 -3.15139890e-01 5.79016805e-02 -2.02591121e-01
3.66362393e-01 1.01892817e+00 -7.59407282e-01 8.09853375e-01
6.01536989e-01 1.53435573e-01 -6.00761324e-02 9.30893481e-01
-4.48474646e-01 9.77555931e-01 -9.33951318e-01 -3.08661431e-01
-1.67971373e-01 3.48521054e-01 3.80764067e-01 1.21837080e+00
8.31905901e-01 6.11641943e-01 -9.32646617e-02 2.39527196e-01
6.87079225e-03 -6.19935572e-01 -2.80391634e-01 3.81324112e-01
6.18572295e-01 8.99300516e-01 -6.93573773e-01 1.46878749e-01
-2.60790318e-01 6.63635790e-01 -7.05522239e-01 8.25781643e-01
-7.74971545e-01 -4.27721113e-01 5.94695091e-01 -3.49170417e-02
3.29818606e-01 -8.66826296e-01 -4.46687609e-01 -9.29799378e-01
2.72255158e-03 -2.88463175e-01 4.52718176e-02 -2.37899840e-01
-9.40646887e-01 3.54063541e-01 -7.25315362e-02 -1.40084541e+00
-4.94920015e-01 -5.27832925e-01 -2.04478487e-01 7.56522954e-01
-1.38576472e+00 -1.15404773e+00 -8.66458774e-01 3.93732876e-01
-9.81721282e-02 1.67856857e-01 7.34241247e-01 7.45194674e-01
-2.97840685e-01 1.18133473e+00 2.57248938e-01 4.80441749e-01
2.57424861e-01 -7.52582192e-01 4.92551118e-01 6.81528389e-01
1.92553669e-01 7.33007312e-01 8.25639784e-01 -7.44771719e-01
-1.92034996e+00 -8.03404808e-01 3.50127131e-01 -2.45527387e-01
2.84730136e-01 -8.21007371e-01 3.08198365e-03 2.61300445e-01
-4.99978453e-01 1.93544269e-01 6.93962932e-01 2.64390290e-01
1.08125443e-02 -3.25393528e-01 -1.32629287e+00 4.33911115e-01
1.75648439e+00 -2.61322170e-01 1.41179621e-01 1.13226883e-01
1.78907245e-01 -6.66449428e-01 -6.13166571e-01 4.92959857e-01
1.32366574e+00 -7.01901793e-01 1.36538851e+00 -1.89878251e-02
-4.15347725e-01 -1.00905374e-01 -6.16973341e-01 -1.11414158e+00
-1.04093216e-01 -4.87086236e-01 -4.20050561e-01 8.85770977e-01
3.18438590e-01 -6.99331105e-01 1.17878819e+00 2.44246736e-01
5.51860146e-02 -4.69438046e-01 -1.40017378e+00 -9.94693160e-01
-4.61790800e-01 -1.36398649e+00 7.90086329e-01 3.00634116e-01
9.12210532e-03 4.91214804e-02 -5.35012007e-01 7.19218314e-01
9.35830057e-01 -3.01284462e-01 8.59173894e-01 -1.01730287e+00
-7.98807859e-01 -1.52035326e-01 -4.01285172e-01 -1.88806784e+00
-2.88201630e-01 -5.03566921e-01 2.27412522e-01 -1.34579182e+00
-3.33610922e-01 -1.00232172e+00 -4.02529597e-01 1.68836102e-01
2.16319084e-01 5.64304531e-01 1.11502305e-01 -1.67688489e-01
-4.02705669e-01 4.25358206e-01 8.08915317e-01 -4.11938161e-01
-3.01271439e-01 8.30913544e-01 -4.85749364e-01 6.82583630e-01
9.09721732e-01 -3.77235919e-01 -1.80039331e-01 -4.87972289e-01
1.06552258e-01 4.43354845e-01 2.60885417e-01 -1.77983904e+00
5.80953918e-02 -1.13330044e-01 6.29920006e-01 -4.13750142e-01
6.87430024e-01 -9.48341072e-01 2.18338385e-01 6.68651760e-01
-5.72262481e-02 -9.23925638e-01 -1.96481481e-01 6.43859327e-01
4.29278523e-01 1.08257048e-02 2.13205561e-01 2.67245203e-01
-3.36947620e-01 2.18444481e-01 -3.64383310e-01 -5.74925065e-01
3.15806031e-01 -5.44386990e-02 -3.19018692e-01 -9.97103751e-01
-5.17337263e-01 -1.29243076e-01 -2.93170661e-01 2.28448212e-01
5.09679675e-01 -1.23919034e+00 -3.58302623e-01 1.39487416e-01
1.05194107e-01 -2.77872205e-01 2.90533662e-01 9.20796812e-01
-1.47694737e-01 7.17725158e-01 1.81156993e-01 -8.38397026e-01
-1.07591391e+00 -2.50919938e-01 4.52760458e-01 -6.95420057e-02
-5.10538042e-01 8.72007310e-01 -1.28684893e-01 -1.41443640e-01
7.00389028e-01 -7.59663105e-01 -1.29888028e-01 -4.34352845e-01
8.47964227e-01 2.81840980e-01 3.50939661e-01 -2.95890808e-01
-5.62657595e-01 7.96584964e-01 6.80057168e-01 -6.79370046e-01
1.26033783e+00 -2.50243992e-01 9.94939923e-01 5.71408048e-02
9.77870405e-01 3.27777088e-01 -1.20388281e+00 -2.91343421e-01
-1.74374297e-01 -6.76898956e-01 2.66561627e-01 -1.04200733e+00
-1.22638285e+00 5.86054742e-01 1.44800198e+00 3.04852687e-02
1.16144359e+00 2.51262873e-01 1.01618433e+00 8.83357584e-01
1.18836665e+00 -7.01407492e-01 -8.60956684e-02 3.24100584e-01
4.48995531e-01 -7.33842015e-01 -3.87065783e-02 -2.73404926e-01
3.53433341e-02 9.54166949e-01 8.58222470e-02 5.23211539e-01
9.24543440e-01 7.06335783e-01 2.95253783e-01 -3.89652210e-03
3.12874377e-01 -3.46598923e-01 7.24808872e-03 1.04992044e+00
2.47706741e-01 3.32246631e-01 -1.41011447e-01 1.21123266e+00
-5.10756671e-01 1.62670657e-01 1.33791551e-01 1.25808430e+00
-4.68606234e-01 -8.85910511e-01 -5.46827912e-01 7.03072429e-01
-2.72137970e-01 3.34635735e-01 7.82066509e-02 4.23793614e-01
2.85989761e-01 1.51115334e+00 -3.52742881e-01 -9.51690912e-01
4.79937941e-01 -4.48431611e-01 9.05588150e-01 2.22096909e-02
-2.16556489e-01 6.10651597e-02 3.85375828e-01 -6.75598264e-01
-4.55660462e-01 -6.83247626e-01 -1.28530169e+00 -1.08262539e-01
-4.01970267e-01 -1.50665253e-01 1.03800321e+00 1.31105566e+00
-6.30982816e-02 8.71865213e-01 5.86572826e-01 -9.99022782e-01
-5.05776703e-01 -1.17297423e+00 -8.62320006e-01 -3.93547453e-02
3.57181191e-01 -9.19938028e-01 -2.83312917e-01 -6.33396149e-01]
|
[6.600583553314209, 0.7499961256980896]
|
94beb042-42b9-4c5a-ab6a-c86baed96797
|
learning-multimodal-data-augmentation-in
|
2212.14453
| null |
https://arxiv.org/abs/2212.14453v2
|
https://arxiv.org/pdf/2212.14453v2.pdf
|
Learning Multimodal Data Augmentation in Feature Space
|
The ability to jointly learn from multiple modalities, such as text, audio, and visual data, is a defining feature of intelligent systems. While there have been promising advances in designing neural networks to harness multimodal data, the enormous success of data augmentation currently remains limited to single-modality tasks like image classification. Indeed, it is particularly difficult to augment each modality while preserving the overall semantic structure of the data; for example, a caption may no longer be a good description of an image after standard augmentations have been applied, such as translation. Moreover, it is challenging to specify reasonable transformations that are not tailored to a particular modality. In this paper, we introduce LeMDA, Learning Multimodal Data Augmentation, an easy-to-use method that automatically learns to jointly augment multimodal data in feature space, with no constraints on the identities of the modalities or the relationship between modalities. We show that LeMDA can (1) profoundly improve the performance of multimodal deep learning architectures, (2) apply to combinations of modalities that have not been previously considered, and (3) achieve state-of-the-art results on a wide range of applications comprised of image, text, and tabular data.
|
['Andrew Gordon Wilson', 'Anshumali Shrivastava', 'Mu Li', 'Aston Zhang', 'Xingjian Shi', 'Zhiqiang Tang', 'Zichang Liu']
|
2022-12-29
| null | null | null | null |
['multimodal-deep-learning']
|
['natural-language-processing']
|
[ 5.96423149e-01 -6.42833710e-02 -1.14174776e-01 -3.67605239e-01
-8.81931007e-01 -8.19943011e-01 8.64926994e-01 2.23951697e-01
-4.62819546e-01 6.88982427e-01 3.32965463e-01 -2.90866733e-01
2.21629322e-01 -4.60838109e-01 -8.45289707e-01 -4.80902135e-01
2.85533875e-01 4.43898499e-01 -3.01338285e-01 -2.68879294e-01
-2.77568340e-01 3.86602342e-01 -1.49226141e+00 4.70849127e-01
4.46431279e-01 1.06483507e+00 -5.37594818e-02 6.27526402e-01
-1.41709045e-01 6.41542554e-01 -2.24247381e-01 -4.14139420e-01
6.95621148e-02 -5.54523289e-01 -9.01656926e-01 4.35436845e-01
5.73802292e-01 -3.51999998e-01 -5.96800745e-01 8.33400130e-01
2.21155956e-01 5.81377745e-02 6.74346745e-01 -1.60667610e+00
-8.40829849e-01 4.77879703e-01 -4.78037477e-01 -3.03262234e-01
3.62318367e-01 5.43149181e-05 1.05779934e+00 -7.45620549e-01
5.84676147e-01 1.17931056e+00 3.45309079e-01 6.43006682e-01
-1.32017934e+00 -2.73339063e-01 2.37344950e-01 -7.33630639e-03
-1.04764819e+00 -5.76712847e-01 7.60383070e-01 -3.44174206e-01
5.19557118e-01 2.59034395e-01 3.59877378e-01 1.26518929e+00
-2.53179699e-01 8.78381371e-01 9.49058950e-01 -4.40578043e-01
-4.92074266e-02 2.13767409e-01 -2.11844265e-01 5.85679948e-01
-2.12200582e-01 -3.03604335e-01 -8.58343482e-01 -1.16750501e-01
5.42874396e-01 -1.34390814e-03 -1.59315377e-01 -6.40186667e-01
-1.67417133e+00 6.32579207e-01 3.18168759e-01 1.71752200e-01
-2.77864933e-01 2.01630041e-01 3.90060365e-01 4.19718623e-01
7.08111599e-02 3.91153455e-01 -4.31962490e-01 -1.61801443e-01
-4.95542109e-01 1.53634146e-01 5.37380993e-01 7.68177450e-01
8.66413593e-01 -7.76174068e-02 9.84418094e-02 8.32765043e-01
2.07488969e-01 6.93440199e-01 3.19319278e-01 -8.23277533e-01
5.27219176e-01 7.69144952e-01 7.90593624e-02 -7.30902433e-01
-4.46810305e-01 -9.14457366e-02 -1.00083542e+00 -2.53324620e-02
5.46084881e-01 -2.56907284e-01 -9.80462790e-01 2.20587659e+00
5.76873161e-02 -2.96577424e-01 2.12234706e-01 8.22137892e-01
1.00522804e+00 5.59079766e-01 1.23689376e-01 1.49256676e-01
1.22858071e+00 -7.49710381e-01 -6.88766062e-01 -4.83873665e-01
5.93543649e-01 -8.39463532e-01 1.09232402e+00 1.02936544e-01
-1.10703599e+00 -3.47975701e-01 -8.93468857e-01 -3.39525759e-01
-5.52786946e-01 5.99994808e-02 6.90987527e-01 4.21233624e-01
-8.33738923e-01 7.33978301e-02 -7.61758268e-01 -4.83650506e-01
3.64519179e-01 4.46182370e-01 -1.07058036e+00 -3.03044796e-01
-1.11790705e+00 9.47721183e-01 2.34089553e-01 -8.23310316e-02
-8.09520662e-01 -5.92280328e-01 -1.01528072e+00 9.88787711e-02
4.53334183e-01 -8.61786366e-01 1.20365763e+00 -1.30950999e+00
-1.16215301e+00 8.60696137e-01 -3.27762842e-01 -1.47762716e-01
3.16611648e-01 7.93741958e-04 -2.84887552e-01 9.83042717e-02
-1.72670096e-01 1.23627746e+00 8.30695152e-01 -1.38169837e+00
-5.49833357e-01 -3.37432563e-01 2.64689833e-01 4.87284720e-01
-8.66652310e-01 -7.80044496e-02 -5.52999914e-01 -3.98454487e-01
6.12693317e-02 -1.07789505e+00 2.55017299e-02 2.40144417e-01
-4.85846072e-01 -4.69797896e-03 9.47061181e-01 -5.53337872e-01
6.80188298e-01 -2.34851360e+00 5.61868131e-01 1.34824559e-01
1.75079808e-01 2.64400747e-02 -5.04062712e-01 5.59231818e-01
2.78765559e-02 9.77041200e-02 -4.11760360e-01 -6.29084170e-01
2.60489315e-01 3.00448656e-01 -3.18869382e-01 3.13138932e-01
4.05847549e-01 9.98287976e-01 -6.69816971e-01 -2.94121593e-01
2.72089183e-01 6.24014258e-01 -3.90317351e-01 1.27857000e-01
-4.06369805e-01 5.99335790e-01 -1.03994586e-01 6.18415296e-01
3.50108176e-01 -4.06834453e-01 1.31879628e-01 -5.13596475e-01
2.03450248e-01 1.26388758e-01 -1.05075061e+00 1.91558826e+00
-4.73848194e-01 8.96494389e-01 1.18446521e-01 -8.74068618e-01
4.81986433e-01 5.03746033e-01 6.95669174e-01 -6.28667951e-01
1.38730571e-01 4.76850867e-02 3.28282118e-02 -5.16080797e-01
4.86766249e-01 -1.81476295e-01 -2.27334738e-01 7.25828230e-01
2.61346638e-01 -1.00374781e-01 3.48007321e-01 2.56173730e-01
8.52328539e-01 -6.00610077e-02 2.69930298e-03 3.91948968e-01
4.01066303e-01 -7.52365813e-02 8.80972371e-02 6.46139443e-01
1.20845638e-01 7.76334584e-01 5.67743301e-01 -1.15179956e-01
-1.16538370e+00 -1.01445663e+00 3.72937024e-02 1.44925523e+00
9.40653533e-02 -1.37181416e-01 -4.60796148e-01 -7.04689264e-01
-8.31010267e-02 3.22888523e-01 -5.66630661e-01 -1.27849370e-01
-1.39756456e-01 -7.24559247e-01 7.06266940e-01 6.12124741e-01
5.09857059e-01 -7.27413833e-01 -2.33356118e-01 -1.51138872e-01
-4.25734937e-01 -1.59415627e+00 -4.79214996e-01 1.52083874e-01
-5.12646556e-01 -9.89523053e-01 -5.84179580e-01 -7.99024880e-01
9.06548262e-01 2.58400500e-01 8.47803533e-01 -1.16676077e-01
9.90216136e-02 8.51884782e-01 -2.63255090e-01 -2.36392185e-01
-4.92987603e-01 2.07932279e-01 -1.19957395e-01 3.94951493e-01
2.10119009e-01 -4.00223166e-01 -1.02346219e-01 2.23336741e-01
-1.34499657e+00 2.63563693e-01 6.46809340e-01 9.58348215e-01
3.90589058e-01 -1.93489388e-01 5.88748038e-01 -7.65270531e-01
4.97598231e-01 -3.72634709e-01 -1.22534640e-01 3.80420029e-01
-3.95481922e-02 9.34581682e-02 3.54863584e-01 -6.59544647e-01
-8.48913431e-01 4.60990727e-01 -2.97728013e-02 -3.07145625e-01
-4.01858598e-01 9.88373458e-01 -4.70784962e-01 -1.46152332e-01
5.30110717e-01 1.38020426e-01 3.07660192e-01 -3.38474393e-01
6.38379991e-01 7.67391622e-01 7.55806625e-01 -5.63928604e-01
8.06168318e-01 6.06139362e-01 2.22210988e-01 -8.91385019e-01
-6.77153885e-01 -1.57158658e-01 -7.47879326e-01 -1.19256645e-01
7.66633093e-01 -9.72471118e-01 -5.81427991e-01 4.83386934e-01
-9.83301461e-01 -2.49237269e-01 -3.21985111e-02 3.99310529e-01
-4.40107554e-01 4.06034142e-01 -3.32703620e-01 -4.66223747e-01
4.45532724e-02 -1.07300162e+00 9.54548240e-01 2.14537919e-01
-3.11294705e-01 -1.13726556e+00 -3.38313371e-01 6.16070151e-01
4.24165994e-01 2.12346256e-01 1.19938517e+00 -7.18619883e-01
-4.71463203e-01 -5.15864670e-01 -3.25542271e-01 3.93477321e-01
4.42504257e-01 1.10308267e-03 -1.13924408e+00 -2.87100315e-01
-5.30156076e-01 -8.21092665e-01 7.04391479e-01 3.39832553e-03
9.97222006e-01 -1.47900358e-01 -8.99594128e-02 3.69064003e-01
8.35191071e-01 -8.97398591e-02 4.08156961e-01 1.94965020e-01
7.86702394e-01 6.40040100e-01 1.45838425e-01 2.34112009e-01
5.70774376e-01 6.65362298e-01 7.15757668e-01 -4.51629013e-01
-1.84630558e-01 -2.77906299e-01 2.18713447e-01 5.22145808e-01
2.76781946e-01 -3.35049331e-01 -8.83629620e-01 5.84880710e-01
-1.91652441e+00 -7.02918112e-01 2.48263568e-01 2.16330218e+00
8.77697587e-01 -1.15329184e-01 1.71825215e-01 -2.08938271e-02
5.63154817e-01 -6.43022684e-03 -6.12702608e-01 -3.20390612e-02
-5.30236721e-01 -2.21637994e-01 1.50571391e-01 3.60889852e-01
-1.31852448e+00 7.04039693e-01 6.31272268e+00 2.61730731e-01
-1.29648602e+00 8.88639390e-02 4.47702557e-01 -1.58214882e-01
-5.63778818e-01 -1.36435390e-01 -2.47132927e-01 1.59122869e-01
6.21759832e-01 -1.32126594e-02 7.31113076e-01 3.97675693e-01
-1.55405357e-01 6.53440058e-02 -1.59832108e+00 1.06710577e+00
3.37949693e-01 -1.08072495e+00 2.26128593e-01 -7.55680958e-03
7.82193899e-01 -4.95473146e-02 5.38045526e-01 2.44890124e-01
-5.06306551e-02 -1.27502215e+00 6.34341300e-01 3.19691777e-01
8.09891820e-01 -5.60908020e-01 6.10120952e-01 2.59718925e-01
-7.02843010e-01 -4.27444428e-02 2.32945699e-02 4.73286994e-02
1.05715029e-01 1.47733495e-01 -5.93203604e-01 4.25468683e-01
4.17972684e-01 5.64238012e-01 -7.17596531e-01 9.10529733e-01
-1.12801135e-01 2.58828223e-01 -4.96279836e-01 1.81446299e-01
2.92203397e-01 2.19184831e-01 4.09951806e-01 9.27436292e-01
2.59374887e-01 -2.54243106e-01 1.14196703e-01 5.41347682e-01
-6.12083256e-01 3.17938775e-02 -7.08360970e-01 -5.76353073e-01
4.67163146e-01 1.21410501e+00 -2.82200605e-01 -7.21969903e-02
-8.93652558e-01 7.82435834e-01 1.53520644e-01 5.82897604e-01
-6.13370180e-01 -9.62421224e-02 6.72879159e-01 -9.86645967e-02
1.46410456e-02 -3.44011813e-01 -3.45546156e-01 -1.25027442e+00
1.18009783e-01 -1.10954535e+00 4.11054075e-01 -1.00510800e+00
-1.39535189e+00 5.07852376e-01 -8.03200975e-02 -1.18561316e+00
-3.92062992e-01 -5.92424691e-01 -1.15087636e-01 8.50043476e-01
-1.29182291e+00 -1.66297317e+00 -3.16489965e-01 7.67359793e-01
2.09222600e-01 -3.48293334e-01 9.92973804e-01 3.84383023e-01
-3.00652474e-01 5.83888769e-01 -1.51287382e-02 2.27427214e-01
9.62176204e-01 -1.00944865e+00 -8.25299993e-02 7.20801413e-01
4.03971702e-01 4.46344823e-01 6.55615330e-01 -2.43180424e-01
-1.85770023e+00 -8.55365813e-01 8.17510009e-01 -4.43691790e-01
7.90634215e-01 -3.34285319e-01 -9.75895107e-01 8.78513038e-01
4.33394313e-01 4.92916955e-03 8.82223308e-01 3.69097501e-01
-8.10062826e-01 -1.71283320e-01 -9.27528322e-01 7.24557579e-01
6.42738402e-01 -9.54738021e-01 -3.16389501e-01 3.35412234e-01
4.43781942e-01 -5.65844417e-01 -8.35001707e-01 5.13654947e-01
6.57716751e-01 -5.04256606e-01 8.52848351e-01 -1.08889210e+00
6.28794193e-01 -2.91221648e-01 -5.23755014e-01 -1.24853277e+00
7.21791238e-02 -4.14746046e-01 -1.32131457e-01 1.37355280e+00
7.18106747e-01 -4.50115055e-01 6.56042397e-01 1.02316868e+00
-5.62051050e-02 -4.26382720e-01 -9.43040371e-01 -4.40548778e-01
2.92323157e-02 -3.59346837e-01 7.12989151e-01 1.18875790e+00
2.42317215e-01 5.25569737e-01 -6.67874694e-01 2.31797501e-01
2.73370832e-01 1.81561485e-02 9.41693902e-01 -9.91452277e-01
-1.36130199e-01 -4.96038228e-01 -4.33131307e-01 -9.32259202e-01
2.91862905e-01 -8.19500327e-01 -6.21449389e-02 -1.60861552e+00
4.15866524e-01 -3.17580223e-01 -3.79015714e-01 1.06074309e+00
-5.22398986e-02 5.71718574e-01 4.20855999e-01 4.81053777e-02
-5.75257301e-01 6.29100859e-01 1.21882164e+00 -4.30246979e-01
6.66149147e-03 -1.69018656e-01 -9.11569834e-01 5.53299844e-01
5.83159924e-01 -7.35281780e-02 -4.53073144e-01 -7.42741764e-01
5.80044985e-01 8.25660676e-02 5.32876432e-01 -6.90275609e-01
2.43732780e-01 -1.99787125e-01 4.12510604e-01 -3.24842811e-01
7.36949980e-01 -1.04337037e+00 8.13556239e-02 -1.05434321e-01
-6.87545598e-01 6.47573993e-02 3.85809183e-01 3.98050100e-01
-4.64002967e-01 -1.02300927e-01 5.11926353e-01 2.11484164e-01
-6.47163570e-01 2.07084537e-01 -3.63453776e-01 -8.85886848e-02
7.20733941e-01 1.86205000e-01 -5.53078234e-01 -7.18056083e-01
-7.77659655e-01 3.29609960e-01 6.53400540e-01 7.27625608e-01
3.92683506e-01 -1.52238917e+00 -6.19042635e-01 8.87564272e-02
4.32894439e-01 -9.58249792e-02 2.67905086e-01 8.27407062e-01
-1.39522254e-02 3.19372356e-01 -3.20028901e-01 -6.66366398e-01
-1.37856030e+00 4.25133795e-01 2.47654229e-01 8.92124921e-02
-2.14002520e-01 4.31020111e-01 2.23233879e-01 -5.74009299e-01
3.89460266e-01 9.45929885e-02 -5.13758697e-02 1.48997828e-01
3.76646280e-01 -1.26122534e-01 4.89614010e-02 -8.81294191e-01
-2.50833571e-01 2.94304252e-01 -8.16066563e-02 -4.90585178e-01
1.21788180e+00 -2.31451079e-01 -1.74724311e-01 5.38702965e-01
1.09430313e+00 -2.61967391e-01 -1.25539970e+00 -5.55081308e-01
-4.29092348e-01 -2.61311650e-01 -6.08328208e-02 -9.73427713e-01
-1.01708400e+00 1.15836549e+00 3.55828434e-01 3.45682085e-01
1.19108689e+00 8.37246701e-02 7.57342219e-01 6.49615884e-01
-6.35351911e-02 -8.87820303e-01 2.94334114e-01 4.93699133e-01
8.05965066e-01 -1.53997195e+00 -1.91607893e-01 -1.50523484e-01
-9.01242435e-01 1.09333372e+00 5.24901867e-01 6.11541569e-01
3.62923324e-01 9.57773402e-02 8.87199491e-02 1.03431627e-01
-8.15736711e-01 -3.61034602e-01 5.09477317e-01 5.61814308e-01
5.35125256e-01 -7.90401474e-02 3.28434408e-01 3.53796750e-01
7.39812106e-02 -1.40935719e-01 4.39221740e-01 9.42262530e-01
-1.23552784e-01 -1.18504143e+00 -3.91261935e-01 3.66374493e-01
-3.16386074e-01 7.97957275e-03 -6.43914938e-01 8.37667406e-01
2.40958706e-02 9.92649376e-01 5.82176559e-02 -3.89380991e-01
1.82542101e-01 3.53198200e-01 6.70904160e-01 -5.03037870e-01
-2.20330551e-01 2.33638454e-02 2.01622605e-01 -7.73173571e-02
-6.30036056e-01 -9.13744032e-01 -1.09740376e+00 -3.34800690e-01
1.37238204e-01 -2.37700582e-01 9.29602683e-01 1.35188186e+00
3.69478315e-01 3.87359768e-01 4.00778651e-01 -8.42710912e-01
-1.83556080e-01 -8.65081072e-01 -1.52802974e-01 7.67097592e-01
5.81470311e-01 -5.52675724e-01 -3.45434695e-02 4.31267589e-01]
|
[10.821537971496582, 1.5795531272888184]
|
48dadc59-dc1c-4244-bd6f-aa8600a89c77
|
instant-multi-view-head-capture-through-1
|
2306.07437
| null |
https://arxiv.org/abs/2306.07437v1
|
https://arxiv.org/pdf/2306.07437v1.pdf
|
Instant Multi-View Head Capture through Learnable Registration
|
Existing methods for capturing datasets of 3D heads in dense semantic correspondence are slow, and commonly address the problem in two separate steps; multi-view stereo (MVS) reconstruction followed by non-rigid registration. To simplify this process, we introduce TEMPEH (Towards Estimation of 3D Meshes from Performances of Expressive Heads) to directly infer 3D heads in dense correspondence from calibrated multi-view images. Registering datasets of 3D scans typically requires manual parameter tuning to find the right balance between accurately fitting the scans surfaces and being robust to scanning noise and outliers. Instead, we propose to jointly register a 3D head dataset while training TEMPEH. Specifically, during training we minimize a geometric loss commonly used for surface registration, effectively leveraging TEMPEH as a regularizer. Our multi-view head inference builds on a volumetric feature representation that samples and fuses features from each view using camera calibration information. To account for partial occlusions and a large capture volume that enables head movements, we use view- and surface-aware feature fusion, and a spatial transformer-based head localization module, respectively. We use raw MVS scans as supervision during training, but, once trained, TEMPEH directly predicts 3D heads in dense correspondence without requiring scans. Predicting one head takes about 0.3 seconds with a median reconstruction error of 0.26 mm, 64% lower than the current state-of-the-art. This enables the efficient capture of large datasets containing multiple people and diverse facial motions. Code, model, and data are publicly available at https://tempeh.is.tue.mpg.de.
|
['Michael J. Black', 'Tianye Li', 'Timo Bolkart']
|
2023-06-12
|
instant-multi-view-head-capture-through
|
http://openaccess.thecvf.com//content/CVPR2023/html/Bolkart_Instant_Multi-View_Head_Capture_Through_Learnable_Registration_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Bolkart_Instant_Multi-View_Head_Capture_Through_Learnable_Registration_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['camera-calibration', '3d-face-reconstruction', 'semantic-correspondence', 'multi-view-3d-shape-retrieval']
|
['computer-vision', 'computer-vision', 'computer-vision', 'computer-vision']
|
[-6.17207550e-02 2.38632441e-01 1.21513925e-01 -6.80025399e-01
-1.11626935e+00 -4.57505703e-01 4.79187518e-01 -9.60130915e-02
-2.16829434e-01 3.51392806e-01 3.62571031e-01 3.65483582e-01
3.23109031e-01 -7.16795683e-01 -1.03013337e+00 -5.08255184e-01
3.43614548e-01 1.09458768e+00 2.01501429e-01 1.51222438e-01
-3.73530686e-02 8.15758288e-01 -1.76362240e+00 -1.68496743e-01
5.78239322e-01 8.85703564e-01 1.42812639e-01 5.58131456e-01
1.03715405e-01 3.48248370e-02 -2.60031279e-02 -6.63081646e-01
2.79129684e-01 2.54474184e-03 -6.37127221e-01 4.52206969e-01
1.00920796e+00 -6.07654393e-01 -5.54646291e-02 5.79608500e-01
7.57826447e-01 -4.30464335e-02 6.59714222e-01 -9.45456326e-01
-4.03374769e-02 -1.15894689e-03 -8.76670659e-01 -1.93671569e-01
8.01250577e-01 4.55363132e-02 6.02183759e-01 -1.15143955e+00
7.09882081e-01 1.25411212e+00 9.69603956e-01 7.65272439e-01
-1.39671731e+00 -7.78826952e-01 -5.36422106e-03 -3.32459122e-01
-1.53802788e+00 -1.10875976e+00 7.33378172e-01 -6.19948566e-01
6.41354740e-01 2.35649571e-01 1.06008112e+00 9.77051318e-01
3.83082256e-02 4.85801876e-01 9.79948938e-01 -8.79892483e-02
1.24124274e-01 -1.04921252e-01 -1.36443749e-01 8.14250290e-01
2.08400771e-01 -6.74340427e-02 -6.98089600e-01 -3.31701636e-01
1.04976022e+00 1.97061256e-01 -2.52625734e-01 -6.77018285e-01
-8.24267149e-01 6.55721664e-01 2.38490880e-01 -9.01600644e-02
-4.32440400e-01 7.88231120e-02 1.15646735e-01 -1.38322040e-01
7.18365133e-01 -2.26428404e-01 -3.88569862e-01 1.82412026e-04
-1.23653436e+00 4.38694924e-01 6.03432715e-01 1.06883872e+00
9.82702315e-01 -2.97558844e-01 3.10544342e-01 7.51327693e-01
5.51660299e-01 8.61304939e-01 1.80535972e-01 -1.13715112e+00
4.24608290e-01 5.49388707e-01 -2.40092695e-01 -5.98425984e-01
-5.83006620e-01 -1.31871589e-02 -6.20706797e-01 -5.78906648e-02
4.95288491e-01 1.27595499e-01 -9.89533842e-01 1.60349905e+00
9.75078583e-01 1.40321016e-01 -5.60284138e-01 9.69718575e-01
8.20845246e-01 -8.60023648e-02 -2.77143568e-01 -1.20026357e-01
1.44504440e+00 -4.53697205e-01 -3.35457951e-01 -4.08593059e-01
3.44613224e-01 -6.65276766e-01 9.19293463e-01 1.92166999e-01
-1.62427986e+00 5.80057851e-04 -6.15174472e-01 -4.70838130e-01
1.42392159e-01 -2.39932925e-01 3.97544444e-01 4.91136134e-01
-1.20859635e+00 5.40318727e-01 -1.20465922e+00 -4.65294302e-01
6.75373793e-01 8.62634897e-01 -8.02487373e-01 -2.21606180e-01
-4.47216779e-01 7.70272553e-01 -3.83505374e-01 -2.00395882e-02
-6.56037092e-01 -9.95476484e-01 -1.17306125e+00 -4.28632051e-01
1.17968328e-01 -1.15504742e+00 1.12331164e+00 -4.42528188e-01
-1.42498255e+00 1.56075692e+00 -6.33099556e-01 -7.55061060e-02
7.80145764e-01 -2.60386169e-01 2.87745725e-02 1.48972154e-01
2.52379358e-01 6.35704994e-01 9.29279625e-01 -1.20389915e+00
5.41651323e-02 -1.20909798e+00 -4.49214041e-01 2.28662118e-01
4.84617390e-02 -1.22245885e-02 -9.49401200e-01 -1.99349657e-01
5.93034029e-01 -1.02241623e+00 5.02958894e-02 4.14488345e-01
-3.53569597e-01 7.98620731e-02 5.19725502e-01 -7.52467930e-01
6.12863660e-01 -1.81486905e+00 2.85062641e-01 3.98507655e-01
4.55429822e-01 -3.12302858e-01 2.11412132e-01 -1.47260994e-01
8.38325918e-02 -2.79998362e-01 -3.31880838e-01 -1.24745500e+00
-1.45325258e-01 1.50759786e-01 1.67969525e-01 1.12470782e+00
-2.38749757e-01 8.47707212e-01 -5.24230361e-01 -7.18617201e-01
3.41371685e-01 9.74075854e-01 -8.89605999e-01 3.68707597e-01
9.77364182e-02 8.93057942e-01 -4.25000817e-01 8.98185015e-01
9.55011666e-01 -3.43730420e-01 1.49084941e-01 -4.05783683e-01
-2.80754976e-02 2.75612772e-01 -1.21746027e+00 2.13085651e+00
-5.61796904e-01 8.15916732e-02 6.30128324e-01 -5.87681115e-01
7.29633152e-01 2.22720429e-01 8.08997393e-01 -3.31120461e-01
3.59017432e-01 2.89408326e-01 -7.57075429e-01 -2.54998684e-01
1.13901101e-01 -3.29304993e-01 1.66080207e-01 3.57408702e-01
7.22739100e-02 -5.02718508e-01 -4.14240479e-01 -4.69188988e-02
8.06521654e-01 3.80636454e-01 7.44001195e-02 -7.68097937e-02
3.72263312e-01 -5.37585795e-01 4.96765882e-01 -1.04647391e-02
1.17181823e-01 1.10756195e+00 2.46965066e-02 -3.17344815e-01
-9.79142189e-01 -1.22317111e+00 -4.80278373e-01 7.88398206e-01
-4.36327532e-02 -3.94573718e-01 -1.01038885e+00 -3.46597403e-01
3.03367853e-01 2.16521472e-01 -6.56803489e-01 2.50815332e-01
-9.08260703e-01 -4.00290132e-01 2.24750608e-01 4.69105572e-01
1.13949209e-01 -4.69300807e-01 -5.79542577e-01 -1.84697695e-02
-9.00760517e-02 -1.17489100e+00 -8.25697482e-01 -1.06567755e-01
-1.05837679e+00 -1.13267028e+00 -8.77126753e-01 -2.77639806e-01
9.06387866e-01 8.67105350e-02 1.23919916e+00 1.88881457e-01
-2.86683798e-01 7.23067701e-01 2.34558135e-01 -7.06142858e-02
-3.02009005e-02 1.48519874e-01 3.54630738e-01 -7.17310309e-02
1.99839145e-01 -1.14601755e+00 -7.81126738e-01 3.43318731e-01
-4.10796106e-01 1.36578739e-01 1.87550604e-01 4.50739175e-01
1.00151050e+00 -8.99179220e-01 -1.92057446e-01 -7.47908831e-01
-2.07808003e-01 -5.09319127e-01 -7.08372891e-01 -5.84915429e-02
-3.79901648e-01 -1.44997602e-02 1.83156565e-01 -2.43229747e-01
-7.28505194e-01 4.94677395e-01 -3.33404064e-01 -9.10585523e-01
-3.46414834e-01 -2.20895439e-01 -4.24813151e-01 -1.24638908e-01
2.69482464e-01 2.43894950e-01 3.19349766e-01 -7.44480968e-01
2.95492142e-01 4.65721548e-01 5.91118157e-01 -6.16755724e-01
7.79309630e-01 8.15227985e-01 1.76934406e-01 -7.53748477e-01
-7.22928047e-01 -3.59211981e-01 -1.12151504e+00 -3.52943063e-01
9.55119610e-01 -1.11463737e+00 -8.94533694e-01 4.35951322e-01
-1.14690828e+00 -3.30944985e-01 -1.46094665e-01 4.54053283e-01
-7.29516268e-01 1.09349132e-01 -4.27863598e-01 -6.32716715e-01
-5.47699273e-01 -1.21772575e+00 1.89216173e+00 -7.35608786e-02
-3.75644892e-01 -9.35527146e-01 1.44245446e-01 8.04254115e-01
9.27935988e-02 4.91897941e-01 1.48168251e-01 -2.35049948e-01
-6.08201563e-01 -1.62970409e-01 1.32553399e-01 -1.86690465e-01
-9.41037759e-03 -2.06986502e-01 -1.37214184e+00 -4.77889746e-01
-3.78814265e-02 -3.40574414e-01 4.15762842e-01 5.86154521e-01
1.00928247e+00 -2.07695663e-01 -4.72630382e-01 1.21344781e+00
1.10597074e+00 -5.31841278e-01 3.71686667e-01 1.18803911e-01
1.06276131e+00 7.83002913e-01 2.28586778e-01 6.73382103e-01
9.76316690e-01 1.05803406e+00 4.36500072e-01 8.39698091e-02
-2.14497492e-01 -4.65012819e-01 2.81924099e-01 8.67325008e-01
-2.14792714e-01 4.05583084e-01 -1.00796533e+00 3.73064458e-01
-1.21467483e+00 -6.84202194e-01 -2.32690141e-01 2.50362301e+00
7.82408357e-01 -1.36305898e-01 4.12193447e-01 3.61451576e-03
5.17136335e-01 1.25731258e-02 -7.43555963e-01 8.18321034e-02
1.52767897e-01 3.24574262e-01 5.69098651e-01 7.11668551e-01
-8.03122938e-01 8.37398410e-01 5.31777954e+00 3.23657632e-01
-1.18068874e+00 3.48909050e-01 4.36220586e-01 -5.14708757e-01
-5.19750416e-01 -8.84326771e-02 -1.01207399e+00 4.14236546e-01
8.07952583e-01 2.71761537e-01 5.01324654e-01 7.07915425e-01
1.68438822e-01 -5.40379770e-02 -1.19030774e+00 1.31119633e+00
3.10730249e-01 -1.07164443e+00 -2.01433972e-01 5.61365664e-01
3.74496460e-01 3.44659060e-01 -9.94962975e-02 -1.84113309e-01
7.70884454e-02 -1.03266251e+00 1.00390160e+00 6.52425885e-01
1.18986285e+00 -5.03027380e-01 1.80079326e-01 3.40375572e-01
-1.48984241e+00 4.26675051e-01 -1.00870088e-01 2.79746532e-01
4.35656697e-01 7.65113652e-01 -5.29818296e-01 3.73643190e-01
8.04581046e-01 7.65736282e-01 -4.71126914e-01 5.31433284e-01
1.33696839e-01 1.63035214e-01 -8.14198494e-01 6.96324587e-01
-4.96091098e-01 -3.38891834e-01 4.90875512e-01 7.95867205e-01
3.89180839e-01 2.14566275e-01 -9.15334839e-03 8.48834455e-01
-1.76056877e-01 5.72323054e-03 -5.82166374e-01 7.00864255e-01
6.73207998e-01 1.23590147e+00 -4.61500496e-01 -9.83861014e-02
-4.17830586e-01 1.10104477e+00 5.38110495e-01 -1.74131662e-01
-7.63370335e-01 4.25976157e-01 7.91467965e-01 8.66167188e-01
2.14590475e-01 -1.99901909e-01 -4.28175569e-01 -1.36565125e+00
3.31215292e-01 -4.44813937e-01 1.19280726e-01 -6.41156197e-01
-1.01710284e+00 4.12323415e-01 8.36965442e-02 -1.05366611e+00
-3.32268834e-01 -1.36721268e-01 -4.61498350e-01 7.60733187e-01
-1.34175849e+00 -1.57483637e+00 -5.56672752e-01 9.03587162e-01
3.65252018e-01 4.01304990e-01 7.32681036e-01 3.45422804e-01
-6.18286312e-01 8.37810814e-01 -3.92764956e-01 -1.32455647e-01
7.72925436e-01 -1.06899500e+00 4.87299621e-01 3.15655768e-01
-1.50771111e-01 6.91180885e-01 5.39878547e-01 -6.54280305e-01
-1.89348364e+00 -9.95945513e-01 7.67664790e-01 -8.59097719e-01
1.66818872e-01 -6.45851374e-01 -9.74182129e-01 9.56974506e-01
-4.61271971e-01 4.78667110e-01 6.03670001e-01 -2.89595183e-02
-3.92036021e-01 -1.25392959e-01 -1.45588493e+00 1.59112558e-01
1.45920956e+00 -5.67665756e-01 -3.68966490e-01 1.43889129e-01
2.14520544e-01 -9.53565836e-01 -1.35330498e+00 2.19715402e-01
8.94090950e-01 -1.16736877e+00 1.18847632e+00 1.02726497e-01
1.30608317e-03 -1.00576945e-01 -3.02144408e-01 -9.47731674e-01
1.59892708e-01 -5.88167310e-01 -2.90557653e-01 1.31300497e+00
-2.52247620e-02 -5.77380002e-01 1.04804659e+00 1.06690192e+00
-1.13953404e-01 -7.77035058e-01 -1.17982268e+00 -3.59482169e-01
8.23983923e-02 -4.62653011e-01 7.75108933e-01 7.06993580e-01
-3.75674516e-01 1.50364935e-01 -1.20600596e-01 2.45571226e-01
1.06910384e+00 5.71946502e-02 1.07084846e+00 -1.55644822e+00
-1.22372061e-01 -9.53745544e-02 -3.81544262e-01 -1.09516907e+00
4.57752973e-01 -8.99322152e-01 -1.61406219e-01 -1.21838641e+00
2.90354729e-01 -4.76407111e-01 5.60037076e-01 4.28448290e-01
1.40107647e-01 4.61085260e-01 -3.14793438e-02 3.60854089e-01
-2.19699994e-01 6.18845880e-01 1.17556381e+00 3.77445728e-01
-1.37242690e-01 1.29473675e-02 -4.05611962e-01 1.08390832e+00
4.62663651e-01 -3.07619035e-01 -1.20193526e-01 -7.10656822e-01
-5.54163046e-02 3.00371349e-01 5.18459141e-01 -8.29393148e-01
2.33765632e-01 3.79139371e-02 7.07651675e-01 -6.80758119e-01
9.96818364e-01 -8.68469715e-01 5.54190934e-01 8.20168555e-02
1.55311123e-01 1.24888562e-01 9.57231410e-03 3.21309388e-01
2.78842151e-01 3.44988167e-01 1.04303002e+00 -1.90073922e-01
1.65484734e-02 8.22798371e-01 1.59260780e-01 1.45373523e-01
7.78759956e-01 -4.81977969e-01 2.69728482e-01 -3.22663754e-01
-8.82914126e-01 7.33443946e-02 1.26066887e+00 8.96534473e-02
7.04206526e-01 -1.28853095e+00 -5.67237377e-01 6.30064905e-01
-3.08546852e-02 6.43781960e-01 2.92489439e-01 1.18811035e+00
-3.14972728e-01 4.02696542e-02 1.50245413e-01 -9.42125261e-01
-1.49066186e+00 2.61490811e-02 5.51066875e-01 1.12659559e-01
-9.11268830e-01 7.70066440e-01 1.25939772e-01 -7.34462678e-01
9.08884332e-02 -2.00352967e-01 8.87935311e-02 2.78413650e-02
4.21718597e-01 4.43086296e-01 2.76411742e-01 -1.21807301e+00
-6.86466336e-01 1.40541017e+00 1.36130273e-01 -2.55534828e-01
1.57982659e+00 -4.31264222e-01 -9.97544974e-02 3.02968621e-01
1.51503122e+00 2.30689496e-01 -1.46912599e+00 -1.09367885e-01
-3.19823682e-01 -4.68603492e-01 2.06898510e-01 -1.46117494e-01
-1.39511406e+00 8.61462593e-01 6.13596380e-01 -5.31265080e-01
8.75646055e-01 4.79139268e-01 1.10641146e+00 -2.47536734e-01
6.58183753e-01 -6.47277176e-01 -3.76964629e-01 2.78223127e-01
8.78406405e-01 -1.14239979e+00 2.98835278e-01 -5.81821918e-01
-2.68270105e-01 9.00698602e-01 5.70697606e-01 -2.10947752e-01
7.70564556e-01 4.89793718e-01 -2.59960562e-01 -5.74142396e-01
-3.97918940e-01 7.17632920e-02 3.89999121e-01 5.02836764e-01
4.85143512e-01 2.02771947e-02 2.42176294e-01 3.16117078e-01
-6.66161716e-01 2.69838627e-02 6.80243298e-02 8.03962588e-01
-1.25881180e-01 -1.02322161e+00 -7.05158234e-01 4.95172232e-01
-4.04032588e-01 1.62418872e-01 -1.13234378e-01 5.39759636e-01
5.85458688e-02 5.36875844e-01 4.58661228e-01 -2.83333480e-01
5.74408114e-01 -2.02121194e-02 8.98694098e-01 -6.24812126e-01
-3.11978191e-01 2.45378539e-01 -1.72882438e-01 -1.14661467e+00
-5.38578391e-01 -1.09320891e+00 -1.19623840e+00 -6.60181344e-01
-2.68989861e-01 -2.85904825e-01 7.60859966e-01 9.07859802e-01
4.12871212e-01 -1.21453121e-01 6.09882832e-01 -1.59118629e+00
-2.65774876e-01 -6.44461334e-01 -5.56809127e-01 4.36015099e-01
5.16335249e-01 -9.67511594e-01 -4.70763505e-01 2.53577065e-02]
|
[13.302314758300781, 0.031331371515989304]
|
4720ab46-82db-4f67-9039-d0a8a95ef94c
|
accurate-gigapixel-crowd-counting-by
|
2305.09271
| null |
https://arxiv.org/abs/2305.09271v1
|
https://arxiv.org/pdf/2305.09271v1.pdf
|
Accurate Gigapixel Crowd Counting by Iterative Zooming and Refinement
|
The increasing prevalence of gigapixel resolutions has presented new challenges for crowd counting. Such resolutions are far beyond the memory and computation limits of current GPUs, and available deep neural network architectures and training procedures are not designed for such massive inputs. Although several methods have been proposed to address these challenges, they are either limited to downsampling the input image to a small size, or borrowing from other gigapixel tasks, which are not tailored for crowd counting. In this paper, we propose a novel method called GigaZoom, which iteratively zooms into the densest areas of the image and refines coarser density maps with finer details. Through experiments, we show that GigaZoom obtains the state-of-the-art for gigapixel crowd counting and improves the accuracy of the next best method by 42%.
|
['Alexandros Iosifidis', 'Qi Zhang', 'Arian Bakhtiarnia']
|
2023-05-16
| null | null | null | null |
['crowd-counting']
|
['computer-vision']
|
[-3.18366468e-01 -2.80682862e-01 2.65930057e-01 -3.81540880e-02
-2.18674779e-01 -1.03978351e-01 5.93994617e-01 1.76557243e-01
-9.46361065e-01 1.08761251e+00 2.43151635e-01 -3.74826714e-02
2.92363256e-01 -1.15639997e+00 -3.68354410e-01 -4.88151819e-01
2.66657081e-02 7.79555261e-01 6.19877875e-01 -7.98043087e-02
1.65744558e-01 4.12461370e-01 -1.76315558e+00 3.84522378e-02
9.35852945e-01 6.68790460e-01 2.15167254e-02 1.11121058e+00
-3.03194731e-01 7.83823013e-01 -1.07367313e+00 -5.35694540e-01
9.80005488e-02 -4.29710373e-02 -5.34125030e-01 -2.77286917e-01
9.70340669e-01 -7.75439024e-01 -6.10669911e-01 1.03297949e+00
6.04336381e-01 1.85057998e-01 4.77267623e-01 -7.89248645e-01
-5.25411308e-01 2.15692565e-01 -1.06416464e+00 1.04476261e+00
1.02156244e-01 1.33524731e-01 2.13823900e-01 -8.38594437e-01
2.81825036e-01 1.44080162e+00 9.37753320e-01 6.76028073e-01
-6.84155345e-01 -8.26720119e-01 -2.03974806e-02 -2.07781315e-01
-1.51331890e+00 -2.69351274e-01 3.74189094e-02 -5.36197424e-01
1.18446076e+00 1.49312299e-02 8.19040120e-01 8.37225854e-01
-5.62216602e-02 4.26744610e-01 1.18430567e+00 -3.39199960e-01
4.18330193e-01 -1.95954517e-01 3.49920660e-01 7.68429637e-01
1.01099908e+00 -1.32242918e-01 -7.37314820e-01 -3.83150488e-01
1.09619045e+00 1.27621964e-01 3.36173326e-02 3.26938897e-01
-1.00015163e+00 1.11187434e+00 5.04257977e-01 3.52513909e-01
-2.69488364e-01 2.63419777e-01 5.25906265e-01 -3.43904614e-01
8.09496880e-01 4.28978860e-01 7.56612420e-02 -4.17698950e-01
-1.29378867e+00 4.12836224e-01 8.00979793e-01 8.06181610e-01
7.93037295e-01 -2.70101428e-02 -2.69286066e-01 5.84420800e-01
-3.24269205e-01 6.42924428e-01 2.96078831e-01 -1.07133210e+00
6.85806513e-01 6.85971677e-01 1.70228556e-01 -1.06870639e+00
-5.75200796e-01 -1.13708407e-01 -1.21322370e+00 2.27120116e-01
7.59882152e-01 -3.59525383e-01 -7.00270772e-01 1.09847319e+00
5.32365084e-01 2.31253773e-01 -4.11010057e-01 8.93520415e-01
7.52192020e-01 6.32598996e-01 2.12781563e-01 2.52131343e-01
1.44594944e+00 -9.56442595e-01 -2.92488664e-01 -4.04517621e-01
4.47723716e-01 -2.23366007e-01 1.21155417e+00 3.14321101e-01
-1.05000830e+00 -5.24217486e-01 -9.37463105e-01 -3.12710345e-01
-4.58557665e-01 -1.08035773e-01 9.55458522e-01 8.90241265e-01
-1.18175614e+00 4.75980699e-01 -9.88877952e-01 -4.15767848e-01
9.29109812e-01 2.64212966e-01 -1.59329429e-01 -7.17221871e-02
-9.81974542e-01 7.73516119e-01 2.47909725e-01 -3.03381711e-01
-4.41398084e-01 -6.22157693e-01 -7.54102111e-01 1.84240505e-01
7.34078214e-02 -8.12618673e-01 1.04577982e+00 -3.05725157e-01
-1.29089415e+00 9.13764119e-01 -3.98770720e-01 -6.29517674e-01
7.49703884e-01 -1.62462682e-01 1.92724884e-01 1.67008147e-01
4.13042039e-01 8.65506232e-01 6.06529355e-01 -8.64098847e-01
-1.07259333e+00 -5.82704723e-01 8.73266682e-02 6.41226843e-02
-6.79096758e-01 1.16624095e-01 -4.12165642e-01 -3.37026447e-01
-5.75216770e-01 -4.97363716e-01 -4.05691713e-01 -8.78277421e-02
-6.19839616e-02 -2.28680953e-01 9.36394274e-01 -4.28601891e-01
1.17428732e+00 -1.75512803e+00 -3.03559661e-01 -1.57985300e-01
9.66288030e-01 6.22004271e-01 1.31958708e-01 -6.96975291e-02
7.79141963e-01 -7.99178379e-04 -1.59244329e-01 -8.53445590e-01
-1.67236298e-01 2.38520145e-01 -1.54820159e-01 4.62236196e-01
-1.93523541e-02 8.34521592e-01 -1.11725271e+00 -6.94219530e-01
2.77817190e-01 8.58419061e-01 -7.22253025e-01 2.51484122e-02
2.66035289e-01 6.60833895e-01 -2.30831772e-01 5.77008605e-01
1.08802903e+00 -5.50765872e-01 -1.38924122e-01 3.69533211e-01
-2.93042570e-01 5.09423986e-02 -7.65519023e-01 1.20274985e+00
-3.57898384e-01 7.41995037e-01 -5.01138307e-02 -2.83157825e-01
6.70324683e-01 -1.52599230e-01 9.70010757e-02 -6.85987711e-01
3.38982791e-01 2.66187787e-01 -3.17851961e-01 -1.38327897e-01
1.10095906e+00 -4.28285897e-02 -1.82883784e-01 3.59900028e-01
-2.28547044e-02 6.58777729e-02 5.79491675e-01 7.85985067e-02
1.14217722e+00 -4.84916061e-01 5.89264691e-01 -3.55446815e-01
2.56024867e-01 -9.26280580e-03 3.16091031e-01 1.15229487e+00
-4.38835949e-01 7.17284858e-01 5.17286301e-01 -9.69813108e-01
-1.33009303e+00 -8.04813445e-01 -1.54611364e-01 1.30626547e+00
1.83779653e-02 -2.24839419e-01 -1.36553347e+00 -4.98139679e-01
7.50674158e-02 2.47522835e-02 -9.53771591e-01 5.81851840e-01
-9.54756796e-01 -1.09771419e+00 1.06302166e+00 7.61587799e-01
1.04658854e+00 -8.62885714e-01 -1.22208023e+00 2.07228765e-01
-2.50853114e-02 -1.51647842e+00 -4.70771521e-01 -1.75282523e-01
-8.53630722e-01 -1.14160442e+00 -1.13383007e+00 -3.15042913e-01
6.84587657e-01 4.75155562e-01 1.44929338e+00 6.02628529e-01
-2.97663927e-01 -2.14579016e-01 -1.70414746e-01 -4.65368479e-01
2.35105958e-02 5.43461621e-01 1.91270486e-01 -3.88058841e-01
8.10638309e-01 -5.56151330e-01 -7.80869067e-01 -1.35510728e-01
-6.15775585e-01 -1.39256850e-01 2.21067265e-01 6.91285908e-01
3.94077897e-01 -1.05884306e-01 4.07423496e-01 -1.08968854e+00
7.87223995e-01 -3.98741513e-01 -8.35336685e-01 -2.84699649e-01
-1.81528524e-01 -1.71263069e-01 8.22259784e-01 -5.04727364e-01
-7.88289845e-01 -3.03635716e-01 -3.22248608e-01 -3.67605180e-01
-3.42212677e-01 -1.39291838e-01 2.30989531e-01 -3.25166196e-01
8.31436634e-01 2.10675858e-02 -3.11685503e-01 -3.05362135e-01
-6.09379485e-02 6.08032167e-01 5.99330246e-01 -5.95041275e-01
3.85298967e-01 9.65008378e-01 -1.18228376e-01 -1.00561142e+00
-1.05460656e+00 -4.47065294e-01 -7.65093803e-01 -1.34174779e-01
1.09109366e+00 -8.93031955e-01 -9.70891416e-01 7.09156275e-01
-1.29538572e+00 -5.14318466e-01 -3.95412356e-01 9.48964283e-02
-9.96520296e-02 3.88690621e-01 -9.12365198e-01 -8.92133653e-01
-6.21736944e-01 -8.63885343e-01 1.40819848e+00 7.11972535e-01
-2.42139325e-01 -8.31712961e-01 4.20571625e-01 2.06108987e-01
7.26827979e-01 2.22684681e-01 3.44408453e-01 -1.92541152e-01
-4.01758164e-01 -2.09579423e-01 -6.60606742e-01 -1.37475535e-01
-2.97696829e-01 -2.78780162e-01 -1.10698891e+00 -2.66197950e-01
-1.52563974e-01 -3.65158439e-01 1.05341649e+00 4.49006200e-01
1.34338987e+00 -3.56513560e-01 -4.51851130e-01 8.65674913e-01
1.20126772e+00 -3.23833555e-01 7.84557581e-01 4.69083101e-01
7.65409589e-01 2.98923731e-01 8.40211660e-02 7.37990081e-01
6.54427469e-01 5.01842022e-01 2.10691795e-01 -1.28548563e-01
-2.24752218e-01 -3.68120342e-01 -2.97743529e-01 7.57228553e-01
-6.10636532e-01 -3.29853296e-01 -1.00474119e+00 5.23995519e-01
-1.71312189e+00 -9.90047753e-01 -9.86112952e-02 1.95653510e+00
3.94309580e-01 -4.74809743e-02 4.99351650e-01 7.76589885e-02
7.50044703e-01 3.72972518e-01 -3.84452641e-01 -2.09264055e-01
-1.62550688e-01 4.03260410e-01 7.13398337e-01 5.93150258e-01
-1.12675703e+00 1.11433995e+00 7.26756239e+00 9.76433277e-01
-8.52593362e-01 5.71286857e-01 8.56677234e-01 -2.62325615e-01
3.56087506e-01 -5.22984564e-01 -1.38814735e+00 7.47775435e-01
7.29227543e-01 2.32651010e-01 3.79492015e-01 8.58886898e-01
-1.44718960e-01 -5.95736504e-01 -6.24155045e-01 1.23999071e+00
1.00727797e-01 -1.49098635e+00 1.13064259e-01 3.64693195e-01
9.62112248e-01 1.88831553e-01 -5.34821525e-02 4.88819629e-01
4.35442299e-01 -1.28478074e+00 4.27512318e-01 3.54425728e-01
8.74717593e-01 -9.48598325e-01 8.91809344e-01 5.46477735e-01
-1.19051266e+00 1.64862722e-01 -8.68961096e-01 -7.33192146e-01
1.86891362e-01 7.22475231e-01 -4.91926074e-01 -2.94331163e-01
1.08654809e+00 1.05660990e-01 -7.76837766e-01 9.93795395e-01
1.49576023e-01 3.61590445e-01 -5.13631165e-01 -4.19974238e-01
3.18845600e-01 5.46969138e-02 2.45168597e-01 1.68441212e+00
5.63310683e-01 2.35569522e-01 9.75640025e-03 6.40730023e-01
-3.36257488e-01 -2.66650915e-01 -6.93973184e-01 4.58011031e-01
7.36007154e-01 1.16980290e+00 -9.72308338e-01 -8.34647298e-01
-3.35437566e-01 9.46391284e-01 9.65586126e-01 1.37346014e-01
-8.12992394e-01 -3.75254869e-01 8.22437346e-01 5.17509580e-01
2.53875494e-01 -3.68853450e-01 -5.40824592e-01 -1.27674782e+00
-1.62184834e-01 -7.13635325e-01 3.64594221e-01 -3.56110543e-01
-1.33119333e+00 7.36168087e-01 5.37352115e-02 -6.24126792e-01
1.27403310e-03 -6.19645715e-01 -7.15225101e-01 6.73863113e-01
-1.41261148e+00 -8.65575969e-01 -9.45041776e-01 4.74168122e-01
3.70762974e-01 -1.11539148e-01 6.89522803e-01 4.99694645e-01
-2.40159854e-01 5.52669585e-01 -2.58494429e-02 3.97386700e-01
1.42019764e-01 -1.09059536e+00 8.67110372e-01 9.41670477e-01
-2.05176458e-01 3.57647121e-01 4.08183545e-01 -5.84366024e-01
-7.36289263e-01 -1.16450047e+00 9.24298465e-01 -6.55813217e-01
4.58370924e-01 -5.88591814e-01 -8.63543332e-01 3.64260584e-01
3.81928682e-02 4.70833182e-01 3.55480134e-01 1.62225068e-01
-2.97628164e-01 2.74212301e-01 -1.33735096e+00 4.46157306e-01
1.19043159e+00 -4.03206766e-01 -2.41151333e-01 1.70128301e-01
4.18576151e-01 -7.12230265e-01 -5.52368581e-01 5.51873818e-02
2.83055663e-01 -1.37885022e+00 1.05235207e+00 -2.29980960e-01
3.32408041e-01 -2.38182783e-01 2.03186981e-02 -1.10573590e+00
-3.93887609e-01 -1.91858158e-01 -4.56750095e-01 7.86234319e-01
-1.38066024e-01 -8.64016354e-01 1.22661400e+00 3.44174415e-01
1.98448747e-01 -5.24330735e-01 -1.28356445e+00 -5.72568893e-01
3.39327008e-01 -3.58043723e-02 1.02703881e+00 7.05283523e-01
-1.08498722e-01 2.17067018e-01 -4.53234673e-01 3.41431089e-02
9.82885003e-01 -2.59096563e-01 1.08953667e+00 -1.22055614e+00
-4.84982096e-02 -6.32831097e-01 -5.73598325e-01 -1.32320070e+00
-7.27922842e-03 -2.83308655e-01 -7.19306916e-02 -1.49281418e+00
6.50885165e-01 -4.41299707e-01 4.10961747e-01 1.07315108e-01
-5.59773564e-01 8.90212655e-01 3.10805500e-01 2.26573735e-01
-9.21779335e-01 3.76473427e-01 1.24575126e+00 -7.71721900e-02
6.42545074e-02 -3.11261088e-01 -6.14245892e-01 1.00426471e+00
7.87046731e-01 -4.74073350e-01 1.31448224e-01 -9.36753571e-01
2.05410853e-01 -2.30199888e-01 4.20480877e-01 -1.64410162e+00
4.61932272e-01 6.99772835e-02 6.99163735e-01 -5.72041571e-01
4.87042546e-01 -1.56692669e-01 -3.10333490e-01 4.66105908e-01
2.37894893e-01 2.79364377e-01 1.63806304e-01 4.34798926e-01
-1.46105900e-01 6.17141323e-03 1.03685975e+00 -3.13786089e-01
-4.71500635e-01 5.01564741e-01 -3.29019040e-01 5.02279460e-01
7.80144215e-01 -4.11057949e-01 -8.02397251e-01 -9.95651037e-02
-2.57618666e-01 4.46016863e-02 6.80588543e-01 -9.22637731e-02
3.89538080e-01 -1.09577811e+00 -7.93006063e-01 9.04872119e-02
-2.75636315e-01 4.29938465e-01 2.63552636e-01 5.01268446e-01
-1.07858038e+00 5.59077084e-01 -3.17331791e-01 -6.05237067e-01
-1.08158600e+00 4.06079769e-01 4.50020432e-01 -7.88230062e-01
-9.64374125e-01 9.80799258e-01 3.78305107e-01 -4.70281452e-01
9.31221768e-02 -3.18906993e-01 -3.22903812e-01 -1.53411865e-01
1.22771752e+00 7.93585837e-01 -1.53068965e-02 -7.72814929e-01
-3.63009006e-01 5.74482977e-01 8.53016078e-02 1.02976926e-01
1.23369360e+00 8.06434732e-03 -2.12795585e-02 9.74356458e-02
8.07371259e-01 7.48394281e-02 -1.58983445e+00 -9.72640701e-03
-3.52412760e-01 -9.49426293e-01 -2.05893174e-01 -6.63530007e-02
-1.12142634e+00 1.23013747e+00 4.49554503e-01 2.48196185e-01
7.52773702e-01 3.45504805e-02 1.13192534e+00 3.19531888e-01
9.01588321e-01 -9.82302487e-01 1.30101234e-01 1.16509628e+00
3.05265576e-01 -1.33102202e+00 6.03209995e-02 -2.34176546e-01
-2.20614776e-01 8.92565548e-01 1.04745173e+00 -5.12952626e-01
3.81215543e-01 7.90687263e-01 -1.78887799e-01 -4.68410045e-01
-3.72246683e-01 -2.13510960e-01 -2.52385944e-01 8.94727290e-01
3.65492046e-01 2.69741505e-01 -1.80098861e-02 3.29953104e-01
-5.20958304e-01 2.10442841e-01 4.47306544e-01 6.25863492e-01
-8.25792134e-01 -3.97259265e-01 -8.32248926e-01 7.72087574e-01
-4.84896928e-01 -1.84322417e-01 -5.67017719e-02 8.21948230e-01
3.85219336e-01 7.85279572e-01 5.75646877e-01 -2.02071264e-01
2.50056535e-01 -5.22034347e-01 6.30356669e-01 -4.91185188e-01
-6.76158905e-01 -5.88668704e-01 -7.10950941e-02 -3.70218664e-01
-2.32080027e-01 -3.80110115e-01 -1.01543295e+00 -1.38833094e+00
-7.60930553e-02 1.51222842e-02 2.07994998e-01 9.08805192e-01
2.11000368e-01 4.36039656e-01 -1.72713444e-01 -1.63770747e+00
-5.32484390e-02 -1.16340661e+00 -4.32796985e-01 1.70292079e-01
2.80545652e-01 -6.95506215e-01 -2.70185471e-01 -4.70648497e-01]
|
[8.40416431427002, -0.33159422874450684]
|
97e533da-766c-4151-bd63-dab4d1ebc473
|
backdoor-attack-is-a-devil-in-federated-gan
|
2207.00762
| null |
https://arxiv.org/abs/2207.00762v2
|
https://arxiv.org/pdf/2207.00762v2.pdf
|
Backdoor Attack is a Devil in Federated GAN-based Medical Image Synthesis
|
Deep Learning-based image synthesis techniques have been applied in healthcare research for generating medical images to support open research. Training generative adversarial neural networks (GAN) usually requires large amounts of training data. Federated learning (FL) provides a way of training a central model using distributed data from different medical institutions while keeping raw data locally. However, FL is vulnerable to backdoor attack, an adversarial by poisoning training data, given the central server cannot access the original data directly. Most backdoor attack strategies focus on classification models and centralized domains. In this study, we propose a way of attacking federated GAN (FedGAN) by treating the discriminator with a commonly used data poisoning strategy in backdoor attack classification models. We demonstrate that adding a small trigger with size less than 0.5 percent of the original image size can corrupt the FL-GAN model. Based on the proposed attack, we provide two effective defense strategies: global malicious detection and local training regularization. We show that combining the two defense strategies yields a robust medical image generation.
|
['Xiaoxiao Li', 'Ruinan Jin']
|
2022-07-02
| null | null | null | null |
['data-poisoning', 'medical-image-generation']
|
['adversarial', 'medical']
|
[ 3.01577508e-01 4.80870157e-01 1.04781927e-03 7.06071854e-02
-9.77501214e-01 -1.03843272e+00 5.54418862e-01 -2.20804617e-01
-2.50204712e-01 7.99491584e-01 1.40747847e-02 -6.96590483e-01
5.51601529e-01 -1.29376888e+00 -9.40839648e-01 -1.08586657e+00
4.65012826e-02 3.64333272e-01 -2.81658798e-01 -9.88027379e-02
-3.10010880e-01 6.46655679e-01 -7.31270969e-01 5.55700898e-01
6.82944953e-01 6.74328625e-01 -3.31551671e-01 9.90083575e-01
1.11382408e-03 1.06278872e+00 -1.39933264e+00 -7.51690149e-01
8.38459074e-01 -9.15000498e-01 -8.56162429e-01 -1.47718057e-01
2.88016856e-01 -6.73594296e-01 -4.45971787e-01 1.18142676e+00
7.88646936e-01 -4.77332503e-01 3.29632312e-01 -1.58722341e+00
-6.94682300e-01 6.63737953e-01 -9.73210931e-02 -7.07239136e-02
-1.36448711e-01 6.22300684e-01 1.87988639e-01 -1.29289433e-01
6.93167627e-01 8.98575127e-01 5.18399775e-01 1.34869123e+00
-1.03691697e+00 -9.76932883e-01 -4.63481545e-01 -5.39087534e-01
-1.07184660e+00 -2.50488997e-01 7.80880988e-01 -2.94646561e-01
4.55176026e-01 6.74083412e-01 2.57073492e-01 1.64018595e+00
6.24844551e-01 3.03363442e-01 1.32320285e+00 -4.06494945e-01
4.78664339e-01 3.39567214e-01 -3.37429553e-01 8.69945824e-01
4.94369626e-01 3.47606152e-01 -1.49512693e-01 -9.38456237e-01
8.23999345e-01 2.83214211e-01 -1.97307169e-01 1.69524506e-01
-8.79840791e-01 1.27705753e+00 6.13861501e-01 2.24614769e-01
-3.04955930e-01 4.40612435e-01 5.03273487e-01 3.95325363e-01
4.03156787e-01 4.75456744e-01 -2.97573417e-01 5.17711103e-01
-6.29852533e-01 3.65707353e-02 7.21227109e-01 7.11075842e-01
3.58670682e-01 4.15228635e-01 -3.25135440e-01 1.30229443e-01
-3.77520435e-02 6.46869600e-01 8.27884376e-01 -8.50778282e-01
3.90577167e-01 2.96442002e-01 -2.73131788e-01 -8.25788021e-01
1.54020339e-01 -3.46832424e-01 -1.18839979e+00 3.49995077e-01
2.23040998e-01 -6.79122210e-01 -9.38352466e-01 1.88363361e+00
5.54925740e-01 2.14618325e-01 5.65876544e-01 3.71070713e-01
4.98834282e-01 4.54171747e-01 -5.65460846e-02 -2.82365642e-03
1.46916866e+00 -9.33982909e-01 -6.81189358e-01 2.68859386e-01
6.54773235e-01 -5.12141347e-01 6.67078257e-01 4.17878777e-01
-9.38762069e-01 -7.17954189e-02 -1.09974992e+00 2.95142502e-01
-6.01178646e-01 -3.51541758e-01 5.40268004e-01 1.36847961e+00
-1.10616589e+00 2.61257052e-01 -9.27920938e-01 8.06161854e-03
1.03997087e+00 4.97228861e-01 -4.50410157e-01 -9.04989317e-02
-1.22122896e+00 3.91212761e-01 1.12292180e-02 -4.53801811e-01
-1.53552318e+00 -6.77190542e-01 -6.13184392e-01 -2.50719637e-01
-6.52040318e-02 -1.23816037e+00 9.94246066e-01 -1.16049194e+00
-1.36946404e+00 8.95280659e-01 5.81641912e-01 -1.05250359e+00
9.81965721e-01 1.23640798e-01 -4.35953617e-01 3.88274670e-01
2.96117160e-02 3.68570805e-01 1.29707360e+00 -1.33819640e+00
-1.40324324e-01 -5.48707664e-01 -1.46592557e-01 -3.61375004e-01
-5.83037913e-01 9.99814495e-02 4.35884297e-01 -9.77730215e-01
-4.82528716e-01 -9.01810527e-01 -5.48070192e-01 -2.81591356e-01
-8.36985409e-01 3.92070055e-01 1.21927857e+00 -5.17017126e-01
6.74703896e-01 -1.95610714e+00 -3.55855703e-01 3.02269846e-01
6.22824013e-01 4.90921855e-01 -7.56093487e-02 3.75918716e-01
-6.36367947e-02 5.39545834e-01 -2.64001578e-01 -2.50297934e-01
-3.67149293e-01 3.17763299e-01 -7.35916018e-01 6.17708147e-01
-1.54844746e-01 1.01607728e+00 -6.38185441e-01 -3.56563002e-01
-1.07179537e-01 6.61080122e-01 -6.88310146e-01 6.29344940e-01
-2.46925905e-01 6.02547824e-01 -4.77319717e-01 7.60010540e-01
7.38473296e-01 -2.05905914e-01 3.15145969e-01 3.55410539e-02
5.37511230e-01 -2.67890543e-01 -5.14121234e-01 1.30929065e+00
-2.34136611e-01 1.05820134e-01 1.66332528e-01 -5.94291210e-01
7.59035587e-01 7.30254054e-01 3.91208857e-01 -1.97079673e-01
4.21941429e-01 9.79706049e-02 -2.43013054e-01 -2.93899566e-01
-4.49806936e-02 -2.85579890e-01 -2.25491151e-01 7.48416662e-01
1.31096140e-01 2.80445695e-01 -6.76066399e-01 4.37883765e-01
1.65579975e+00 -5.32823324e-01 1.18821613e-01 -6.05662540e-02
4.43161070e-01 7.35493824e-02 3.84731144e-01 9.95750427e-01
-3.19660246e-01 5.21339178e-01 4.23602134e-01 -6.60669029e-01
-1.09422815e+00 -9.69237864e-01 3.09913337e-01 6.65483415e-01
-3.40153039e-01 -3.57003540e-01 -1.32005727e+00 -1.38295734e+00
-1.37445658e-01 5.89870691e-01 -8.09246004e-01 -5.44215083e-01
-6.36821747e-01 -7.11256504e-01 1.51913786e+00 3.58212590e-01
6.63057923e-01 -1.03112006e+00 -6.81407690e-01 -3.53381820e-02
-6.37105256e-02 -7.77339339e-01 -5.46734869e-01 3.69977318e-02
-8.33612144e-01 -1.18789208e+00 -8.13351989e-01 -5.75889528e-01
1.07874107e+00 3.67192291e-02 9.81054723e-01 3.06718439e-01
-6.98214829e-01 3.65621537e-01 -1.55451670e-01 -6.50586426e-01
-1.37646687e+00 -1.39884770e-01 8.65124390e-02 1.73199803e-01
-8.84404108e-02 -3.72894138e-01 -6.90636098e-01 1.72393285e-02
-1.29408360e+00 -1.33146539e-01 3.23423058e-01 1.09545600e+00
4.13569480e-01 1.51311066e-02 5.98997891e-01 -1.59054542e+00
6.82522058e-01 -6.94584310e-01 -5.73136568e-01 2.70818353e-01
-6.10865593e-01 5.58563620e-02 1.08774555e+00 -6.00933790e-01
-7.34978855e-01 1.25491908e-02 -2.11807281e-01 -1.10241175e+00
-3.11379373e-01 -1.61358818e-01 -3.11283529e-01 -3.75685424e-01
9.35259998e-01 3.05328667e-01 4.09908652e-01 -3.17536473e-01
5.03408551e-01 6.51265681e-01 5.23470223e-01 -4.80017632e-01
8.80332708e-01 7.89081156e-01 1.90181527e-02 -2.13731825e-01
-3.32426310e-01 2.80665278e-01 1.61301985e-01 3.68811190e-02
1.04814148e+00 -8.25907707e-01 -5.83104432e-01 6.50074661e-01
-1.21189654e+00 -3.19925606e-01 -5.23316205e-01 6.30049184e-02
-4.83553827e-01 1.09730037e-02 -8.43952775e-01 -4.99112755e-01
-1.08777773e+00 -1.10153496e+00 6.64634049e-01 -1.80213958e-01
2.09878564e-01 -9.38027859e-01 4.28123504e-01 7.15580463e-01
7.50159740e-01 1.10391557e+00 9.44218099e-01 -1.05319059e+00
-5.22817194e-01 -3.84879351e-01 5.66633165e-01 5.85849226e-01
4.87812370e-01 -2.11999819e-01 -1.09946990e+00 -6.36417449e-01
7.86483049e-01 -4.51332599e-01 5.19188225e-01 4.51081879e-02
1.30030215e+00 -1.32099223e+00 -3.26572776e-01 1.02481973e+00
1.64844441e+00 2.54107285e-02 7.69964516e-01 1.33440033e-01
7.97816873e-01 2.79709339e-01 4.85127699e-03 4.37819034e-01
-1.44813523e-01 7.89312869e-02 7.25243390e-01 -3.98224026e-01
-7.48898601e-03 -2.69214481e-01 4.28958595e-01 3.07815552e-01
1.16977997e-01 -1.89611539e-01 -7.07959473e-01 1.01602234e-01
-1.41784179e+00 -1.04646051e+00 1.60661027e-01 2.19389248e+00
1.09451139e+00 -2.39604011e-01 1.22789323e-01 -1.30613565e-01
7.55059361e-01 -1.44982010e-01 -5.41255593e-01 -4.04087782e-01
-1.35001376e-01 6.17758393e-01 1.05556750e+00 1.57770708e-01
-9.50671673e-01 7.02448547e-01 6.34777212e+00 7.06198931e-01
-1.31363833e+00 7.89740801e-01 1.09091890e+00 -1.89898729e-01
-3.53417009e-01 -2.88058162e-01 -3.22626352e-01 6.44114494e-01
1.18088019e+00 -1.72015339e-01 2.48434350e-01 1.15396440e+00
-2.23916113e-01 6.53314888e-01 -8.20931971e-01 8.73193681e-01
1.34896606e-01 -1.82340944e+00 4.24478650e-01 5.51743567e-01
9.29901838e-01 -4.76599373e-02 3.16394001e-01 1.31282220e-02
8.78730059e-01 -1.42759764e+00 3.49095076e-01 1.59968659e-01
1.05123079e+00 -1.12933874e+00 6.82185888e-01 4.44713771e-01
-5.00717044e-01 -1.72409698e-01 -2.37049878e-01 6.51276112e-01
-3.03544611e-01 3.00260186e-01 -8.77135575e-01 4.86184150e-01
4.37464297e-01 -3.07679653e-01 -4.79503483e-01 1.89562783e-01
6.86729699e-03 7.86045313e-01 -9.53587219e-02 5.90909898e-01
2.30213448e-01 1.54191419e-01 4.52352047e-01 7.15563774e-01
2.60803163e-01 1.43351378e-02 8.63128304e-02 1.02597106e+00
-6.08481646e-01 -1.04888305e-01 -1.23534107e+00 -1.86386309e-03
3.85378718e-01 1.20947325e+00 -5.38377404e-01 -3.21893632e-01
-4.03515883e-02 1.18957043e+00 6.08850233e-02 -1.17641212e-02
-9.36454773e-01 -1.89340070e-01 7.06717372e-01 1.42954364e-01
2.98144706e-02 1.42044172e-01 -2.46839464e-01 -1.06781220e+00
-4.07605946e-01 -1.66237319e+00 6.99099839e-01 -2.48887986e-01
-1.42910028e+00 1.09951997e+00 -4.54354703e-01 -1.03914225e+00
-3.37332606e-01 -2.66900659e-01 -7.54385114e-01 7.59197235e-01
-9.49247956e-01 -1.70040429e+00 -1.30365014e-01 1.49619901e+00
1.54586270e-01 -7.22430110e-01 1.34016848e+00 4.33596931e-02
-6.61416829e-01 1.15312672e+00 9.64116752e-02 6.19799972e-01
5.00821531e-01 -9.51680899e-01 2.53802091e-01 1.10916305e+00
1.32763192e-01 7.38426507e-01 5.19358158e-01 -8.47726524e-01
-1.58066154e+00 -1.61931419e+00 2.87670493e-01 -4.33093458e-01
1.04317613e-01 -4.68011320e-01 -5.64963162e-01 9.27962005e-01
4.96357888e-01 3.81839335e-01 1.12698936e+00 -7.61023343e-01
-6.42871499e-01 -8.33216906e-02 -2.04433942e+00 4.05100852e-01
3.27932745e-01 -4.84355509e-01 -4.33564559e-02 8.82479131e-01
9.98404145e-01 -3.10010225e-01 -7.17699707e-01 -6.47715479e-02
-3.92241403e-02 -8.46088469e-01 8.85869563e-01 -9.70610559e-01
2.39638150e-01 -6.39885291e-02 -1.38191283e-01 -1.04909158e+00
1.80667296e-01 -1.21474648e+00 -3.88975888e-01 1.17666924e+00
-1.68905351e-02 -1.08998203e+00 1.05918705e+00 5.52700460e-01
2.69661844e-01 -4.53307092e-01 -1.08293796e+00 -7.77776182e-01
4.21692103e-01 7.23465756e-02 9.86991525e-01 1.23113441e+00
-4.50201422e-01 -1.40209019e-01 -6.10582948e-01 2.92736173e-01
1.00324917e+00 -4.42761749e-01 9.16475892e-01 -5.49778521e-01
-5.54238021e-01 1.37091264e-01 -4.16620582e-01 1.23665102e-01
3.61769497e-02 -8.82594585e-01 -3.61949116e-01 -7.95660794e-01
2.51352452e-02 -4.74504143e-01 -3.37749779e-01 8.75764787e-01
4.91279401e-02 5.02735138e-01 2.71319568e-01 4.52011675e-01
1.69577107e-01 7.58304074e-02 9.14541900e-01 -3.40282023e-01
1.02228798e-01 7.29751438e-02 -9.36898708e-01 3.64468038e-01
1.15055871e+00 -1.06622159e+00 -5.94724536e-01 -2.23294258e-01
-1.39348611e-01 6.80422559e-02 6.53791308e-01 -9.95324910e-01
1.55362800e-01 -6.64116815e-02 4.01789367e-01 1.11945821e-02
-1.24187075e-01 -1.01584625e+00 7.10998178e-01 1.13778222e+00
-3.46276313e-01 1.29868045e-01 -1.57727331e-01 4.79223400e-01
5.17098308e-02 2.44237337e-04 9.81479943e-01 -6.11417830e-01
2.77898222e-01 4.64901835e-01 -2.56508768e-01 -8.63890946e-02
1.50690830e+00 1.58126816e-01 -6.66663647e-01 -2.44483173e-01
-6.42820477e-01 -2.44105935e-01 6.49562418e-01 4.31836285e-02
7.12372065e-01 -1.24062192e+00 -8.93850505e-01 5.32999635e-01
-2.43610665e-01 -7.25103691e-02 1.61675394e-01 3.50510687e-01
-9.35190499e-01 1.21934615e-01 -3.36613506e-01 -1.65029958e-01
-1.55841410e+00 1.10975909e+00 7.51637816e-01 -4.58306789e-01
-7.88585901e-01 8.21733296e-01 2.51969367e-01 -3.10203552e-01
1.18299916e-01 1.67974189e-01 4.09644842e-01 -4.88061070e-01
7.20696270e-01 2.93818384e-01 1.17345281e-01 -4.05660242e-01
-2.40221992e-01 -2.84198552e-01 -8.39721188e-02 -8.00073519e-02
1.03295815e+00 3.21024567e-01 -3.02381873e-01 -2.93286413e-01
1.33862793e+00 2.56512225e-01 -7.93285072e-01 1.44793406e-01
-7.39106655e-01 -4.18430179e-01 -8.66898596e-02 -8.08904469e-01
-1.61099267e+00 5.23469269e-01 8.35918903e-01 3.38304102e-01
1.14629114e+00 -1.02669664e-01 1.16088688e+00 1.64444908e-01
5.97680807e-01 -4.74029481e-01 1.84708431e-01 -2.62781292e-01
5.91839492e-01 -9.25289810e-01 -3.09259802e-01 -1.57711491e-01
-6.55221820e-01 8.84772003e-01 5.73809981e-01 -3.45831156e-01
5.59366643e-01 6.52732432e-01 5.15986621e-01 -2.26587564e-01
-6.72458768e-01 6.28349245e-01 -3.97164911e-01 8.56601238e-01
-1.80094451e-01 1.89897686e-01 5.90759404e-02 7.73157895e-01
-2.01323748e-01 -6.81192353e-02 6.72332168e-01 1.12449074e+00
1.82247519e-01 -1.47244179e+00 -8.61122727e-01 3.57682467e-01
-1.21928847e+00 -1.61766335e-01 -4.77445275e-01 3.93295646e-01
4.04513150e-01 8.23397219e-01 -1.26815423e-01 -4.20168400e-01
-2.10699841e-01 9.05159637e-02 2.40140989e-01 -4.21909481e-01
-1.34564269e+00 -6.68025166e-02 -3.49565387e-01 -5.74651837e-01
-1.46642402e-01 3.84137174e-03 -9.96992290e-01 -5.98583341e-01
-2.41527304e-01 2.58171886e-01 7.53003299e-01 4.62160856e-01
6.22554302e-01 5.32358289e-01 1.18238842e+00 -2.10378617e-01
-1.05976188e+00 -4.52539831e-01 -6.00781620e-01 4.94764537e-01
4.53303367e-01 1.92041799e-01 -4.27662522e-01 1.99754715e-01]
|
[5.982293128967285, 7.07939338684082]
|
c43db37a-3d09-4e32-bfa2-bd2c26a715aa
|
mher-model-based-hindsight-experience-replay
|
2107.00306
| null |
https://arxiv.org/abs/2107.00306v2
|
https://arxiv.org/pdf/2107.00306v2.pdf
|
MHER: Model-based Hindsight Experience Replay
|
Solving multi-goal reinforcement learning (RL) problems with sparse rewards is generally challenging. Existing approaches have utilized goal relabeling on collected experiences to alleviate issues raised from sparse rewards. However, these methods are still limited in efficiency and cannot make full use of experiences. In this paper, we propose Model-based Hindsight Experience Replay (MHER), which exploits experiences more efficiently by leveraging environmental dynamics to generate virtual achieved goals. Replacing original goals with virtual goals generated from interaction with a trained dynamics model leads to a novel relabeling method, model-based relabeling (MBR). Based on MBR, MHER performs both reinforcement learning and supervised learning for efficient policy improvement. Theoretically, we also prove the supervised part in MHER, i.e., goal-conditioned supervised learning with MBR data, optimizes a lower bound on the multi-goal RL objective. Experimental results in several point-based tasks and simulated robotics environments show that MHER achieves significantly higher sample efficiency than previous model-free and model-based multi-goal methods.
|
['Xiu Li', 'Feng Luo', 'Yali Du', 'Lei Han', 'Meng Fang', 'Rui Yang']
|
2021-07-01
| null | null | null | null |
['multi-goal-reinforcement-learning']
|
['methodology']
|
[-1.43389761e-01 1.95303649e-01 -2.27962688e-01 -8.91612843e-02
-9.19316590e-01 -3.32152873e-01 3.64477307e-01 4.50976397e-04
-6.80959582e-01 1.24378455e+00 4.14785296e-01 2.60950495e-02
-4.04215813e-01 -6.13704979e-01 -8.55612040e-01 -7.80286789e-01
-3.50813389e-01 3.29154909e-01 -2.04197004e-01 -4.62330401e-01
3.50771487e-01 1.33413181e-01 -1.57933187e+00 -1.76758513e-01
1.11693203e+00 5.57359755e-01 8.34829748e-01 7.04872370e-01
3.96101624e-01 1.47360682e+00 -3.49700958e-01 5.25334477e-01
4.83104467e-01 -7.13074505e-01 -5.69192171e-01 5.37152290e-02
-2.37950817e-01 -7.24264741e-01 -2.96865404e-01 9.82384980e-01
6.91931725e-01 1.02174163e+00 2.99255788e-01 -1.47429430e+00
-6.34282172e-01 6.93116546e-01 -5.56782603e-01 -2.37134531e-01
5.29748082e-01 1.71152443e-01 8.84497106e-01 -5.31518400e-01
6.20887041e-01 1.34508789e+00 2.02295914e-01 9.21064615e-01
-1.09691572e+00 -4.69503641e-01 4.23865318e-01 1.73679125e-02
-9.10527110e-01 -5.93491457e-02 6.98614120e-01 -1.34916469e-01
1.01166260e+00 -4.70462590e-02 8.66905332e-01 1.03290582e+00
7.81863853e-02 1.12286305e+00 1.38429105e+00 -5.16298652e-01
8.03796887e-01 -9.42274481e-02 -2.65215605e-01 8.55211318e-01
9.00045335e-02 6.83137953e-01 -5.81990957e-01 -6.65147007e-02
1.09545302e+00 2.01334924e-01 -2.02143222e-01 -7.52755761e-01
-1.17913318e+00 1.01843596e+00 5.29688299e-01 -3.70763317e-02
-7.82854974e-01 4.18692708e-01 1.74404010e-01 5.31204164e-01
5.56037799e-02 7.82619119e-01 -3.12685072e-01 -2.71485358e-01
-3.95218939e-01 6.46075308e-01 5.92665315e-01 9.89332259e-01
9.60104823e-01 4.50587630e-01 -3.05228710e-01 8.54044378e-01
2.09786341e-01 6.78996086e-01 9.54729974e-01 -1.30668378e+00
3.43911976e-01 2.45120302e-01 5.46767712e-01 -5.42442977e-01
-6.64774716e-01 -2.36153319e-01 -3.19852561e-01 7.75969088e-01
2.71641791e-01 -6.51316762e-01 -9.17358875e-01 2.05769277e+00
5.78706682e-01 2.11562291e-01 5.46337485e-01 1.22352099e+00
3.98760378e-01 5.70666015e-01 1.74302951e-01 -3.08313638e-01
3.81250232e-01 -1.29141426e+00 -5.42955399e-01 -4.02327925e-01
8.75310242e-01 -1.79388031e-01 1.33148694e+00 5.49706221e-01
-9.40557122e-01 -4.71435279e-01 -9.08509493e-01 3.61414522e-01
8.33105221e-02 -5.82429245e-02 9.11726356e-01 2.75560141e-01
-1.01863468e+00 8.92235279e-01 -8.35884869e-01 -2.39331931e-01
1.74754158e-01 3.78088266e-01 -2.63060898e-01 -1.54491350e-01
-8.08775246e-01 9.45757687e-01 5.43774366e-01 -3.30526441e-01
-1.65822148e+00 -3.21587741e-01 -1.19060481e+00 -7.92002380e-02
8.54685068e-01 -4.51630265e-01 1.59778249e+00 -7.45132506e-01
-1.98174822e+00 2.29170591e-01 2.48540208e-01 -4.99645203e-01
2.08620638e-01 -4.95226771e-01 2.33406667e-02 1.22589029e-01
3.34940910e-01 6.75961673e-01 8.90888870e-01 -1.57796454e+00
-6.93123758e-01 -2.16284409e-01 3.68606031e-01 8.38886023e-01
-2.02678010e-01 -4.16880339e-01 1.63610369e-01 -2.59964734e-01
-1.00748599e-01 -1.03407061e+00 -9.02320504e-01 -3.09219360e-01
2.67558713e-02 -9.72388536e-02 4.51171279e-01 -4.05484647e-01
6.48519397e-01 -1.91771305e+00 3.68713707e-01 9.94532704e-02
1.34784237e-01 1.90177530e-01 -5.35355926e-01 5.37606537e-01
4.08144474e-01 -4.57059026e-01 1.45325899e-01 -4.80362356e-01
1.21679679e-01 5.49840927e-01 -4.59807366e-01 3.85925144e-01
-2.45950714e-01 9.20533359e-01 -1.59460866e+00 -4.17627484e-01
5.75452089e-01 -6.97744219e-03 -8.31166208e-01 5.02761304e-01
-5.10102332e-01 8.93652618e-01 -7.75345504e-01 5.80253959e-01
2.60135502e-01 -7.76563287e-02 3.42387617e-01 6.40826583e-01
-2.32431553e-02 2.73274500e-02 -1.11754656e+00 2.11690259e+00
-7.42242515e-01 -3.61942016e-02 3.32310535e-02 -9.60277736e-01
9.66287732e-01 1.36273682e-01 6.37131453e-01 -9.51128483e-01
7.60489702e-02 8.97806883e-02 -3.07137907e-01 -5.08990765e-01
9.64524031e-01 -1.75455630e-01 -3.20517749e-01 5.38779974e-01
3.62950832e-01 -2.90259629e-01 5.95998541e-02 2.54008234e-01
1.10611784e+00 9.90908682e-01 6.02205098e-01 1.51014268e-01
-1.05131984e-01 1.78934366e-01 7.83102751e-01 1.28784978e+00
-5.37569582e-01 2.87559718e-01 -2.63691042e-02 4.01867628e-02
-8.50547373e-01 -1.02739680e+00 5.88413358e-01 1.39230680e+00
5.31505466e-01 -2.32254118e-01 -3.24764580e-01 -6.01490974e-01
4.74613532e-02 9.28566873e-01 -5.60312510e-01 -2.68363506e-01
-6.14758015e-01 -5.53290784e-01 -2.17782743e-02 5.04727304e-01
4.30224955e-01 -1.58727837e+00 -1.10116148e+00 4.99207735e-01
-1.68512434e-01 -7.96538353e-01 -2.97745824e-01 3.59915674e-01
-8.47227871e-01 -8.08010280e-01 -8.82333755e-01 -6.80281818e-01
5.03524303e-01 5.36412001e-01 7.21569955e-01 -2.05433413e-01
-1.34819731e-01 8.19070458e-01 -7.71274209e-01 -3.39644998e-01
-1.44222990e-01 -4.29165125e-01 4.96066809e-01 -3.67734760e-01
-5.95373809e-02 -8.30399811e-01 -3.99271488e-01 3.23226536e-03
-5.55794597e-01 5.57416305e-02 7.80682325e-01 1.15208888e+00
7.64021575e-01 -2.11048961e-01 1.03539145e+00 -4.29240525e-01
9.44245756e-01 -5.73179007e-01 -7.06828833e-01 9.04864967e-02
-6.35943055e-01 3.81404072e-01 7.46964216e-01 -8.85698795e-01
-1.18484688e+00 1.44799471e-01 9.20840949e-02 -6.97849512e-01
-6.97057098e-02 4.37289834e-01 1.59586325e-01 -1.70487672e-01
8.37180853e-01 3.41331005e-01 -3.03342799e-03 -1.53076679e-01
6.51845813e-01 3.24684411e-01 3.05869728e-01 -8.93577635e-01
4.21147257e-01 3.38315070e-01 3.04332450e-02 -5.56221604e-01
-9.75329876e-01 -5.21873116e-01 6.61249608e-02 -5.64798355e-01
5.82063735e-01 -1.02119815e+00 -1.03322029e+00 2.23810837e-01
-4.98618811e-01 -1.19449830e+00 -8.18250477e-01 1.14934409e+00
-1.47516143e+00 1.61457837e-01 -4.58814204e-01 -1.39406598e+00
-1.36089399e-01 -8.01052809e-01 7.99172342e-01 5.38181245e-01
1.13332681e-01 -8.83033991e-01 3.75852466e-01 -2.21353322e-02
2.86855251e-01 3.90637815e-01 2.45305091e-01 -3.99912924e-01
-4.18425739e-01 1.99029386e-01 2.42506713e-01 8.58674496e-02
-8.66740718e-02 -9.02212977e-01 -6.28637195e-01 -6.07388556e-01
9.83917341e-02 -1.33136404e+00 6.51968777e-01 5.75266838e-01
7.86690652e-01 -3.27728301e-01 -1.13031305e-01 2.51521438e-01
1.62071490e+00 3.01598996e-01 4.91121024e-01 4.74320024e-01
4.97635514e-01 3.14478010e-01 1.52651632e+00 1.15235281e+00
3.84652406e-01 4.16801989e-01 9.44597721e-01 1.32409215e-01
1.59109861e-01 -7.46102631e-01 6.56435668e-01 3.53486061e-01
-1.94749758e-01 -1.83803514e-02 -3.26902688e-01 6.00640655e-01
-2.51956940e+00 -1.06400609e+00 3.53765160e-01 2.28295946e+00
6.63132429e-01 -1.83634549e-01 3.14436048e-01 -3.05189699e-01
4.88520443e-01 2.34651640e-02 -1.02818394e+00 -1.41274869e-01
1.68024793e-01 2.39846110e-01 4.68289226e-01 5.65323770e-01
-7.84268737e-01 1.23158860e+00 5.93036890e+00 7.85093307e-01
-6.64437532e-01 1.89539671e-01 -2.74713635e-02 -4.26800191e-01
-3.66455056e-02 9.32468176e-02 -7.42759407e-01 8.47974941e-02
5.62220454e-01 -2.63705611e-01 1.05916512e+00 1.36098766e+00
4.69740659e-01 -4.61022466e-01 -8.35500658e-01 9.61459935e-01
-3.97016853e-02 -8.20583224e-01 -4.13185984e-01 1.26092777e-01
9.42939699e-01 1.34740453e-02 5.26578128e-02 1.19175339e+00
1.48648655e+00 -7.43514240e-01 8.24519813e-01 5.33420622e-01
4.11953121e-01 -8.29831123e-01 2.62629837e-01 7.84843922e-01
-9.29533243e-01 -5.92815161e-01 -6.00864768e-01 -5.96348643e-01
3.07641923e-01 1.63802311e-01 -8.39801252e-01 6.46380663e-01
4.26534474e-01 7.49971032e-01 1.08554631e-01 8.96120012e-01
-5.43122232e-01 2.63794303e-01 -1.49722949e-01 -4.92967516e-01
5.76756001e-01 -3.42255205e-01 6.22586787e-01 4.13422883e-01
4.48467076e-01 8.93887505e-02 8.47290099e-01 6.97566867e-01
2.99702555e-01 7.74429590e-02 -7.29789317e-01 -7.26932362e-02
3.94787490e-01 1.28498995e+00 -3.63497049e-01 -2.49172598e-01
-3.67520824e-02 9.52830374e-01 9.20331120e-01 4.15995270e-01
-8.42435896e-01 -1.54076219e-01 6.21005714e-01 -4.52006221e-01
2.55564630e-01 -4.71765578e-01 2.52008855e-01 -1.20378828e+00
-1.45499468e-01 -7.59174168e-01 3.03400666e-01 -9.95097041e-01
-9.40932870e-01 2.65165508e-01 -5.64121902e-02 -1.41297054e+00
-7.20961988e-01 -2.69964248e-01 -2.65713423e-01 6.13496125e-01
-1.57641840e+00 -1.01927400e+00 -2.45706692e-01 8.14447224e-01
5.63889802e-01 -1.33859172e-01 9.90432322e-01 -1.85285136e-01
-1.65052071e-01 2.39117652e-01 2.85263181e-01 -4.04203057e-01
6.73727453e-01 -1.45748651e+00 -2.47867957e-01 6.91443503e-01
-1.08743161e-01 3.49630684e-01 7.73371339e-01 -7.60380447e-01
-1.38260400e+00 -1.19267988e+00 7.83370361e-02 1.27350632e-02
5.90267658e-01 -4.41039866e-03 -4.41032469e-01 8.42300832e-01
-3.72964963e-02 -1.44559845e-01 4.85521644e-01 1.82650521e-01
7.46469479e-03 2.70449519e-01 -1.32202232e+00 9.10920501e-01
1.19278193e+00 -1.22413881e-01 -6.37911320e-01 4.16655898e-01
8.85001242e-01 -6.41011357e-01 -6.29371643e-01 1.49157658e-01
3.94007266e-01 -7.60501206e-01 8.44098687e-01 -8.83356690e-01
4.89619046e-01 -1.43847078e-01 -3.26343417e-01 -1.92451525e+00
-3.06088448e-01 -7.90756106e-01 -4.24954116e-01 7.33935475e-01
-1.02188483e-01 -6.01929605e-01 5.84511280e-01 3.42290908e-01
-4.25913900e-01 -6.88685060e-01 -6.38956904e-01 -1.26520169e+00
4.85366061e-02 -3.00652087e-01 4.69911873e-01 7.70285428e-01
4.24860060e-01 2.07802504e-01 -1.07214320e+00 1.95350815e-02
8.04469824e-01 2.69851148e-01 1.02546084e+00 -7.70629108e-01
-8.44382584e-01 7.61635751e-02 2.02049688e-02 -1.27612865e+00
2.34750688e-01 -6.86602652e-01 5.57659328e-01 -1.82567251e+00
1.09402962e-01 -7.47473598e-01 -3.84277612e-01 6.22166157e-01
-2.64673293e-01 -2.24224821e-01 4.15540338e-01 1.91292211e-01
-1.26992130e+00 1.14039075e+00 1.76245117e+00 2.92850375e-01
-8.50816846e-01 -1.15456820e-01 -7.93231905e-01 6.52235389e-01
1.20454752e+00 -4.87221450e-01 -6.23145819e-01 -1.32890373e-01
7.07045421e-02 5.78890741e-01 2.60069311e-01 -1.21004105e+00
1.33661069e-02 -7.41715014e-01 4.96605150e-02 -3.47800583e-01
6.45869493e-01 -4.62270886e-01 -3.36815119e-01 5.14187574e-01
-6.42580152e-01 -3.30148458e-01 -4.83703539e-02 1.03251743e+00
1.58762366e-01 -4.37432170e-01 5.34087837e-01 -5.20789444e-01
-1.07438552e+00 2.68951029e-01 -4.88750994e-01 2.00328246e-01
1.25326622e+00 2.00210541e-01 -1.30536914e-01 -7.43789375e-01
-1.00254679e+00 6.63354635e-01 2.47968510e-01 2.84549892e-01
8.57104301e-01 -1.46060884e+00 -4.78554994e-01 -1.66236579e-01
1.28081650e-01 -8.64650980e-02 4.27062511e-01 4.97211695e-01
1.85757756e-01 7.85949454e-02 -4.90941137e-01 -1.73041016e-01
-9.71036732e-01 8.77052844e-01 8.72689635e-02 -7.79397190e-01
-7.46343851e-01 6.56797647e-01 -2.08441936e-03 -8.00236702e-01
2.15158001e-01 -5.33616580e-02 -3.07595551e-01 -5.08586586e-01
5.13631999e-01 4.52083588e-01 -5.29295146e-01 -1.31153256e-01
2.15713963e-01 6.25500381e-02 -3.01153883e-02 -6.59922719e-01
1.38951123e+00 -1.40413389e-01 5.53407848e-01 5.11966527e-01
6.12209618e-01 -3.27390343e-01 -1.87955856e+00 -3.82224470e-01
-3.67495924e-01 -6.57887340e-01 -3.13825458e-02 -7.81344354e-01
-5.44625521e-01 5.19192219e-01 5.04638970e-01 -1.36611417e-01
9.85819399e-01 -2.15307325e-01 7.02570200e-01 8.72102201e-01
1.22572923e+00 -1.57170284e+00 8.35496664e-01 6.09530687e-01
9.08696175e-01 -1.38783979e+00 -2.28708580e-01 1.65762559e-01
-1.11578786e+00 5.37157834e-01 9.46829855e-01 -4.90325660e-01
2.15195581e-01 -6.85382560e-02 -1.60845369e-01 6.31437302e-02
-6.79384172e-01 -8.18634152e-01 -4.71446663e-01 9.97217298e-01
-2.99586177e-01 2.79199779e-01 -2.05739453e-01 6.05439484e-01
-1.44557029e-01 3.50470930e-01 7.21741438e-01 1.47913790e+00
-9.42221522e-01 -9.24996197e-01 -3.17485690e-01 4.36909378e-01
-4.44426239e-02 1.03183985e-01 1.43701091e-01 5.89289248e-01
-3.28769714e-01 1.31197393e+00 -3.14727843e-01 -5.47377050e-01
7.83156306e-02 -3.58349264e-01 9.63798881e-01 -7.43277550e-01
-3.76379102e-01 -1.20504573e-01 -3.63355093e-02 -9.28643346e-01
-3.80151570e-01 -5.50626516e-01 -1.75550890e+00 -1.15722410e-01
-3.30668658e-01 1.81252360e-01 5.97801864e-01 9.87069726e-01
3.39042515e-01 6.59828722e-01 7.95137465e-01 -1.17437720e+00
-1.24076712e+00 -9.69171524e-01 -1.03269732e+00 4.29608136e-01
3.56443495e-01 -1.08951259e+00 -2.86680341e-01 -3.51365983e-01]
|
[4.097320079803467, 1.704741358757019]
|
e5dc285f-f063-435f-afa7-bc8667bd4544
|
disentangle-align-and-fuse-for-multimodal-and
|
1911.04417
| null |
https://arxiv.org/abs/1911.04417v5
|
https://arxiv.org/pdf/1911.04417v5.pdf
|
Disentangle, align and fuse for multimodal and semi-supervised image segmentation
|
Magnetic resonance (MR) protocols rely on several sequences to assess pathology and organ status properly. Despite advances in image analysis, we tend to treat each sequence, here termed modality, in isolation. Taking advantage of the common information shared between modalities (an organ's anatomy) is beneficial for multi-modality processing and learning. However, we must overcome inherent anatomical misregistrations and disparities in signal intensity across the modalities to obtain this benefit. We present a method that offers improved segmentation accuracy of the modality of interest (over a single input model), by learning to leverage information present in other modalities, even if few (semi-supervised) or no (unsupervised) annotations are available for this specific modality. Core to our method is learning a disentangled decomposition into anatomical and imaging factors. Shared anatomical factors from the different inputs are jointly processed and fused to extract more accurate segmentation masks. Image misregistrations are corrected with a Spatial Transformer Network, which non-linearly aligns the anatomical factors. The imaging factor captures signal intensity characteristics across different modality data and is used for image reconstruction, enabling semi-supervised learning. Temporal and slice pairing between inputs are learned dynamically. We demonstrate applications in Late Gadolinium Enhanced (LGE) and Blood Oxygenation Level Dependent (BOLD) cardiac segmentation, as well as in T2 abdominal segmentation. Code is available at https://github.com/vios-s/multimodal_segmentation.
|
['Rohan Dharmakumar', 'Scott Semple', 'Chengjia Wang', 'Agisilaos Chartsias', 'Sotirios A. Tsaftaris', 'Giorgos Papanastasiou', 'David E. Newby']
|
2019-11-11
| null | null | null | null |
['cardiac-segmentation']
|
['medical']
|
[ 6.78028107e-01 -8.00820962e-02 -3.29855055e-01 -5.27073920e-01
-1.01442397e+00 -8.61349761e-01 3.42879385e-01 2.56903321e-01
-4.64654893e-01 4.54839051e-01 2.59468406e-01 -1.96219802e-01
-1.95003480e-01 -2.79073268e-01 -4.46583569e-01 -9.73409772e-01
-2.73892283e-01 2.87356198e-01 1.33587539e-01 2.16617092e-01
-8.72424394e-02 5.42532146e-01 -7.60459244e-01 4.67463046e-01
1.02149379e+00 8.66512895e-01 4.36701536e-01 8.34550023e-01
-6.85335472e-02 7.80317664e-01 1.74015462e-02 -1.09272087e-02
3.06078702e-01 -6.79491580e-01 -1.03827763e+00 4.15354334e-02
4.44202304e-01 -8.44100937e-02 -2.83218712e-01 1.14319336e+00
5.26027024e-01 2.27862932e-02 6.10907793e-01 -7.60343075e-01
-1.11059517e-01 7.18500733e-01 -7.10863173e-01 7.51624107e-01
-1.95061907e-01 2.87343502e-01 3.81378919e-01 -6.46127582e-01
6.09574378e-01 4.40971255e-01 6.95501268e-01 4.45403934e-01
-1.57552516e+00 -3.74867111e-01 -3.18231471e-02 -1.73322201e-01
-1.03833389e+00 -4.21254754e-01 6.62448108e-01 -8.17365766e-01
6.62797928e-01 3.52907509e-01 5.37336349e-01 6.72121942e-01
4.66707319e-01 5.70033908e-01 1.40452051e+00 -2.82467008e-01
-2.86580414e-01 -1.41035616e-01 9.43102017e-02 7.61604488e-01
-8.91836956e-02 3.23028229e-02 -5.15006542e-01 -5.51714972e-02
8.70109618e-01 3.10959350e-02 -5.55492878e-01 -5.82011461e-01
-1.75831962e+00 4.51021105e-01 5.57385564e-01 7.67149746e-01
-6.52901888e-01 2.87789349e-02 4.00894642e-01 1.23413712e-01
2.65181035e-01 3.64844561e-01 -3.17266375e-01 1.70780972e-01
-1.27471983e+00 -4.30677652e-01 4.40774679e-01 3.31639022e-01
7.45233119e-01 -3.42875384e-02 -1.49755955e-01 8.33828390e-01
2.05348313e-01 3.91264468e-01 7.68954396e-01 -1.03802514e+00
9.47547406e-02 3.25453192e-01 -3.44179153e-01 -7.17168093e-01
-7.94282913e-01 -5.58818877e-01 -1.01503015e+00 1.74803898e-01
7.53272295e-01 -1.74302503e-01 -1.26042283e+00 1.74601877e+00
3.78109932e-01 2.69440562e-01 -3.93163979e-01 1.31679726e+00
6.36869252e-01 -1.18348068e-02 4.18518513e-01 -3.32343429e-01
1.43824267e+00 -7.60922372e-01 -5.73017836e-01 -3.67453337e-01
7.67504454e-01 -6.54459178e-01 7.28993952e-01 2.00195044e-01
-1.25598168e+00 -1.56418502e-01 -9.46337640e-01 2.25999624e-01
-1.51842356e-01 -1.55736491e-01 6.32826865e-01 5.70261598e-01
-1.00127709e+00 6.74882233e-01 -1.40314615e+00 -2.75323521e-02
4.99307722e-01 7.46077418e-01 -6.02127612e-01 -2.20291868e-01
-1.16546690e+00 1.25766969e+00 2.44653538e-01 2.90203571e-01
-7.84961224e-01 -1.33571243e+00 -7.99776971e-01 -4.37790304e-01
3.24310422e-01 -7.76796222e-01 9.34128225e-01 -1.25020707e+00
-1.12087190e+00 1.12625074e+00 -5.47718294e-02 -2.94203371e-01
4.66175646e-01 3.50222290e-01 -4.01132315e-01 7.00797856e-01
1.93024263e-01 6.84260130e-01 7.78275669e-01 -1.20683002e+00
-1.82921916e-01 -6.81110740e-01 -2.46125162e-01 4.40767586e-01
1.66540354e-01 1.96633175e-01 -2.43744686e-01 -5.05987704e-01
5.97230256e-01 -9.72772777e-01 -4.15562212e-01 1.84307452e-02
-2.68430948e-01 7.17696249e-01 3.48966718e-01 -1.21158326e+00
8.55273366e-01 -1.88430619e+00 3.65575165e-01 4.11244750e-01
6.97741270e-01 -1.69877797e-01 1.51421074e-02 -3.21599543e-01
-4.68664825e-01 4.35043545e-03 -6.60200775e-01 -4.82448153e-02
-5.19046962e-01 1.44980192e-01 3.99223685e-01 8.87665510e-01
4.92066443e-02 1.08829963e+00 -9.74428594e-01 -6.82609141e-01
5.16241729e-01 4.63463813e-01 -1.72685668e-01 -2.82360557e-02
3.31366628e-01 1.43710423e+00 -4.61648017e-01 6.25228643e-01
5.16415656e-01 -2.80963421e-01 5.24437606e-01 -7.20233679e-01
-9.02978629e-02 1.07729234e-01 -7.84880996e-01 2.26576066e+00
-4.34926242e-01 1.05772905e-01 5.06513298e-01 -1.17711377e+00
4.62969005e-01 6.95976913e-01 1.32122552e+00 -8.60264003e-01
2.99563140e-01 4.66128349e-01 4.45814252e-01 -6.62885070e-01
-1.34947553e-01 -6.16136968e-01 1.89419389e-01 5.21505713e-01
4.12063330e-01 -3.69422585e-02 1.52679339e-01 1.62320659e-01
1.05698657e+00 1.59694731e-01 3.87866274e-02 -5.48118353e-01
4.51440632e-01 4.79664765e-02 5.66085875e-01 7.10627496e-01
-5.90367854e-01 9.20141757e-01 3.80084246e-01 -3.12280923e-01
-9.84834671e-01 -1.16463351e+00 -4.43936378e-01 8.13644648e-01
9.87621993e-02 1.57698020e-01 -6.12712383e-01 -7.57680357e-01
-1.90362751e-01 1.66612729e-01 -7.96767116e-01 1.80092454e-02
-8.43003809e-01 -1.18224442e+00 6.64454520e-01 4.87578303e-01
-2.90424265e-02 -7.92049766e-01 -8.05153012e-01 3.83907169e-01
-4.08391982e-01 -1.16656888e+00 -5.46397507e-01 5.57374775e-01
-1.20462847e+00 -1.14702630e+00 -8.91858041e-01 -3.20746988e-01
8.28539610e-01 3.20827290e-02 1.18452179e+00 4.18482013e-02
-5.08397996e-01 6.31686866e-01 -8.90991911e-02 4.63560037e-02
-4.03254420e-01 1.41966254e-01 -1.52323902e-01 1.03292398e-01
-2.18663305e-01 -7.31388986e-01 -8.45278203e-01 1.70800328e-01
-1.02947605e+00 3.36186349e-01 6.51972055e-01 9.54629898e-01
7.11866081e-01 -5.45642197e-01 3.52906853e-01 -1.06063759e+00
7.82513022e-02 -5.53734064e-01 -2.30004609e-01 5.33029497e-01
-3.13569516e-01 3.50886583e-02 1.77913934e-01 -4.03554827e-01
-1.04234016e+00 2.60883719e-01 1.12164043e-01 -4.36777115e-01
-3.78070503e-01 7.57678866e-01 2.34934121e-01 -5.55933535e-01
6.58970118e-01 2.26047486e-01 3.63961965e-01 -2.11742580e-01
1.72692835e-01 9.32033062e-02 8.22602987e-01 -6.74981654e-01
3.97097886e-01 6.66574061e-01 1.89425439e-01 -4.84541804e-01
-7.55559921e-01 -4.71235722e-01 -1.43777931e+00 -4.87066299e-01
1.02342379e+00 -6.22622252e-01 -3.03140074e-01 4.37547863e-01
-6.83764756e-01 -5.12529552e-01 -2.85993248e-01 9.19241726e-01
-3.80710274e-01 3.58746529e-01 -7.09753036e-01 -2.05886081e-01
-2.98485219e-01 -1.53330302e+00 6.32122695e-01 2.51547933e-01
-1.42339095e-01 -1.55067408e+00 9.06060487e-02 4.64531153e-01
6.04485273e-01 5.95820785e-01 8.00170600e-01 -5.73376477e-01
-4.65851665e-01 7.76748359e-02 -9.38551575e-02 3.08558166e-01
3.34950268e-01 -4.30025846e-01 -8.31034005e-01 -8.59010294e-02
2.27384597e-01 -1.26517907e-01 8.12873185e-01 9.48756993e-01
1.05504251e+00 2.20945641e-01 2.58921459e-02 8.80591512e-01
1.28723443e+00 5.48731955e-03 3.80518794e-01 2.54187360e-03
1.02353573e+00 7.52428532e-01 1.59831673e-01 -4.48225476e-02
3.51747006e-01 4.87196505e-01 2.20020086e-01 -6.49308383e-01
-3.25025350e-01 2.01459244e-01 -7.93526024e-02 9.08586919e-01
-2.06403583e-01 4.12239552e-01 -1.34805107e+00 5.49152017e-01
-1.48762226e+00 -6.79933608e-01 -2.24990681e-01 2.23107243e+00
1.01845598e+00 -2.70181239e-01 9.19706374e-02 -2.03520611e-01
6.55994356e-01 9.44885910e-02 -7.15529323e-01 -2.12182943e-02
-2.22476482e-01 2.65281707e-01 8.42056155e-01 6.84656978e-01
-1.20283413e+00 4.64328706e-01 6.34059906e+00 3.29750061e-01
-1.56375277e+00 6.00710928e-01 8.10612857e-01 -9.31006968e-02
-4.31093752e-01 4.72257333e-03 5.05433679e-02 2.78294504e-01
8.30811620e-01 2.55952328e-02 5.59764385e-01 1.08215427e-02
2.03894630e-01 -4.66862261e-01 -1.00107455e+00 7.26169765e-01
7.52319172e-02 -1.23522735e+00 -5.21206319e-01 -8.06653425e-02
6.29003882e-01 3.87679487e-01 1.40553012e-01 -2.30487436e-01
-1.34776026e-01 -1.10954642e+00 3.47542763e-01 7.13003218e-01
9.84465063e-01 -2.27377847e-01 5.74629664e-01 9.77736190e-02
-9.36643839e-01 2.91532218e-01 3.28673065e-01 4.89451170e-01
2.16964349e-01 6.04454517e-01 -5.87343395e-01 8.06678236e-01
4.24472481e-01 7.36779392e-01 -5.24766266e-01 9.87087309e-01
-2.50471476e-02 4.92070198e-01 -2.65947014e-01 9.81229782e-01
1.35329083e-01 -2.75188118e-01 6.21687472e-01 1.16496420e+00
-1.39367711e-02 3.96182746e-01 2.76019484e-01 8.62911761e-01
1.40214011e-01 3.84649299e-02 -2.12636560e-01 1.23048283e-01
-6.15257099e-02 1.78951144e+00 -1.11243677e+00 -4.64968652e-01
-5.40911138e-01 8.05083930e-01 -4.95232344e-02 4.60203409e-01
-7.43297517e-01 3.34704459e-01 1.11365609e-01 2.74529576e-01
-2.28803873e-01 -2.28620738e-01 -7.59956360e-01 -1.40683234e+00
-2.57742882e-01 -7.42073596e-01 6.54213011e-01 -6.22746170e-01
-1.29682338e+00 4.66319770e-01 1.41875029e-01 -9.51781273e-01
-9.99076217e-02 -3.43150198e-01 -4.44338500e-01 1.27116728e+00
-1.62032735e+00 -1.30572569e+00 -2.53052801e-01 6.42078042e-01
-1.93127729e-02 3.79048765e-01 6.96961999e-01 6.86204731e-01
-4.15182292e-01 3.25281560e-01 -2.05468029e-01 2.10475191e-01
8.66360724e-01 -1.42668605e+00 -4.42568749e-01 8.34518135e-01
-9.86263603e-02 6.07281864e-01 3.52868795e-01 -6.95082009e-01
-1.33412457e+00 -7.02667475e-01 4.73960757e-01 -3.31362784e-01
8.47863257e-01 1.31777182e-01 -1.15204859e+00 6.37055397e-01
1.66130990e-01 7.12897539e-01 9.71437931e-01 -1.47696227e-01
-5.25679849e-02 5.26169054e-02 -1.33035588e+00 1.91122025e-01
6.31409228e-01 -6.16454124e-01 -4.59309667e-01 1.68298498e-01
2.05829024e-01 -9.21736836e-01 -1.46962273e+00 5.32154262e-01
6.30666435e-01 -7.32935369e-01 9.41342056e-01 -4.86219645e-01
2.64361590e-01 -3.37722898e-01 1.47440836e-01 -1.19462490e+00
-1.30663678e-01 -3.46910805e-01 5.10995276e-02 6.90135300e-01
4.49404776e-01 -6.24137580e-01 6.44376218e-01 1.04807258e+00
-3.91317666e-01 -5.89174449e-01 -9.96194303e-01 -1.95056126e-01
3.13155621e-01 -4.53312963e-01 1.49009228e-01 1.48253357e+00
5.11909761e-02 -4.44140770e-02 -2.18311310e-01 2.73573458e-01
8.84463370e-01 1.05609812e-01 -7.42200911e-02 -9.02102947e-01
-2.66399115e-01 -4.06063318e-01 -1.38849452e-01 -4.22554135e-01
3.86554785e-02 -1.49033678e+00 -5.28175198e-02 -1.27123511e+00
4.58663613e-01 -6.22915983e-01 -8.73649836e-01 7.95231640e-01
-1.51824072e-01 6.95133269e-01 1.89499736e-01 3.19839656e-01
-3.32874298e-01 4.45336327e-02 1.69495714e+00 -2.17665061e-02
-1.28565043e-01 -2.81491369e-01 -5.76139688e-01 6.88794374e-01
6.78589225e-01 -5.72967768e-01 -1.20245270e-01 -5.13362527e-01
-2.13079602e-01 6.50470495e-01 6.89988315e-01 -9.40842092e-01
3.44153255e-01 -3.32408361e-02 6.82559073e-01 -2.60630161e-01
3.79999243e-02 -7.87684619e-01 2.27908075e-01 5.16510844e-01
-4.42331910e-01 -7.46384785e-02 1.47614464e-01 -2.81426515e-02
-2.86053807e-01 -2.16744900e-01 1.02532506e+00 -4.43364590e-01
-3.72593462e-01 5.79288006e-01 -3.44019443e-01 1.91869274e-01
6.68983161e-01 -1.75151780e-01 -1.73649453e-02 -4.54213517e-03
-1.47960174e+00 2.89844453e-01 3.04954916e-01 1.35204762e-01
4.34113294e-01 -1.06672978e+00 -6.63831294e-01 5.25060557e-02
-3.01265895e-01 -1.77987814e-01 1.04517782e+00 2.02486706e+00
-3.58007014e-01 2.25730136e-01 -4.59626406e-01 -1.01005638e+00
-9.59688187e-01 1.16666690e-01 1.09037924e+00 -5.10297060e-01
-4.95491445e-01 5.56661725e-01 2.75577545e-01 -6.44163847e-01
-3.23559433e-01 -9.17462483e-02 -3.83891873e-02 2.02896409e-02
3.43484133e-01 3.33702639e-02 2.68969238e-01 -9.77487445e-01
-5.13472557e-01 5.87478817e-01 -6.96502952e-03 -2.98404604e-01
1.28642178e+00 -3.65704745e-01 -2.93796986e-01 5.47394335e-01
1.14679170e+00 -1.95291519e-01 -1.31921017e+00 -4.35025573e-01
-6.02808930e-02 -4.00662869e-01 5.23455501e-01 -1.21479309e+00
-1.43965864e+00 1.04548621e+00 8.60714972e-01 -1.22345626e-01
1.16936493e+00 -1.37841970e-01 5.83206892e-01 -4.83463645e-01
1.26737803e-01 -6.77055359e-01 -3.18710804e-01 2.27496058e-01
5.57722688e-01 -1.26723826e+00 4.56279032e-02 -3.49438995e-01
-9.93203998e-01 1.13108659e+00 3.90135318e-01 1.16985351e-01
5.21575034e-01 7.12807536e-01 4.20710415e-01 -3.42741609e-01
-2.51118749e-01 -1.04892224e-01 6.35088921e-01 5.43131590e-01
7.24825442e-01 7.45136216e-02 -1.71101972e-01 4.41350222e-01
3.36430639e-01 -4.84923050e-02 2.37366453e-01 9.68252718e-01
-4.80081551e-02 -1.22657561e+00 -3.82800728e-01 5.76408267e-01
-8.61038685e-01 -2.28168696e-01 1.42296225e-01 5.46078324e-01
2.27074251e-01 3.40949267e-01 -1.27317846e-01 5.17292544e-02
-3.63044953e-03 2.08295479e-01 9.12060380e-01 -5.89612424e-01
-9.84611511e-01 4.55393225e-01 -2.24750683e-01 -6.63204134e-01
-7.93921232e-01 -9.00114059e-01 -1.47881913e+00 1.47229135e-01
4.89431992e-02 -2.10177958e-01 7.97506332e-01 1.07322526e+00
1.26417400e-02 7.86173344e-01 6.06529474e-01 -1.02154708e+00
-1.26146376e-01 -6.68443263e-01 -5.64770222e-01 4.47153747e-01
3.46814364e-01 -5.64407051e-01 -2.55183846e-01 2.91861504e-01]
|
[13.999917030334473, -2.352886438369751]
|
9aa2e826-fde2-48cf-81c6-5b5e886f16df
|
won-t-get-fooled-again-answering-questions
|
2307.02394
| null |
https://arxiv.org/abs/2307.02394v1
|
https://arxiv.org/pdf/2307.02394v1.pdf
|
Won't Get Fooled Again: Answering Questions with False Premises
|
Pre-trained language models (PLMs) have shown unprecedented potential in various fields, especially as the backbones for question-answering (QA) systems. However, they tend to be easily deceived by tricky questions such as "How many eyes does the sun have?". Such frailties of PLMs often allude to the lack of knowledge within them. In this paper, we find that the PLMs already possess the knowledge required to rebut such questions, and the key is how to activate the knowledge. To systematize this observation, we investigate the PLMs' responses to one kind of tricky questions, i.e., the false premises questions (FPQs). We annotate a FalseQA dataset containing 2365 human-written FPQs, with the corresponding explanations for the false premises and the revised true premise questions. Using FalseQA, we discover that PLMs are capable of discriminating FPQs by fine-tuning on moderate numbers (e.g., 256) of examples. PLMs also generate reasonable explanations for the false premise, which serve as rebuttals. Further replaying a few general questions during training allows PLMs to excel on FPQs and general questions simultaneously. Our work suggests that once the rebuttal ability is stimulated, knowledge inside the PLMs can be effectively utilized to handle FPQs, which incentivizes the research on PLM-based QA systems.
|
['Maosong Sun', 'Zhiyuan Liu', 'Xingyi Cheng', 'Huadong Wang', 'Yifan Luo', 'Shengding Hu']
|
2023-07-05
| null | null | null | null |
['question-answering']
|
['natural-language-processing']
|
[ 8.09981376e-02 7.25806594e-01 1.25948250e-01 -4.90011483e-01
-9.30626571e-01 -1.03516722e+00 5.71414232e-01 -6.88241348e-02
-1.32671371e-01 8.89255047e-01 -1.07035577e-01 -9.31708395e-01
-8.93637538e-03 -9.40689504e-01 -8.41644943e-01 -3.04177761e-01
4.63530540e-01 5.25139213e-01 4.15484875e-01 -7.99535990e-01
4.58773375e-01 2.36026317e-01 -1.29020488e+00 7.56729841e-01
1.42315102e+00 6.64193809e-01 8.37728903e-02 7.90931106e-01
-5.84514320e-01 1.13852775e+00 -1.15786934e+00 -1.05944216e+00
-2.08517969e-01 -4.39730048e-01 -1.41892958e+00 -2.23510906e-01
6.01831079e-01 -4.77986991e-01 -1.94255725e-01 9.42499638e-01
1.20271593e-01 -3.80451456e-02 4.03033823e-01 -1.44429839e+00
-1.20497990e+00 6.48904204e-01 4.82234098e-02 2.24564761e-01
5.95911264e-01 4.06857133e-01 1.29268003e+00 -8.74745727e-01
1.14951402e-01 1.42321980e+00 3.86760712e-01 1.01158798e+00
-7.75222361e-01 -4.89733368e-01 -9.10919160e-02 4.78972286e-01
-1.06828821e+00 -5.08587778e-01 4.65969950e-01 -4.13648821e-02
1.02467537e+00 6.75781012e-01 5.52749187e-02 1.17503715e+00
3.19732606e-01 8.24265599e-01 1.14804542e+00 -5.27147532e-01
6.59978464e-02 5.15138268e-01 3.90171379e-01 6.21458054e-01
2.72086889e-01 -2.67175883e-02 -4.64072138e-01 -3.54005396e-01
3.97784203e-01 -4.53821331e-01 -3.49613219e-01 3.38725239e-01
-1.05429900e+00 1.20103395e+00 2.14571670e-01 1.89042121e-01
-7.57686347e-02 -2.26496354e-01 4.77705449e-02 7.32433975e-01
-1.84386313e-01 1.16101253e+00 -8.16842675e-01 -4.13567945e-02
-4.27499086e-01 2.12792262e-01 9.06773865e-01 7.73081303e-01
7.53232837e-01 -3.88532095e-02 -3.97100747e-01 6.07017636e-01
1.73478976e-01 7.07047045e-01 5.49867094e-01 -1.26517642e+00
4.68364388e-01 8.97387087e-01 4.26701784e-01 -8.19280982e-01
-8.03881437e-02 -2.59989560e-01 -4.58786845e-01 -2.87051588e-01
7.47097492e-01 -2.13747863e-02 -5.82330287e-01 1.82358694e+00
2.38881603e-01 -1.63320377e-01 4.03173655e-01 7.96428621e-01
9.40982997e-01 7.43783116e-01 -4.24825065e-02 -1.71495490e-02
1.67120636e+00 -9.48122561e-01 -7.07264841e-01 -4.11728561e-01
8.72047663e-01 -5.94588995e-01 1.82904828e+00 5.30221462e-01
-1.15003645e+00 -5.04450738e-01 -7.08666384e-01 -2.46748254e-01
-2.75465876e-01 2.79324502e-02 5.56604028e-01 7.97276020e-01
-1.09207380e+00 8.21132362e-02 -2.00634077e-01 -4.21984755e-02
2.36714199e-01 2.01476097e-01 -1.57970861e-01 -2.28254825e-01
-1.79419076e+00 1.11798942e+00 2.55084038e-01 1.15903869e-01
-9.85379875e-01 -3.82500410e-01 -6.82734072e-01 2.85876721e-01
6.73533916e-01 -6.90401077e-01 1.54067707e+00 -9.80050683e-01
-1.55150223e+00 8.77845407e-01 -3.79557163e-01 -4.55070764e-01
2.38650829e-01 -2.62829006e-01 -7.68191755e-01 5.00715494e-01
2.41081625e-01 6.94544613e-01 1.16246891e+00 -1.10909402e+00
-3.76142472e-01 -2.44314559e-02 7.21079886e-01 -1.67302266e-01
-2.78366119e-01 1.05024293e-01 3.35927218e-01 -2.26210549e-01
4.86302786e-02 -6.40310228e-01 3.28408897e-01 -2.62306184e-01
-5.36012828e-01 -7.30444670e-01 5.55884182e-01 -4.76607323e-01
1.22923732e+00 -1.97365952e+00 -4.10865664e-01 1.23136640e-02
3.82543117e-01 6.98682725e-01 -4.26321268e-01 3.40009093e-01
7.37641379e-02 4.00188774e-01 -8.58604759e-02 2.51385301e-01
2.07100749e-01 6.91300690e-01 -1.00423777e+00 -1.76254690e-01
7.20035672e-01 1.37494051e+00 -9.62815702e-01 -4.45544749e-01
-2.82525778e-01 -1.38837323e-01 -6.00872934e-01 5.62444806e-01
-6.50985301e-01 2.81137556e-01 -4.38153565e-01 4.71905321e-01
6.23197019e-01 -6.49877369e-01 -1.19237542e-01 9.36639309e-02
3.11999470e-01 8.23109567e-01 -7.08436191e-01 1.00285184e+00
-1.35479704e-01 4.62233812e-01 -4.44878265e-02 -9.40533042e-01
8.74045670e-01 3.38633835e-01 -5.84749222e-01 -6.80109978e-01
-1.06200263e-01 3.98302704e-01 2.27654457e-01 -7.67308474e-01
5.89963078e-01 -4.29997474e-01 -1.07509166e-01 5.85236192e-01
-6.36011511e-02 -3.58620048e-01 6.27310947e-04 6.27892494e-01
9.88271058e-01 -4.56343353e-01 2.10439697e-01 -1.03173099e-01
1.01864922e+00 1.87117234e-01 2.20871657e-01 1.14024901e+00
-3.14920485e-01 2.64403522e-01 5.54508090e-01 -3.56970638e-01
-5.16295433e-01 -1.33143568e+00 2.55938396e-02 1.34684134e+00
7.58130802e-03 -3.58143419e-01 -6.94783628e-01 -1.06018984e+00
-2.41800025e-01 1.33610308e+00 -3.23348731e-01 -5.79332411e-01
-5.16918778e-01 -2.13020504e-01 1.14876306e+00 3.06894481e-01
6.36179507e-01 -1.15795350e+00 -5.88202119e-01 -6.98049553e-03
-6.83547378e-01 -9.49408829e-01 -2.05572620e-01 -2.10213754e-02
-7.28489459e-01 -1.10542428e+00 -1.75772727e-01 -5.19878089e-01
6.87518299e-01 3.70299071e-01 1.41284227e+00 6.48062110e-01
4.66280758e-01 5.41237116e-01 -4.37340826e-01 -4.64168519e-01
-9.31838155e-01 -3.66055220e-02 1.48585485e-02 -1.95359915e-01
6.35291040e-01 -2.76686728e-01 -2.24720657e-01 6.65138066e-01
-1.13591373e+00 -1.43443808e-01 5.98717332e-01 9.86179769e-01
4.35409620e-02 -1.88763529e-01 1.12266576e+00 -9.45264101e-01
9.92536664e-01 -4.94973540e-01 -4.13019419e-01 7.79071450e-01
-4.34596032e-01 3.12879711e-01 8.77703786e-01 -6.08799160e-01
-1.15891755e+00 -8.35054517e-01 -2.91044593e-01 7.35037550e-02
-2.42962331e-01 4.27970111e-01 -2.60900676e-01 -1.08068585e-01
1.01676261e+00 3.78476620e-01 -6.89501837e-02 -1.10981919e-01
5.43864310e-01 7.06128061e-01 6.64498746e-01 -1.01468098e+00
9.44119990e-01 5.06739952e-02 -4.88090873e-01 -6.42865121e-01
-1.34404016e+00 -1.07991077e-01 -1.40199615e-02 9.65183154e-02
4.91158098e-01 -5.40839434e-01 -1.02207184e+00 9.65975970e-02
-1.47772920e+00 -2.20394418e-01 -3.34348679e-01 1.02204464e-01
-3.55734944e-01 6.33862197e-01 -7.15117157e-01 -9.06584203e-01
-1.96027637e-01 -8.41676712e-01 7.99699485e-01 4.48628545e-01
-4.70494330e-01 -8.73146296e-01 -2.05317631e-01 1.01008785e+00
5.91578960e-01 -5.44344068e-01 1.57608819e+00 -1.15311611e+00
-5.36944330e-01 1.92066710e-02 -4.75062206e-02 6.13170147e-01
-4.35522608e-02 -1.30964950e-01 -1.12094617e+00 -1.03853196e-01
4.91665125e-01 -9.17541802e-01 5.25050044e-01 -2.41522193e-01
1.11902690e+00 -7.87053168e-01 1.88994765e-01 8.11602026e-02
6.37320876e-01 -6.93060234e-02 9.08649504e-01 1.40782684e-01
1.14796840e-01 6.17896557e-01 6.51397943e-01 -5.58763705e-02
7.28233993e-01 1.78926572e-01 4.97110993e-01 3.56181175e-01
2.05180913e-01 -3.78276020e-01 6.97764218e-01 8.74519289e-01
3.31998110e-01 -3.99057686e-01 -9.19634163e-01 4.15065676e-01
-1.43583083e+00 -8.12224925e-01 -4.49271888e-01 1.97296774e+00
1.19445884e+00 1.71114951e-01 -5.55898309e-01 1.52766228e-01
5.64863026e-01 -6.61690161e-03 -4.90217000e-01 -8.85266662e-01
-4.66011614e-01 3.88001174e-01 -2.96220779e-01 6.74090087e-01
-4.39049155e-01 1.03057230e+00 6.52783966e+00 7.53277421e-01
-8.63037765e-01 9.41431075e-02 4.63066578e-01 4.25854653e-01
-8.00196648e-01 2.17313230e-01 -7.33224273e-01 2.71192670e-01
1.25808632e+00 -4.86153886e-02 4.68638390e-01 6.95538640e-01
-1.27991170e-01 -1.93797484e-01 -1.04816306e+00 4.96769965e-01
2.35298738e-01 -1.18643129e+00 6.95948184e-01 -3.03997993e-01
5.27249873e-01 -3.95295858e-01 2.23893628e-01 8.37584436e-01
2.97422141e-01 -1.39863098e+00 5.36611736e-01 4.23733979e-01
5.50563335e-01 -4.63533700e-01 9.13953066e-01 9.95340466e-01
-3.31803590e-01 -2.81559557e-01 -7.08113492e-01 -4.73280191e-01
-9.96022969e-02 2.85237581e-01 -1.06819379e+00 4.73996133e-01
4.40046191e-01 -3.74919772e-01 -9.35094595e-01 4.50609505e-01
-9.15507138e-01 9.93495941e-01 -1.04824938e-01 -1.86465010e-01
2.79519916e-01 4.26938236e-02 3.04508507e-01 5.88293433e-01
2.17959791e-01 4.56227809e-01 -4.57870156e-01 1.23254561e+00
-1.42674387e-01 -2.33589172e-01 -1.85609505e-01 -3.70241672e-01
5.54548383e-01 1.16203439e+00 1.89041886e-02 -5.98445475e-01
-3.48874807e-01 9.19212103e-01 5.12853384e-01 3.36062968e-01
-7.43528903e-01 -2.91807234e-01 3.26347917e-01 8.60115886e-02
-8.43709856e-02 -4.29209322e-02 -9.72799137e-02 -1.41367674e+00
6.49967492e-02 -1.53523636e+00 5.93825758e-01 -1.33221781e+00
-1.57652605e+00 5.16369462e-01 -1.94470197e-01 -6.07326210e-01
-4.06805515e-01 -7.56482184e-01 -7.51020908e-01 9.10088778e-01
-1.72278821e+00 -9.47736442e-01 -1.39930144e-01 8.18683088e-01
2.91251183e-01 2.68881768e-02 1.07403386e+00 -3.09895743e-02
-1.64398730e-01 8.83893251e-01 -5.36440909e-01 1.72162980e-01
9.32669163e-01 -1.11724472e+00 3.18115979e-01 7.39847541e-01
5.29746532e-01 1.21754766e+00 6.91199124e-01 -3.69873494e-01
-1.42479670e+00 -8.46373677e-01 1.49167800e+00 -1.10746276e+00
6.10912800e-01 -1.91299304e-01 -1.54762471e+00 7.77843475e-01
1.39251158e-01 -1.17419116e-01 8.23746741e-01 -1.02721654e-01
-6.87139988e-01 2.22373798e-01 -1.21882987e+00 5.91474593e-01
5.28212547e-01 -8.48079681e-01 -1.55279660e+00 4.36149746e-01
1.04553688e+00 -2.06852511e-01 -2.63152957e-01 2.36199260e-01
8.16501901e-02 -9.37100828e-01 8.68472397e-01 -1.29855263e+00
4.45972562e-01 -3.16715568e-01 -1.03933699e-01 -1.17052734e+00
-1.19706385e-01 -6.34753764e-01 -3.93888980e-01 1.12677765e+00
5.76111078e-01 -1.00914788e+00 5.14966130e-01 8.48979354e-01
-1.98070526e-01 -5.06361246e-01 -1.12592244e+00 -7.12594092e-01
4.07859743e-01 -5.83077073e-01 6.89915121e-01 1.03069305e+00
9.20481980e-02 5.21456063e-01 -2.41749197e-01 4.96911198e-01
3.83304618e-02 8.38167369e-02 7.88585246e-01 -1.07224572e+00
-3.85955453e-01 -1.20601907e-01 1.30953684e-01 -1.37868881e+00
3.02730113e-01 -8.29840839e-01 -2.38723233e-02 -1.14499748e+00
-6.83275983e-02 -2.19599858e-01 1.83981638e-02 5.74017346e-01
-6.01340234e-01 -1.14403442e-01 -2.45548375e-02 5.50276302e-02
-7.08354354e-01 4.56719220e-01 1.30094504e+00 -3.32286842e-02
1.01906933e-01 3.13437991e-02 -1.14306641e+00 7.81812012e-01
9.12363589e-01 -3.85702014e-01 -5.33550620e-01 -6.72057211e-01
6.51610255e-01 1.43033221e-01 7.18840003e-01 -4.82890785e-01
3.44140470e-01 -4.79807466e-01 3.44138853e-02 -3.16413283e-01
2.51539499e-01 -4.61497456e-01 -5.05420685e-01 4.44115698e-01
-5.05098462e-01 1.10307612e-01 1.28531098e-01 2.93897688e-01
-3.79885644e-01 -5.67770362e-01 4.82533216e-01 -2.94503152e-01
-5.03989697e-01 -2.09830746e-01 -4.53150302e-01 5.41989923e-01
5.45895934e-01 -7.85072818e-02 -1.09886837e+00 -8.19503486e-01
-3.65619332e-01 5.11306047e-01 6.24323860e-02 2.96275407e-01
8.57992589e-01 -1.07771218e+00 -7.73108542e-01 2.78682351e-01
2.84429789e-01 -3.40201333e-02 3.24572295e-01 8.03866863e-01
-2.24209219e-01 7.53763855e-01 3.31664495e-02 -4.34342682e-01
-8.64852905e-01 5.76862037e-01 4.05543149e-01 -1.45532325e-01
-5.50566167e-02 1.13592684e+00 3.67347836e-01 -7.20760286e-01
-1.60178825e-01 -3.62116426e-01 -1.66125298e-01 -2.73342997e-01
7.57341683e-01 5.63208498e-02 -7.49189183e-02 -3.24364662e-01
-3.77549767e-01 4.57848571e-02 -1.45158648e-01 5.64366952e-02
7.05432534e-01 -1.98177025e-01 -3.30063820e-01 2.37293527e-01
6.75737143e-01 3.37139457e-01 -7.02311873e-01 -2.81079978e-01
1.46843746e-01 -4.35901046e-01 -6.75903797e-01 -1.24342608e+00
-4.01457846e-01 1.33396864e+00 -1.21183760e-01 5.02772808e-01
9.05591011e-01 2.65444517e-01 9.77329433e-01 1.00091171e+00
4.68517005e-01 -7.88000882e-01 4.57700908e-01 8.41548264e-01
9.27626193e-01 -1.36604738e+00 -5.26467681e-01 -3.00083309e-01
-7.01071560e-01 1.17838788e+00 8.76801193e-01 2.47448489e-01
-2.79707104e-01 -1.97806522e-01 3.79422843e-01 -2.46804073e-01
-1.08902109e+00 2.32278809e-01 1.99171051e-01 4.93801773e-01
2.18336612e-01 1.11364454e-01 -1.47226006e-01 1.19899726e+00
-6.09191358e-01 -3.60389680e-01 7.94149935e-01 7.33683288e-01
-8.25153351e-01 -9.46842670e-01 -7.18748033e-01 3.90362352e-01
-4.61210698e-01 -4.03766811e-01 -7.30337560e-01 5.61777532e-01
1.10752471e-02 1.52956712e+00 -3.42990756e-01 -3.09451073e-01
1.71652302e-01 5.67706525e-01 2.65076905e-01 -7.39394903e-01
-7.57698715e-01 -9.14648354e-01 2.04563692e-01 -4.45517957e-01
-4.96335588e-02 -5.77094518e-02 -1.40320027e+00 -3.56581718e-01
-5.53767323e-01 6.76363170e-01 7.72592574e-02 1.35164690e+00
3.11661661e-01 9.77432076e-03 4.43086922e-01 3.44215125e-01
-1.34430528e+00 -8.69186759e-01 -1.34937242e-01 4.11248147e-01
3.85261059e-01 -3.34556401e-01 -7.74820805e-01 -2.75814861e-01]
|
[10.991774559020996, 7.9343485832214355]
|
904fda51-a8a8-4d3a-86c1-50f02012f8d8
|
reader-guided-passage-reranking-for-open
|
2101.00294
| null |
https://arxiv.org/abs/2101.00294v3
|
https://arxiv.org/pdf/2101.00294v3.pdf
|
Rider: Reader-Guided Passage Reranking for Open-Domain Question Answering
|
Current open-domain question answering systems often follow a Retriever-Reader architecture, where the retriever first retrieves relevant passages and the reader then reads the retrieved passages to form an answer. In this paper, we propose a simple and effective passage reranking method, named Reader-guIDEd Reranker (RIDER), which does not involve training and reranks the retrieved passages solely based on the top predictions of the reader before reranking. We show that RIDER, despite its simplicity, achieves 10 to 20 absolute gains in top-1 retrieval accuracy and 1 to 4 Exact Match (EM) gains without refining the retriever or reader. In addition, RIDER, without any training, outperforms state-of-the-art transformer-based supervised rerankers. Remarkably, RIDER achieves 48.3 EM on the Natural Questions dataset and 66.4 EM on the TriviaQA dataset when only 1,024 tokens (7.8 passages on average) are used as the reader input after passage reranking.
|
['Weizhu Chen', 'Jiawei Han', 'Jianfeng Gao', 'Yelong Shen', 'Xiaodong Liu', 'Pengcheng He', 'Yuning Mao']
|
2021-01-01
| null | null | null | null |
['triviaqa']
|
['miscellaneous']
|
[ 8.06396753e-02 8.59768223e-03 -5.84629104e-02 -2.21806262e-02
-1.82988548e+00 -8.97390246e-01 7.26969004e-01 4.91261452e-01
-9.24320161e-01 8.88741136e-01 6.50829077e-01 -3.68689150e-01
-3.58502954e-01 -7.69426227e-01 -7.96770334e-01 -8.32454041e-02
3.24006349e-01 1.18837047e+00 8.65428090e-01 -8.96836162e-01
6.00013316e-01 -1.02119029e-01 -1.41459167e+00 6.08974695e-01
1.31119132e+00 7.94359684e-01 2.31280744e-01 1.14815915e+00
-4.35726851e-01 1.22571707e+00 -6.57595575e-01 -6.49719119e-01
-1.02059118e-01 -4.58704889e-01 -1.60182595e+00 -8.55291545e-01
6.61556959e-01 -6.72631919e-01 -6.28302038e-01 4.81702268e-01
5.47960341e-01 5.40768743e-01 6.04686379e-01 -5.71417987e-01
-1.21844018e+00 8.23007166e-01 -1.08690701e-01 6.24259174e-01
9.72765088e-01 -3.99396002e-01 1.79584253e+00 -1.03766382e+00
6.21002793e-01 7.98206449e-01 3.23725790e-01 4.87444401e-01
-8.80792558e-01 -2.19161406e-01 -8.16640109e-02 6.14427090e-01
-1.22650146e+00 -3.32410425e-01 2.39156976e-01 -1.21552041e-02
1.08386159e+00 6.61688149e-01 1.32406220e-01 4.75665390e-01
-9.07965750e-02 1.00239253e+00 6.00151718e-01 -4.97128159e-01
-5.58698326e-02 -1.84438631e-01 6.69653833e-01 3.82577568e-01
1.48801534e-02 -2.33340457e-01 -5.91332853e-01 -4.59497839e-01
7.33622313e-02 -1.12297289e-01 -3.81885678e-01 2.45841876e-01
-1.26354635e+00 6.14787579e-01 5.98585367e-01 2.06416756e-01
-3.79973680e-01 -8.66460204e-02 3.25736701e-01 9.10090327e-01
4.41213489e-01 1.09853995e+00 -5.06712437e-01 -1.32315189e-01
-1.11544788e+00 6.51099741e-01 1.04370952e+00 8.14667463e-01
7.58174419e-01 -8.22554111e-01 -9.27604258e-01 1.10450339e+00
3.92144844e-02 6.91717625e-01 5.71767569e-01 -9.93949175e-01
4.97890115e-01 5.38196445e-01 5.73550582e-01 -6.66941941e-01
-1.14741743e-01 -4.66950655e-01 -3.52419287e-01 -7.49513447e-01
4.28379565e-01 1.64738595e-01 -8.01692367e-01 1.14238656e+00
1.70924723e-01 -2.02321142e-01 2.42918193e-01 9.27150071e-01
1.29000366e+00 9.33124483e-01 3.83522548e-02 1.75055996e-01
1.53189075e+00 -1.42309844e+00 -3.96599501e-01 -2.11441424e-02
7.92847514e-01 -1.10016644e+00 1.18233240e+00 8.52601305e-02
-1.30500245e+00 -3.93502474e-01 -7.07979739e-01 -5.84362149e-01
-2.84212261e-01 -1.28611565e-01 -1.85386866e-01 -7.65176341e-02
-1.33179259e+00 5.76975465e-01 -1.17639005e-01 -3.22055876e-01
-1.94337964e-01 2.38714404e-02 -6.48355410e-02 -3.68895769e-01
-1.74051237e+00 1.09489822e+00 -5.38955331e-02 -3.90587479e-01
-7.94582605e-01 -1.12985635e+00 -3.53656977e-01 3.20939183e-01
1.32255867e-01 -1.05669165e+00 1.89885902e+00 -4.06579673e-01
-1.35218728e+00 8.68289709e-01 -4.57481354e-01 -6.10087276e-01
3.35014731e-01 -7.90232003e-01 -4.27883863e-01 5.56939662e-01
3.01057339e-01 4.92870152e-01 3.97523493e-01 -8.88542414e-01
-8.93698215e-01 1.63243666e-01 4.59037215e-01 5.49672842e-01
-6.91207647e-02 1.27150699e-01 -5.00908732e-01 -4.04233038e-01
1.21508818e-02 -5.44502199e-01 4.60092165e-02 -3.59833568e-01
-1.99631318e-01 -8.21374595e-01 1.93208113e-01 -9.40819025e-01
1.52209330e+00 -1.68596971e+00 -1.00294389e-01 5.32589965e-02
4.83467609e-01 4.05274302e-01 -6.46039546e-01 1.01561892e+00
4.90976959e-01 -1.85396168e-02 2.19205767e-02 1.58896297e-01
1.89334840e-01 -1.14516743e-01 -8.63379121e-01 4.44920845e-02
-2.28121668e-01 1.24699914e+00 -1.30624020e+00 -5.07001996e-01
-3.17537993e-01 8.94745812e-02 -6.52746081e-01 4.52847749e-01
-5.34776926e-01 2.40684792e-01 -6.19138122e-01 1.68390781e-01
2.71881372e-01 -6.02643013e-01 -3.56869757e-01 2.75995750e-02
2.26490304e-01 1.14744353e+00 -4.10014749e-01 1.49142289e+00
-5.19407749e-01 6.71286523e-01 -3.98944199e-01 -4.13637966e-01
9.27391112e-01 3.68812501e-01 2.53844351e-01 -1.55078113e+00
-2.86417603e-01 5.91192901e-01 -4.06869024e-01 -3.95502120e-01
1.29863608e+00 1.41457707e-01 -5.73871583e-02 8.73920202e-01
-7.73377717e-02 4.86804359e-02 4.46072340e-01 8.74036729e-01
1.57587528e+00 -6.16885312e-02 -2.99883969e-02 -1.91325679e-01
7.65633941e-01 3.68347973e-01 -1.80893615e-02 1.28219461e+00
1.60976052e-01 6.34556472e-01 -2.88146995e-02 -2.34571725e-01
-1.14596128e+00 -1.22295880e+00 1.64419353e-01 1.84695458e+00
4.38318014e-01 -4.34166759e-01 -6.08533740e-01 -5.38394690e-01
1.44662067e-01 8.61853063e-01 -2.13853911e-01 -2.17024446e-01
-8.47140253e-01 -2.88415905e-02 7.94353545e-01 2.76483804e-01
3.68895859e-01 -1.06622267e+00 -1.52387336e-01 3.56053054e-01
-9.70656216e-01 -8.24292719e-01 -8.65415812e-01 -2.37221301e-01
-7.79290259e-01 -9.54594731e-01 -1.28604257e+00 -8.76469970e-01
4.10181522e-01 5.17934918e-01 1.78798342e+00 5.02643168e-01
3.64758641e-01 5.75036108e-01 -1.03682685e+00 1.51304707e-01
-3.27674448e-01 7.57257760e-01 -4.10469979e-01 -5.05626917e-01
4.58020240e-01 -2.38768503e-01 -9.59255159e-01 4.36876625e-01
-7.81304121e-01 -2.16101795e-01 6.04080498e-01 1.01080859e+00
5.46768188e-01 -6.93303823e-01 1.02979267e+00 -8.73990178e-01
9.83971059e-01 -6.13266647e-01 -3.29300433e-01 8.51091504e-01
-8.23644936e-01 4.16559786e-01 6.52829707e-01 -2.48848170e-01
-9.73681748e-01 -8.18724751e-01 -4.52172846e-01 1.78554818e-01
3.23691934e-01 6.49228156e-01 6.78671777e-01 1.13675043e-01
1.06247342e+00 4.72214609e-01 -4.38421071e-01 -8.48104179e-01
6.95481837e-01 9.28222358e-01 6.10256851e-01 -6.02877319e-01
9.37499404e-01 -3.35507691e-02 -6.32335067e-01 -4.00730014e-01
-1.38301504e+00 -1.11636233e+00 -2.27133632e-01 -1.21122837e-01
5.52176654e-01 -8.65677297e-01 -5.72351873e-01 1.74471691e-01
-1.23228478e+00 -2.44907364e-01 -5.67269623e-01 1.69756681e-01
-4.01176780e-01 3.52135956e-01 -9.29395437e-01 -3.45679909e-01
-1.08351576e+00 -4.07985389e-01 9.79178786e-01 2.94239014e-01
-3.72975260e-01 -7.26112127e-01 4.52207714e-01 9.40814614e-01
7.65914083e-01 -6.91481054e-01 1.04073906e+00 -1.22725976e+00
-7.20707536e-01 -2.60510296e-01 -3.06419432e-01 5.41771725e-02
-3.81514207e-02 -6.11517131e-01 -6.91252947e-01 -3.08397472e-01
-6.11176908e-01 -5.21298230e-01 1.01671386e+00 -3.13112259e-01
7.29047358e-01 -6.55294359e-01 -6.26991168e-02 -7.10627511e-02
1.02161849e+00 -1.07992873e-01 8.19306850e-01 5.92629075e-01
3.85736138e-01 3.93472254e-01 6.92508042e-01 2.31659397e-01
8.16083074e-01 3.32531363e-01 -8.34886543e-03 1.75906345e-01
-2.78904498e-01 -7.12126851e-01 1.99100897e-01 1.20468855e+00
1.97851688e-01 -4.69624490e-01 -7.25414634e-01 8.92394722e-01
-1.53112710e+00 -1.01426399e+00 -4.87509929e-02 2.26259041e+00
1.25495696e+00 -1.66238934e-01 2.84030791e-02 -1.10213198e-01
5.86108923e-01 1.67415440e-01 -5.14232397e-01 -3.18167448e-01
-3.12426314e-02 5.15003681e-01 4.20649260e-01 8.44969451e-01
-5.74540436e-01 1.08924735e+00 6.62008667e+00 8.99326205e-01
-4.07580525e-01 -3.52244005e-02 3.17894012e-01 1.34141380e-02
-7.54844189e-01 1.21967614e-01 -8.44249070e-01 2.18681455e-01
1.07046521e+00 -5.50065339e-01 6.47881985e-01 5.08933961e-01
-3.51941437e-01 -8.07622224e-02 -1.00616133e+00 6.09661162e-01
2.29005694e-01 -1.48114347e+00 6.02537513e-01 -6.75126851e-01
7.30460405e-01 2.91763961e-01 -1.82097793e-01 7.02088714e-01
5.65933228e-01 -9.28272903e-01 5.38549364e-01 8.84933472e-01
4.64638501e-01 -5.66008866e-01 8.20910037e-01 4.59830165e-01
-9.20768380e-01 1.35437310e-01 -4.63341326e-01 -4.00864072e-02
2.31720477e-01 2.66868204e-01 -8.36844981e-01 5.63186765e-01
7.80697346e-01 3.26955974e-01 -5.66653252e-01 1.41470909e+00
-2.46754125e-01 8.82957458e-01 -2.68070966e-01 -4.97768193e-01
3.43593091e-01 2.39946574e-01 6.97639108e-01 1.06226254e+00
1.56689882e-01 4.12020385e-01 -1.32632896e-01 3.88875276e-01
-5.85979581e-01 3.57555717e-01 1.10335648e-01 1.03093669e-01
8.05528402e-01 8.86195183e-01 -9.83922090e-03 -5.16305506e-01
-2.45366722e-01 1.21440971e+00 5.87717533e-01 4.43529010e-01
-4.45297807e-01 -9.04664040e-01 2.98904747e-01 2.20130250e-01
4.58513677e-01 2.34234825e-01 3.25241804e-01 -1.03488338e+00
2.85368592e-01 -1.10088694e+00 8.87087405e-01 -8.93667281e-01
-1.74570739e+00 8.03657174e-01 -2.50921488e-01 -1.10157764e+00
-3.16108197e-01 -2.07657263e-01 -3.86536002e-01 9.70647275e-01
-2.09741807e+00 -7.19289482e-01 -1.13755822e-01 4.85892951e-01
5.27987838e-01 1.12893336e-01 7.92694390e-01 5.76404214e-01
1.36199832e-01 9.31763113e-01 5.70458293e-01 2.97195107e-01
9.63098645e-01 -1.33988941e+00 5.32567561e-01 4.44307148e-01
2.89835095e-01 6.90525472e-01 6.93305254e-01 -3.07307780e-01
-1.30385530e+00 -8.37858975e-01 1.87764835e+00 -7.65939593e-01
8.34751725e-01 1.47233367e-01 -1.18747723e+00 3.48648697e-01
4.20718133e-01 -3.18745017e-01 6.16679192e-01 3.02594811e-01
-6.37693942e-01 -3.48505050e-01 -9.05403852e-01 7.07696140e-01
8.36942017e-01 -9.82128382e-01 -1.32855690e+00 4.82532203e-01
1.19749367e+00 -6.41005337e-01 -1.01955569e+00 2.89743751e-01
5.07205248e-01 -5.18051922e-01 1.11905408e+00 -7.78478146e-01
2.68989086e-01 -3.11705887e-01 -1.77501276e-01 -1.06816351e+00
-4.35480952e-01 -7.28548646e-01 -3.58475596e-01 9.95278120e-01
9.33000743e-01 -7.18560040e-01 4.91514742e-01 5.30910075e-01
-7.37792253e-02 -8.42558205e-01 -9.81512904e-01 -6.34656847e-01
5.74568033e-01 1.18675366e-01 8.86629045e-01 3.88410240e-01
2.04645753e-01 6.58172488e-01 -1.28350854e-01 -5.32357879e-02
2.21398845e-01 3.95685226e-01 5.76771796e-01 -1.09116518e+00
-2.21341580e-01 -3.68816853e-01 2.07019955e-01 -2.13572955e+00
-7.38848075e-02 -1.00162184e+00 2.93873668e-01 -2.04491663e+00
3.90833646e-01 -6.32945538e-01 -6.70614541e-01 1.69354931e-01
-6.03641927e-01 1.79729313e-01 4.64761667e-02 5.79495966e-01
-1.57788587e+00 7.48536587e-01 1.40550876e+00 -3.27094316e-01
-5.73206842e-02 5.75286942e-03 -1.05312324e+00 1.18681110e-01
5.12195706e-01 -5.50918996e-01 -3.15115988e-01 -8.59309673e-01
7.17498362e-01 2.78754711e-01 2.74200559e-01 -7.32271075e-01
7.43161917e-01 1.88778415e-01 -8.10736343e-02 -8.60190928e-01
-8.95871222e-03 -2.94708371e-01 -4.28834289e-01 1.08322397e-01
-1.05987036e+00 4.48789895e-01 -5.23403510e-02 4.66940969e-01
-4.36260223e-01 -4.77909058e-01 3.05252045e-01 1.39294490e-02
-6.59850061e-01 1.69918180e-01 -4.70948249e-01 1.01007831e+00
2.00494334e-01 2.63734490e-01 -8.79420161e-01 -7.04707682e-01
-2.43586481e-01 6.47452474e-01 1.72174305e-01 4.56018955e-01
6.44015193e-01 -1.14595544e+00 -1.25958622e+00 -4.47073668e-01
4.94538784e-01 -2.28987351e-01 4.01551396e-01 4.80904639e-01
-6.28244162e-01 9.37385261e-01 4.90049034e-01 -2.26168722e-01
-1.04858363e+00 2.41568103e-01 3.52752984e-01 -9.16292727e-01
-6.80105507e-01 9.59178686e-01 -2.73212999e-01 -7.49507785e-01
1.03426620e-01 -1.34049132e-01 -4.64090526e-01 9.50271115e-02
8.97883296e-01 6.72968745e-01 3.33789170e-01 -3.46808314e-01
-1.96102545e-01 4.38859046e-01 -6.29554629e-01 -2.53792226e-01
8.47115397e-01 -3.01148325e-01 -2.54288077e-01 8.42239931e-02
1.19656909e+00 4.17674109e-02 -5.46864390e-01 -9.19500291e-01
2.76202530e-01 -1.92660689e-01 -8.32310691e-02 -1.14723682e+00
-4.54355866e-01 3.56514513e-01 2.70516183e-02 7.71769583e-02
1.09177613e+00 2.56071270e-01 1.70659125e+00 1.17500865e+00
4.87281144e-01 -9.26741540e-01 2.02100843e-01 1.13148427e+00
9.24839318e-01 -9.39476490e-01 -1.57929838e-01 2.73573250e-01
-5.45414031e-01 6.26184285e-01 3.68575543e-01 -1.35375828e-01
3.17727804e-01 -4.81544286e-01 2.43939504e-01 -1.79271549e-01
-9.93865848e-01 -3.27908933e-01 8.54765356e-01 8.56398940e-02
4.53720212e-01 -1.99219197e-01 -4.16064918e-01 4.77062702e-01
-5.31036019e-01 -1.38775736e-01 1.42911494e-01 7.73678780e-01
-9.20167863e-01 -9.65964556e-01 -9.52927247e-02 8.29115212e-01
-5.80877721e-01 -6.53253615e-01 -2.61838138e-01 1.85474396e-01
-6.47234619e-01 1.37727559e+00 9.84124243e-02 -3.49509835e-01
3.81418854e-01 1.12691022e-01 2.64347762e-01 -4.54077929e-01
-1.01627576e+00 -7.66834319e-01 3.16844106e-01 -3.63740802e-01
-1.00040063e-01 -2.97817469e-01 -1.28760159e+00 -5.60445726e-01
-5.84682584e-01 1.10616279e+00 6.66086422e-03 1.05258095e+00
7.28020847e-01 2.42967278e-01 5.98288357e-01 8.46291333e-02
-8.67372692e-01 -1.00458515e+00 -2.38080397e-01 3.52241158e-01
7.52355158e-01 -1.35023281e-01 -4.36317384e-01 -1.96446687e-01]
|
[11.448826789855957, 7.760507583618164]
|
408f46b9-1e58-4d23-829c-8060f96c1ea3
|
pre-trained-contextual-embedding-of-source-1
|
2001.00059
| null |
https://arxiv.org/abs/2001.00059v3
|
https://arxiv.org/pdf/2001.00059v3.pdf
|
Learning and Evaluating Contextual Embedding of Source Code
|
Recent research has achieved impressive results on understanding and improving source code by building up on machine-learning techniques developed for natural languages. A significant advancement in natural-language understanding has come with the development of pre-trained contextual embeddings, such as BERT, which can be fine-tuned for downstream tasks with less labeled data and training budget, while achieving better accuracies. However, there is no attempt yet to obtain a high-quality contextual embedding of source code, and to evaluate it on multiple program-understanding tasks simultaneously; that is the gap that this paper aims to mitigate. Specifically, first, we curate a massive, deduplicated corpus of 7.4M Python files from GitHub, which we use to pre-train CuBERT, an open-sourced code-understanding BERT model; and, second, we create an open-sourced benchmark that comprises five classification tasks and one program-repair task, akin to code-understanding tasks proposed in the literature before. We fine-tune CuBERT on our benchmark tasks, and compare the resulting models to different variants of Word2Vec token embeddings, BiLSTM and Transformer models, as well as published state-of-the-art models, showing that CuBERT outperforms them all, even with shorter training, and with fewer labeled examples. Future work on source-code embedding can benefit from reusing our benchmark, and from comparing against CuBERT models as a strong baseline.
|
['Aditya Kanade', 'Petros Maniatis', 'Kensen Shi', 'Gogul Balakrishnan']
|
2019-12-21
| null |
https://proceedings.icml.cc/static/paper_files/icml/2020/5401-Paper.pdf
|
https://proceedings.icml.cc/static/paper_files/icml/2020/5401-Paper.pdf
|
icml-2020-1
|
['contextual-embedding-for-source-code', 'program-repair', 'variable-misuse', 'exception-type', 'swapped-operands', 'function-docstring-mismatch', 'wrong-binary-operator', 'program-repair']
|
['computer-code', 'computer-code', 'computer-code', 'computer-code', 'computer-code', 'computer-code', 'computer-code', 'reasoning']
|
[ 1.09479934e-01 4.53702062e-02 -3.93339515e-01 -3.51715982e-01
-1.10107112e+00 -6.13729239e-01 4.23197687e-01 3.58751714e-01
-3.00429940e-01 2.67843217e-01 6.42394900e-01 -8.41663957e-01
1.08380541e-01 -5.76300979e-01 -8.09584856e-01 -1.33678228e-01
-1.76816970e-01 1.29825482e-02 2.32713446e-01 -3.29709679e-01
5.22786081e-01 -2.00546250e-01 -1.40305662e+00 4.79290783e-01
9.84283686e-01 4.78916675e-01 2.95130014e-01 9.63838875e-01
-1.47995472e-01 9.59596097e-01 -3.38351220e-01 -5.84438384e-01
-2.08985880e-01 -1.03719331e-01 -1.17698383e+00 -4.39953715e-01
2.24953458e-01 -1.82522327e-01 -1.29133269e-01 6.21901155e-01
2.55941778e-01 -2.79826611e-01 3.97623479e-01 -9.24233496e-01
-1.35856771e+00 8.05765271e-01 -3.96157503e-01 2.88244545e-01
2.75950819e-01 2.17418283e-01 1.29284072e+00 -9.17498469e-01
5.03549755e-01 9.61168826e-01 1.11354184e+00 7.36247122e-01
-1.37634051e+00 -3.01830173e-01 -9.09274891e-02 2.05104068e-01
-9.75668728e-01 -2.42620587e-01 3.94524634e-01 -5.79890370e-01
1.71772325e+00 -5.98568358e-02 1.82053804e-01 1.27128649e+00
2.15546727e-01 7.61326849e-01 7.37711489e-01 -6.77916527e-01
-2.63767272e-01 2.55443245e-01 3.41518581e-01 7.97483802e-01
-2.08222792e-02 1.53460532e-01 -5.44290915e-02 -2.78187305e-01
2.46466592e-01 -6.67718705e-03 -3.38634044e-01 -5.90686083e-01
-1.35888076e+00 1.04492438e+00 5.33168793e-01 6.78420722e-01
7.12909549e-02 5.50269723e-01 9.58016753e-01 5.58500946e-01
6.42603457e-01 7.19289362e-01 -9.45194960e-01 -9.26348031e-01
-8.95943582e-01 7.63217956e-02 9.56616938e-01 1.05503058e+00
8.37075889e-01 3.54119577e-03 -6.51447251e-02 8.92666936e-01
1.54817045e-01 1.81161925e-01 7.50945687e-01 -6.64932489e-01
7.37964511e-01 7.67381847e-01 -2.75206268e-01 -6.93932712e-01
-1.33441091e-01 -2.63667971e-01 -2.79567748e-01 1.26146957e-01
2.74445117e-01 -2.24119108e-02 -7.07325876e-01 1.80066764e+00
-3.33014965e-01 1.04311377e-01 1.25843108e-01 4.07487899e-01
6.47021651e-01 6.73857808e-01 1.28146857e-01 5.62370896e-01
1.22131908e+00 -1.31975269e+00 -2.54427195e-01 -4.13166404e-01
1.42518282e+00 -7.94172525e-01 1.46505427e+00 3.29923987e-01
-7.27366924e-01 -6.89668119e-01 -1.14696920e+00 -4.49262321e-01
-7.06132114e-01 2.10928798e-01 8.99457335e-01 8.37853909e-01
-1.37431669e+00 7.09629536e-01 -9.51998353e-01 -5.26126385e-01
4.36607033e-01 9.08922851e-02 -5.20219684e-01 -3.25682431e-01
-8.58021438e-01 1.12278152e+00 2.76035905e-01 -4.01375502e-01
-1.19597983e+00 -1.33247530e+00 -1.23198414e+00 2.47259051e-01
1.66231245e-01 -5.33541024e-01 1.35011625e+00 -8.68604183e-01
-1.12983263e+00 1.00119042e+00 -1.30498156e-01 -4.92586821e-01
-1.01578467e-01 -3.30528289e-01 -2.53948718e-01 -4.11689103e-01
5.05568795e-02 4.41969663e-01 5.25242746e-01 -1.17666519e+00
-3.09328973e-01 -1.18438244e-01 4.93992090e-01 -3.41272831e-01
-6.67971611e-01 1.75078526e-01 -3.21401775e-01 -5.24006903e-01
-6.94721222e-01 -7.40184486e-01 -1.51570663e-01 -2.76443124e-01
-1.65416643e-01 -3.45996618e-01 8.11247110e-01 -9.28820610e-01
1.39153671e+00 -2.39797139e+00 3.56987894e-01 -4.54773307e-01
3.16249132e-01 5.00562131e-01 -5.72923660e-01 7.16137350e-01
-3.60258013e-01 5.70186615e-01 -5.56689441e-01 -5.41426122e-01
2.11782739e-01 2.48758987e-01 -5.43033540e-01 1.82611763e-01
5.97815514e-01 1.13483465e+00 -1.00150585e+00 -2.86524862e-01
1.58514827e-01 4.50007260e-01 -1.08778274e+00 2.19195202e-01
-3.88604552e-01 -9.10371169e-02 -3.94138575e-01 4.86560374e-01
2.95979559e-01 -3.23456645e-01 -1.39030799e-01 1.70437142e-01
-2.33174875e-01 6.27816379e-01 -4.12432224e-01 2.26037216e+00
-1.33677793e+00 8.64512920e-01 -2.47377902e-01 -1.22062373e+00
7.26945102e-01 3.48952681e-01 5.78707866e-02 -5.32421708e-01
-1.19390436e-01 3.94980550e-01 -6.38906239e-03 -8.73882413e-01
5.53547919e-01 1.81578040e-01 -4.65590686e-01 7.14063466e-01
4.65280950e-01 -2.45507553e-01 1.27041757e-01 4.11762565e-01
1.60099018e+00 4.06669766e-01 2.20400840e-01 -3.21426272e-01
5.41107714e-01 2.19263613e-01 4.02669385e-02 5.59346318e-01
-7.91233853e-02 5.27638614e-01 7.53332794e-01 -3.47704887e-01
-1.25681722e+00 -7.45251417e-01 -1.93319723e-01 1.49208105e+00
-3.55598927e-01 -7.12787151e-01 -8.28609526e-01 -9.76248264e-01
1.19255841e-01 8.42928767e-01 -9.96952355e-01 -2.39257708e-01
-6.89482987e-01 -6.62233472e-01 8.78590047e-01 8.94856632e-01
7.95626938e-02 -9.34236705e-01 -3.89523953e-01 2.39176571e-01
2.34748106e-02 -7.28919208e-01 -4.81363118e-01 5.22703528e-01
-7.96148121e-01 -1.17625523e+00 -4.24411833e-01 -8.14581394e-01
3.37222785e-01 2.17402354e-01 1.77612686e+00 7.15951502e-01
-3.81031632e-01 5.87959170e-01 -6.93651378e-01 -2.52354443e-01
-7.75050521e-01 5.10592997e-01 -4.91200089e-01 -8.30546737e-01
7.21236289e-01 -6.64140344e-01 -2.44313255e-01 4.02014814e-02
-1.03290081e+00 -1.38392538e-01 6.29929543e-01 1.16596806e+00
-1.02210052e-01 -5.03527880e-01 7.74278283e-01 -1.00992560e+00
6.87507868e-01 -8.03447843e-01 -4.28999692e-01 2.55570561e-01
-6.15152359e-01 3.54089677e-01 9.28018093e-01 -1.19182311e-01
-1.04458964e+00 -4.09242511e-01 -6.56950951e-01 -7.00753033e-02
-1.44308686e-01 8.52859020e-01 2.15774521e-01 -2.67207414e-01
9.41454053e-01 1.64653838e-01 -1.55727223e-01 -6.63067877e-01
8.89325082e-01 8.66053283e-01 4.41700011e-01 -9.36911404e-01
9.62946832e-01 1.35417879e-01 -7.43129611e-01 -4.72432524e-01
-7.47055352e-01 -5.50156176e-01 -6.71285689e-01 5.69956958e-01
9.25595641e-01 -9.51840103e-01 -2.68602848e-01 1.05475731e-01
-1.30480182e+00 -6.77159727e-01 -1.79910824e-01 1.35629162e-01
-6.26646161e-01 4.55032349e-01 -8.04639697e-01 -2.36199215e-01
-1.73606366e-01 -1.40569413e+00 1.17333579e+00 -1.98987395e-01
-3.39628696e-01 -1.40660882e+00 3.21492553e-01 2.95374602e-01
9.41364765e-01 9.65328589e-02 1.51467860e+00 -6.56630635e-01
-6.83820903e-01 -1.97983891e-01 -3.99205923e-01 7.65240431e-01
-4.81663868e-02 7.55508170e-02 -1.17626572e+00 -2.98360884e-01
-3.37895080e-02 -8.00895214e-01 1.05484951e+00 -1.82206836e-02
1.28830874e+00 5.89844286e-02 -4.87919569e-01 7.42973804e-01
1.64148438e+00 -9.82936099e-02 7.09924996e-01 4.37468469e-01
7.22003162e-01 3.24380070e-01 3.77991050e-01 1.97403282e-01
6.54649258e-01 5.29973328e-01 5.70415020e-01 -6.52962551e-02
-2.49739841e-01 -3.65325779e-01 4.41510737e-01 1.23765600e+00
1.56990111e-01 -5.47988489e-02 -1.14167440e+00 1.09832263e+00
-1.57070160e+00 -6.15160465e-01 -2.10987806e-01 1.99894369e+00
1.05882144e+00 -1.08706415e-01 -1.95560396e-01 -1.14909708e-02
5.47798164e-02 1.91967532e-01 -3.42008978e-01 -8.04302335e-01
3.92879456e-01 5.54833949e-01 2.59444833e-01 4.38835442e-01
-1.02821577e+00 9.80398953e-01 6.24955654e+00 8.55716646e-01
-9.62595344e-01 5.50858617e-01 3.18840235e-01 3.68628263e-01
-5.31547368e-01 2.81504780e-01 -6.13979816e-01 2.43443564e-01
1.41243434e+00 -3.80470544e-01 4.58566815e-01 1.33629620e+00
-1.73864767e-01 2.57668924e-02 -1.68542969e+00 5.69215655e-01
1.60194263e-01 -1.29642880e+00 -3.79143447e-01 -1.72291175e-01
8.09291184e-01 4.66822118e-01 2.22267546e-02 1.13796854e+00
6.11523211e-01 -1.22817302e+00 3.91582370e-01 -7.76259303e-02
7.93074131e-01 -4.47213382e-01 9.53996778e-01 2.11922303e-01
-1.20335722e+00 -2.50466019e-01 -4.58759636e-01 -1.82179332e-01
-1.65662244e-01 5.13706505e-01 -6.73540235e-01 6.22523248e-01
7.00076997e-01 1.18308282e+00 -1.01372170e+00 8.54353309e-01
-3.38412791e-01 7.19256520e-01 2.44541198e-01 -7.10410178e-02
3.24314088e-01 2.60014236e-01 1.20520309e-01 1.57282984e+00
3.29384863e-01 -4.59540427e-01 -4.82184254e-02 1.38448787e+00
-3.38896364e-01 -8.78560096e-02 -7.93545306e-01 -3.93439472e-01
2.23207518e-01 1.25756288e+00 -1.02159254e-01 -4.50685531e-01
-1.05477166e+00 8.51468742e-01 6.66497588e-01 2.28531614e-01
-1.03927529e+00 -6.82249069e-01 7.92268455e-01 -2.44112179e-01
4.17208254e-01 -3.52306336e-01 -3.07007968e-01 -1.34684420e+00
5.21665625e-02 -1.16943347e+00 2.11167082e-01 -6.81001008e-01
-1.32142484e+00 6.67488158e-01 -9.16133076e-03 -9.48800862e-01
-3.68260354e-01 -7.91886747e-01 -7.69969344e-01 1.06125152e+00
-1.77332258e+00 -1.17898834e+00 -1.56378865e-01 2.09835112e-01
6.53329790e-01 -1.05512977e-01 1.19466388e+00 5.58294177e-01
-4.09391254e-01 7.31373131e-01 2.16418400e-01 2.23366305e-01
7.32634962e-01 -1.54027259e+00 8.92617047e-01 8.35126460e-01
3.50044966e-02 9.78130877e-01 5.37834644e-01 -1.80205613e-01
-1.57665217e+00 -1.20581543e+00 8.83429825e-01 -1.16004920e+00
1.16500545e+00 -6.59998953e-01 -1.20615149e+00 1.05847931e+00
4.86430526e-01 4.92703281e-02 7.15792239e-01 5.54644942e-01
-8.92856598e-01 3.55879962e-01 -8.09645951e-01 2.96192616e-01
1.01290405e+00 -9.03776586e-01 -8.96782935e-01 3.95536423e-01
1.19710505e+00 -3.31170857e-01 -1.26632202e+00 1.89764827e-01
2.77140290e-01 -9.90991533e-01 1.04058969e+00 -6.87821448e-01
1.22003317e+00 6.29105046e-02 -1.36492044e-01 -1.71060276e+00
-1.38895571e-01 -1.55797914e-01 1.00106351e-01 1.39630270e+00
6.51399016e-01 -5.72948992e-01 4.63290244e-01 3.76998961e-01
-6.46746695e-01 -9.00659382e-01 -6.20271683e-01 -8.08754981e-01
7.17132747e-01 -6.88799500e-01 5.22615135e-01 9.52097416e-01
4.05648112e-01 2.68962085e-01 -1.37632430e-01 -1.79794937e-01
2.02455580e-01 1.90428242e-01 9.88896847e-01 -7.88419425e-01
-6.25101030e-01 -5.03614843e-01 -3.22201729e-01 -1.11309004e+00
5.14902234e-01 -1.20226288e+00 1.08350925e-01 -1.46474981e+00
5.00918388e-01 -4.88609105e-01 -4.34923321e-02 8.42959642e-01
-3.62687320e-01 1.06255323e-01 2.04055123e-02 -4.99900021e-02
-4.37144399e-01 6.55565500e-01 7.77344227e-01 -2.71548837e-01
1.48473039e-01 -4.52232689e-01 -9.04890180e-01 5.00124216e-01
5.26349306e-01 -6.57157302e-01 -3.09260249e-01 -9.24226522e-01
2.06637114e-01 -7.58660510e-02 3.56090099e-01 -8.80397201e-01
-2.32784823e-01 4.24964517e-01 -1.75695062e-01 -1.10286973e-01
-9.41615459e-03 -5.62573314e-01 -4.48426127e-01 4.25971389e-01
-4.22325224e-01 1.25025183e-01 6.43962979e-01 4.19254601e-01
-3.78858149e-01 -7.45788991e-01 5.91807008e-01 -1.66723967e-01
-9.97463048e-01 -5.69684505e-02 -2.62902200e-01 3.51202786e-01
7.87894547e-01 2.55969632e-03 -6.75277948e-01 -5.93455471e-02
-2.34681964e-01 9.64628607e-02 7.63134599e-01 9.09564972e-01
3.39210838e-01 -1.08216846e+00 -6.47751033e-01 9.43409801e-02
6.25935495e-01 -3.11786085e-01 1.56671226e-01 7.48822987e-01
-3.70555371e-01 6.71860337e-01 4.69079092e-02 -5.00811219e-01
-1.04491079e+00 8.93437564e-01 1.29197612e-01 -6.92075193e-01
-4.25518721e-01 8.25006306e-01 1.99776337e-01 -9.81938243e-01
-1.10337622e-01 -6.78117812e-01 2.64129248e-02 -3.59009445e-01
6.29632473e-01 -9.94665325e-02 2.73322791e-01 -1.16032355e-01
-2.84012198e-01 5.40805876e-01 -8.61755013e-02 3.97392273e-01
1.65855813e+00 2.47826621e-01 -3.05670381e-01 3.79305780e-01
1.75458479e+00 -2.84870830e-03 -8.71092618e-01 -3.00905198e-01
3.62739146e-01 -5.38119733e-01 -4.61195633e-02 -7.92153835e-01
-8.65801394e-01 1.28195632e+00 4.20374840e-01 2.85134703e-01
9.21171844e-01 3.69169638e-02 6.90867960e-01 3.30557883e-01
5.69896698e-01 -5.03983498e-01 3.03765774e-01 8.18885744e-01
7.51627505e-01 -1.40157461e+00 -3.45020950e-01 -4.45249043e-02
-3.33477318e-01 1.25776768e+00 6.52397513e-01 -2.20252305e-01
3.97597939e-01 4.97843355e-01 -1.33940488e-01 -2.32959718e-01
-9.60416377e-01 -6.78557828e-02 2.64292601e-02 6.82957113e-01
1.08758175e+00 -1.93228111e-01 -2.04040688e-02 6.13001287e-01
-2.29870081e-02 2.81567037e-01 7.12348759e-01 9.36694324e-01
-2.33221695e-01 -1.52828217e+00 -1.46318823e-01 5.80748022e-01
-4.61974740e-01 -6.74589694e-01 4.79523018e-02 9.49107707e-01
-7.09893852e-02 8.49518001e-01 -2.34117553e-01 -5.39172113e-01
3.93886924e-01 1.54267699e-01 3.64358038e-01 -1.14197516e+00
-8.05854917e-01 -7.44902372e-01 7.52291679e-02 -6.04875028e-01
-2.06941754e-01 -3.43158394e-01 -9.07894909e-01 -3.24299693e-01
-2.50779718e-01 1.69819295e-01 6.89703703e-01 8.32445383e-01
5.13877392e-01 8.81007135e-01 2.76926398e-01 -9.84157145e-01
-7.30310917e-01 -1.15804613e+00 5.29612638e-02 5.52201807e-01
3.97439837e-01 -3.94567490e-01 -5.43486416e-01 1.82577834e-01]
|
[7.612033843994141, 7.865365982055664]
|
f4955f9b-410c-4507-9571-d6fe1d0c3355
|
towards-scene-understanding-for-autonomous
| null | null |
https://openaccess.thecvf.com/content/ACCV2022W/MLCSA/html/Steininger_Towards_Scene_Understanding_for_Autonomous_Operations_on_Airport_Aprons_ACCVW_2022_paper.html
|
https://openaccess.thecvf.com/content/ACCV2022W/MLCSA/papers/Steininger_Towards_Scene_Understanding_for_Autonomous_Operations_on_Airport_Aprons_ACCVW_2022_paper.pdf
|
Towards Scene Understanding for Autonomous Operations on Airport Aprons
|
Enhancing logistics vehicles on airport aprons with assistant and autonomous capabilities offers the potential to significantly increase
safety and efficiency of operations. However, this research area is still underrepresented compared to other automotive domains, especially regarding available image data, which is essential for training and benchmarking AI-based approaches. To mitigate this gap, we introduce a novel dataset specialized on static and dynamic objects commonly encountered while navigating apron areas. We propose an efficient approach for image acquisition as well as annotation of object instances and environmental
parameters. Furthermore, we derive multiple dataset variants on which we conduct baseline classification and detection experiments. The resulting models are evaluated with respect to their overall performance and robustness against specific environmental conditions. The results are quite promising for future applications and provide essential insights regarding the selection of aggregation strategies as well as current potentials and limitations of similar approaches in this research domain.
|
['Oliver Zendel', 'Julia Simon', 'Verena Widhalm', 'Wolfgang Pointner', 'Andreas Kriegler', 'Daniel Steininger']
|
2022-12-04
| null | null | null |
asian-conference-on-computer-vision-accv
|
['fine-grained-image-classification']
|
['computer-vision']
|
[ 1.93923950e-01 -2.79631585e-01 4.05414179e-02 -5.06528437e-01
-2.30379969e-01 -7.98102319e-01 7.90232539e-01 4.15422350e-01
-6.38330162e-01 5.91890097e-01 -4.34677839e-01 -2.75886744e-01
-4.25838441e-01 -9.07984138e-01 -6.30340457e-01 -8.00704241e-01
-3.69677752e-01 6.29739523e-01 3.68504286e-01 -7.34990656e-01
2.93038309e-01 1.09285414e+00 -1.95764124e+00 2.69804627e-01
6.54794812e-01 1.17130637e+00 2.77797103e-01 5.02512276e-01
2.42849365e-01 4.42388713e-01 -8.32810581e-01 -3.17927450e-01
6.52520537e-01 2.77464986e-02 -2.67278939e-01 3.88983965e-01
5.20414293e-01 -2.09809411e-02 -2.27228105e-01 8.01150680e-01
1.83454171e-01 6.08219862e-01 5.67394674e-01 -1.48624277e+00
-5.13804257e-02 -6.69140294e-02 -3.52702220e-03 2.11913392e-01
-1.04810745e-02 5.13802171e-01 6.47725224e-01 -6.80381119e-01
7.22470582e-01 8.55209410e-01 3.92157435e-01 1.80608973e-01
-8.28229785e-01 -4.48579073e-01 6.01335406e-01 6.39094830e-01
-1.21187377e+00 -5.03548086e-01 6.69978380e-01 -1.99428320e-01
8.00457716e-01 3.41732204e-01 8.19698572e-01 9.17136371e-01
3.09848458e-01 8.16668212e-01 9.18074965e-01 -1.95538878e-01
3.37843418e-01 5.05676866e-01 -9.90165025e-02 7.49912620e-01
6.32272124e-01 1.91666231e-01 -3.18619728e-01 2.72133350e-01
2.40838349e-01 -2.54293799e-01 1.09083675e-01 -7.99581051e-01
-1.15292621e+00 6.12000585e-01 4.25591052e-01 3.59027982e-02
-5.64636171e-01 -8.32325444e-02 4.22912329e-01 2.29535013e-01
2.14699194e-01 7.99728751e-01 -3.66094559e-01 -1.05464505e-02
-5.34311771e-01 6.26128256e-01 4.82310236e-01 1.25221479e+00
5.74006200e-01 1.71069294e-01 1.04607768e-01 5.69886386e-01
-1.68395955e-02 5.34144163e-01 -3.71855676e-01 -9.18036044e-01
5.65223277e-01 7.50675738e-01 4.47582483e-01 -1.19678831e+00
-6.36877954e-01 -3.49712849e-01 -2.51272738e-01 3.28301579e-01
3.00473034e-01 -1.29923418e-01 -7.99106181e-01 1.03781998e+00
2.70176083e-01 -1.29973963e-01 1.37746289e-01 9.17705417e-01
5.40944397e-01 5.44959843e-01 3.76012437e-02 -5.60336299e-02
1.32407665e+00 -9.44837332e-01 -7.27532089e-01 -4.65340495e-01
5.47514737e-01 -6.86227918e-01 8.12747538e-01 6.16550386e-01
-9.37090755e-01 -6.24087214e-01 -9.94649410e-01 4.69487160e-01
-8.38685513e-01 2.36975640e-01 7.03692138e-01 7.84896791e-01
-7.21407235e-01 1.51178569e-01 -7.13664412e-01 -4.61730033e-01
2.49972701e-01 4.37226266e-01 -4.03890789e-01 -5.07550649e-02
-1.11214304e+00 1.25108147e+00 5.49542189e-01 5.21582723e-01
-9.02720332e-01 -5.55520236e-01 -1.01996720e+00 -4.03794318e-01
7.75658846e-01 -1.39904469e-01 9.54814196e-01 -5.00997543e-01
-1.13549268e+00 5.59449375e-01 2.17668951e-01 -7.89367318e-01
4.97282743e-01 -2.48417988e-01 -6.82850063e-01 4.55751456e-02
-1.39921457e-01 6.71028852e-01 4.62746412e-01 -1.62453365e+00
-1.14354753e+00 -2.40840092e-01 4.37460124e-01 2.67705828e-01
-2.77685523e-01 -6.78576306e-02 -4.26514834e-01 -2.71455735e-01
-2.41654828e-01 -1.11949897e+00 -5.08685827e-01 -1.28987739e-02
-4.99905758e-02 1.90030932e-01 9.52355921e-01 -4.43387836e-01
1.02564657e+00 -1.93914950e+00 -5.10187028e-03 4.64692593e-01
-3.24940950e-01 4.40750122e-01 -2.12715641e-01 5.76147139e-01
3.08975190e-01 -2.93192148e-01 -1.41085787e-02 7.82760680e-02
-1.50433630e-01 2.83865958e-01 -2.02256635e-01 4.61398751e-01
4.58667248e-01 5.81718326e-01 -8.72501373e-01 -2.07807153e-01
7.83351243e-01 1.65986359e-01 -2.04771787e-01 -2.33298559e-02
-2.64050096e-01 4.89767343e-01 -5.11739552e-01 1.15032089e+00
6.20262742e-01 6.20368838e-01 1.74480423e-01 -4.67835903e-01
-3.31257701e-01 -9.14443508e-02 -1.20259893e+00 1.13472426e+00
-5.56205690e-01 7.94443250e-01 1.53450146e-01 -1.13052809e+00
1.13685548e+00 -8.45304355e-02 5.96153021e-01 -8.38989794e-01
3.78721535e-01 6.75751790e-02 6.76555559e-02 -8.88685107e-01
1.15582240e+00 1.15957998e-01 -4.43334669e-01 -2.85808444e-01
-3.00601840e-01 -2.65506774e-01 6.52015984e-01 -1.36338949e-01
9.99831259e-01 1.71347439e-01 1.28953144e-01 -2.31960580e-01
6.81896925e-01 7.39657462e-01 3.63461912e-01 5.00450075e-01
-4.34021622e-01 8.52874368e-02 2.44939923e-01 -7.39992976e-01
-9.75490987e-01 -9.67844009e-01 -1.76341102e-01 8.73018682e-01
8.37972999e-01 -1.40663832e-01 -5.60227334e-01 -7.82516539e-01
2.17329457e-01 1.03255808e+00 -4.64532405e-01 -2.14877039e-01
-7.37478495e-01 -7.69088984e-01 3.75432342e-01 7.81438291e-01
4.56258148e-01 -9.72868741e-01 -9.74469662e-01 9.33423191e-02
-1.68268487e-01 -1.54741502e+00 2.76261657e-01 1.59447208e-01
-5.90958238e-01 -1.24305880e+00 -4.82193418e-02 -4.48050350e-01
7.73262680e-01 5.66543162e-01 9.95721936e-01 1.10314295e-01
-4.24817979e-01 5.29548764e-01 -4.59680468e-01 -9.91437078e-01
-3.59744638e-01 1.15131870e-01 2.40718827e-01 8.85733441e-02
4.29461122e-01 2.57520944e-01 -4.91657943e-01 8.51188004e-01
-6.29089415e-01 -3.96200567e-01 7.59376764e-01 5.73175907e-01
4.38418567e-01 3.93391222e-01 3.30522269e-01 -6.54821575e-01
5.23052275e-01 -3.91329497e-01 -1.08344007e+00 2.47965336e-01
-6.03946865e-01 -3.61190259e-01 6.31335735e-01 -1.26261991e-02
-1.11717200e+00 1.03310391e-01 1.49343461e-01 -2.20098883e-01
-7.08381534e-01 3.46445709e-01 -3.55893403e-01 -3.70539814e-01
5.43601513e-01 -7.52851740e-02 1.41736746e-01 -2.29114354e-01
5.98252043e-02 4.99240905e-01 4.82800215e-01 -5.53022683e-01
8.54864120e-01 6.83851361e-01 3.71750414e-01 -1.00966215e+00
-4.93512511e-01 -6.36709630e-01 -6.74133837e-01 -6.72016501e-01
6.45050645e-01 -5.64066231e-01 -5.60498595e-01 1.37045205e-01
-7.73461044e-01 -1.63043037e-01 -1.26741946e-01 5.01935184e-01
-4.69555587e-01 8.09923112e-02 -4.21326123e-02 -9.88164902e-01
3.91498685e-01 -1.43554723e+00 9.99968171e-01 -3.06424238e-02
7.79723236e-03 -9.33791876e-01 -3.89046729e-01 6.56206787e-01
3.01081359e-01 3.79813582e-01 7.35050321e-01 -6.33256793e-01
-8.73003602e-01 -6.68159425e-01 9.19868425e-02 3.30200285e-01
-7.25714490e-02 7.47985840e-02 -7.08835006e-01 -4.04444993e-01
-4.02858257e-01 -1.45670632e-02 7.48715878e-01 1.10505380e-01
9.22621369e-01 1.98090315e-01 -6.22191489e-01 8.45835134e-02
1.28296602e+00 5.57760537e-01 5.95174432e-01 9.26171660e-01
1.51662365e-01 1.20244837e+00 1.81267333e+00 4.70180273e-01
1.79488108e-01 9.21567678e-01 9.63497937e-01 -4.03493866e-02
1.77466214e-01 2.63709009e-01 3.42927307e-01 1.34651124e-01
-4.63230342e-01 -6.59496605e-01 -1.04405904e+00 7.68840432e-01
-1.65382862e+00 -9.47173178e-01 -1.57572672e-01 2.14385676e+00
-1.32647321e-01 2.49575630e-01 1.94070071e-01 1.88157618e-01
5.63946724e-01 7.62082189e-02 -3.26144964e-01 -2.82864213e-01
-2.70765983e-02 -1.78117588e-01 8.80959511e-01 1.55949160e-01
-1.38029242e+00 8.93800974e-01 6.56432247e+00 4.31237459e-01
-8.36050987e-01 -2.67113805e-01 2.78399825e-01 -1.61149591e-01
1.49448991e-01 -1.83512032e-01 -1.08869708e+00 1.96281493e-01
1.06891382e+00 2.42761463e-01 3.01159769e-01 9.09681976e-01
2.39943832e-01 -3.66321623e-01 -9.07917202e-01 4.52411622e-01
2.27705166e-01 -1.28194857e+00 -8.53482187e-02 1.13031045e-01
5.43833077e-01 -2.76658654e-01 1.32801116e-01 2.63101041e-01
-1.27943486e-01 -6.91385746e-01 9.13965344e-01 4.21802044e-01
2.24262133e-01 -1.16570008e+00 1.14679921e+00 7.26066753e-02
-1.29266965e+00 -5.13182104e-01 -4.38992918e-01 5.36297075e-02
2.92833894e-01 -3.85407135e-02 -1.01991618e+00 9.08282816e-01
7.92055190e-01 4.60630864e-01 -6.97861314e-01 1.23007488e+00
1.83622818e-02 2.81682193e-01 -3.17390025e-01 -2.23858297e-01
5.25328338e-01 -3.72933984e-01 6.46561444e-01 1.10517073e+00
3.56870115e-01 -3.56498063e-01 3.44188571e-01 3.15952688e-01
4.75322783e-01 1.27523541e-02 -9.82932091e-01 7.93300197e-02
5.93978286e-01 1.29507625e+00 -1.09283412e+00 -1.19148225e-01
-4.95619297e-01 3.67328256e-01 -1.91134900e-01 3.94376695e-01
-1.22647488e+00 -4.60576147e-01 8.41447711e-01 3.89617473e-01
2.64776617e-01 -5.26885390e-01 -2.78078884e-01 -5.98218679e-01
1.35401234e-01 -7.01831400e-01 3.00362825e-01 -4.84528333e-01
-7.84223795e-01 6.42093003e-01 6.10835671e-01 -1.69878435e+00
8.68824348e-02 -1.18883944e+00 -1.08586304e-01 1.60319656e-01
-1.59106052e+00 -1.25251746e+00 -8.44931602e-01 1.59766197e-01
8.19978654e-01 -6.26692057e-01 4.98659462e-01 3.71150702e-01
-6.66603684e-01 2.80757517e-01 -2.76098587e-02 -1.96256399e-01
6.95406199e-01 -7.97046244e-01 1.90277234e-01 9.49349403e-01
-1.07116446e-01 4.39603120e-01 9.33465600e-01 -6.21427953e-01
-1.64691126e+00 -1.32700884e+00 3.39139223e-01 -7.36935616e-01
6.39826119e-01 -4.08915281e-01 -5.63021362e-01 6.72675908e-01
1.60907805e-01 -1.46744549e-01 4.70729887e-01 -2.30554342e-02
4.02902663e-01 -5.29985487e-01 -1.15140247e+00 7.64454901e-01
1.00826955e+00 6.85101002e-02 -2.28823438e-01 3.29421431e-01
2.10290700e-01 -5.68629622e-01 -7.50174165e-01 6.84669375e-01
4.39739794e-01 -9.96895492e-01 1.03239250e+00 -6.50758088e-01
-5.19806817e-02 -4.50369358e-01 -2.73583233e-01 -1.15527272e+00
-7.63779804e-02 -2.50823617e-01 1.46044478e-01 9.71243143e-01
5.18963635e-01 -5.33986568e-01 8.88672411e-01 6.12800539e-01
-5.82393885e-01 -7.78524220e-01 -5.85947037e-01 -1.11286235e+00
-5.18132925e-01 -6.82450116e-01 6.40308559e-01 4.97868508e-01
-5.38462698e-01 -1.99082717e-01 -3.52207571e-01 4.94381726e-01
4.61062551e-01 1.01512496e-03 1.08345461e+00 -1.20814848e+00
4.45089519e-01 -1.98196903e-01 -8.95298302e-01 -5.00986397e-01
2.85373867e-01 -6.03974521e-01 3.61797541e-01 -1.41677976e+00
-3.89790505e-01 -6.46852612e-01 -3.40360165e-01 2.24835262e-01
1.60673022e-01 2.94413745e-01 2.21897230e-01 -1.01596564e-01
-6.72007442e-01 3.18907946e-01 1.03456604e+00 -2.10385218e-01
-1.05640121e-01 2.54764348e-01 -3.42266113e-01 6.90246463e-01
9.49030101e-01 -1.34589300e-01 -5.56598902e-01 -2.34740049e-01
5.26830480e-02 -3.30711275e-01 5.14373720e-01 -1.18672907e+00
1.95665553e-01 -6.03793740e-01 2.27088645e-01 -8.41928244e-01
6.70156538e-01 -1.45343673e+00 -7.69032352e-03 4.85493898e-01
-1.22690767e-01 5.61653733e-01 5.57191074e-01 6.45496607e-01
-3.03564966e-01 -3.37333649e-01 6.86099529e-01 -2.55684722e-02
-1.55085552e+00 -2.02833619e-02 -8.11853051e-01 -3.65693659e-01
1.87767506e+00 -5.38147926e-01 -1.63174078e-01 -1.50298297e-01
-4.12392557e-01 4.72797245e-01 5.77065825e-01 8.03162336e-01
7.26399064e-01 -1.02663529e+00 -4.70501542e-01 4.01461810e-01
4.92160201e-01 -1.75404444e-01 2.84966707e-01 9.72942352e-01
-8.79853427e-01 7.59900987e-01 -6.84466541e-01 -6.04955912e-01
-1.38153076e+00 7.27507532e-01 1.76017910e-01 5.65613210e-02
-3.38923484e-01 4.58481908e-01 -6.45720661e-02 -3.62511694e-01
2.76979268e-01 -1.94606379e-01 -3.71639580e-01 1.69823915e-01
3.19946885e-01 6.81259513e-01 4.40216154e-01 -7.04123497e-01
-5.48335195e-01 3.12237144e-01 -2.53499262e-02 2.57371485e-01
1.03912163e+00 -2.17376485e-01 2.36875072e-01 2.57526308e-01
5.01305580e-01 -5.83973527e-02 -1.22985005e+00 3.27177972e-01
2.05172777e-01 -6.86973453e-01 7.25314301e-03 -6.65696979e-01
-9.32340503e-01 6.88989401e-01 7.24947989e-01 1.77088808e-02
1.05653989e+00 -2.16455534e-01 4.88988787e-01 8.63480568e-01
6.37303352e-01 -1.34543645e+00 5.31795472e-02 3.92090291e-01
8.84169579e-01 -1.37003362e+00 1.72258660e-01 -7.01285362e-01
-9.69706953e-01 1.06319773e+00 9.49905753e-01 -6.79432079e-02
2.26988897e-01 4.78047609e-01 3.17791134e-01 -3.63045305e-01
-7.46847510e-01 -4.62432891e-01 2.49363318e-01 9.16136265e-01
7.37229437e-02 -1.13988720e-01 -1.01838574e-01 3.96425389e-02
2.24397257e-02 -3.12736154e-01 4.86223280e-01 1.35362756e+00
-5.05785823e-01 -1.04918265e+00 -4.60358053e-01 3.32791477e-01
-1.24161700e-02 4.13718820e-01 -3.11533958e-01 1.40406883e+00
3.82472157e-01 9.94735956e-01 1.13723308e-01 -5.52377284e-01
8.70811164e-01 -2.58423239e-01 4.83731329e-01 -4.20991778e-01
-6.84826255e-01 -2.87375569e-01 6.50133431e-01 -5.44699371e-01
-4.41096842e-01 -8.41152191e-01 -1.07959795e+00 -6.09738305e-02
-4.37209010e-01 1.55348733e-01 9.23309386e-01 6.82803690e-01
2.82273799e-01 7.28957474e-01 5.30089498e-01 -1.08737075e+00
-3.58843029e-01 -5.40350914e-01 -3.55394334e-01 2.61795968e-01
2.17311084e-02 -1.26196468e+00 7.94006959e-02 -2.21247738e-03]
|
[8.029773712158203, -1.1346460580825806]
|
c9b33952-8f69-40f9-8067-62139b3132aa
|
protecting-the-protected-group-circumventing
|
1905.10546
| null |
https://arxiv.org/abs/1905.10546v3
|
https://arxiv.org/pdf/1905.10546v3.pdf
|
Protecting the Protected Group: Circumventing Harmful Fairness
|
Machine Learning (ML) algorithms shape our lives. Banks use them to determine if we are good borrowers; IT companies delegate them recruitment decisions; police apply ML for crime-prediction, and judges base their verdicts on ML. However, real-world examples show that such automated decisions tend to discriminate against protected groups. This potential discrimination generated a huge hype both in media and in the research community. Quite a few formal notions of fairness were proposed, which take a form of constraints a "fair" algorithm must satisfy. We focus on scenarios where fairness is imposed on a self-interested party (e.g., a bank that maximizes its revenue). We find that the disadvantaged protected group can be worse off after imposing a fairness constraint. We introduce a family of \textit{Welfare-Equalizing} fairness constraints that equalize per-capita welfare of protected groups, and include \textit{Demographic Parity} and \textit{Equal Opportunity} as particular cases. In this family, we characterize conditions under which the fairness constraint helps the disadvantaged group. We also characterize the structure of the optimal \textit{Welfare-Equalizing} classifier for the self-interested party, and provide an algorithm to compute it. Overall, our \textit{Welfare-Equalizing} fairness approach provides a unified framework for discussing fairness in classification in the presence of a self-interested party.
|
['Moshe Tennenholtz', 'Omer Ben-Porat', 'Fedor Sandomirskiy']
|
2019-05-25
| null | null | null | null |
['crime-prediction']
|
['miscellaneous']
|
[ 6.48750663e-02 4.01233047e-01 -8.49051416e-01 -6.42235100e-01
-5.08183658e-01 -5.27536690e-01 4.23395097e-01 3.93762767e-01
-8.47679019e-01 1.06324661e+00 2.43541971e-01 -7.67722547e-01
-2.78776646e-01 -1.06230474e+00 -2.45977696e-02 -7.68731654e-01
3.62863123e-01 5.78544736e-01 -5.67996562e-01 -5.54464422e-02
5.56304932e-01 5.86519063e-01 -1.20131874e+00 5.67728616e-02
1.21158683e+00 6.37340784e-01 -4.66178030e-01 3.40283871e-01
5.42812273e-02 9.89034832e-01 -4.82411027e-01 -1.17796659e+00
6.88154042e-01 -2.74717122e-01 -1.08971846e+00 -3.05798531e-01
2.15623200e-01 -5.86864710e-01 -9.98235792e-02 1.17621386e+00
4.38969880e-01 1.74621582e-01 1.03925490e+00 -1.59033012e+00
-7.19001770e-01 9.86410022e-01 -5.13209105e-01 1.39750808e-01
1.66117221e-01 4.86914255e-02 1.26964569e+00 -2.02816933e-01
5.01462519e-01 1.29607415e+00 4.94207770e-01 8.87749732e-01
-1.23432803e+00 -9.19265687e-01 -1.76341999e-02 -8.92937928e-02
-9.63603914e-01 -7.16979802e-01 4.33488607e-01 -6.24143898e-01
6.25724018e-01 8.15291405e-01 4.81171936e-01 5.82185030e-01
1.73562720e-01 6.20249748e-01 1.16192484e+00 -4.62208062e-01
3.88789177e-01 2.03589633e-01 7.25674212e-01 3.67695361e-01
8.81744683e-01 7.85076693e-02 -4.65917826e-01 -7.62712002e-01
2.85113543e-01 2.87128948e-02 -4.48741317e-02 9.52174217e-02
-3.85273963e-01 1.30370259e+00 3.81108187e-02 1.20636664e-01
-1.29528433e-01 -9.92654935e-02 4.10069853e-01 3.58939499e-01
7.44680822e-01 4.46886063e-01 -1.68740794e-01 1.08745620e-01
-1.18018019e+00 5.14714122e-01 9.15383399e-01 4.72498834e-01
6.72849774e-01 -3.06778222e-01 -4.87178057e-01 7.37811923e-01
6.08876608e-02 4.94125158e-01 7.50846579e-04 -1.46302402e+00
7.25301683e-01 4.21357691e-01 2.65869021e-01 -1.01867437e+00
-3.62494498e-01 -2.41998494e-01 -9.24539089e-01 4.06389892e-01
1.06977487e+00 -1.97440714e-01 -2.07861021e-01 2.00158978e+00
1.88566409e-02 -5.14992774e-01 -1.59701884e-01 7.76529551e-01
1.88842699e-01 2.35720277e-01 4.81431812e-01 -8.37087572e-01
1.02779114e+00 -2.56956398e-01 -5.47738433e-01 -7.53481910e-02
7.77938247e-01 -4.36314404e-01 7.53946841e-01 2.23715991e-01
-1.28038347e+00 1.19528078e-01 -4.32208747e-01 -2.90190857e-02
6.72463328e-03 -3.62026542e-01 8.26205075e-01 1.42234790e+00
-8.11264932e-01 9.12856281e-01 -3.23676705e-01 -1.78884357e-01
9.46397007e-01 3.91774476e-01 -2.72879153e-01 1.85755655e-01
-1.10345674e+00 9.68195021e-01 -2.48893872e-01 -3.09367239e-01
-3.81312042e-01 -6.48261666e-01 -6.69440448e-01 2.98366457e-01
2.42820770e-01 -7.55393505e-01 8.88969600e-01 -1.22387290e+00
-7.59659588e-01 1.64282167e+00 -1.44109085e-01 -4.97358203e-01
1.10529149e+00 1.43398225e-01 -4.26767394e-02 -1.56310007e-01
6.55473232e-01 3.10269836e-02 3.45199943e-01 -9.83013391e-01
-7.84254372e-01 -8.61057162e-01 3.16860825e-01 1.55224532e-01
-5.28608680e-01 7.74242222e-01 4.63746607e-01 -5.24509788e-01
-1.94061518e-01 -6.59842551e-01 -3.59974742e-01 -1.10009469e-01
-7.40534902e-01 -3.72046769e-01 2.49811679e-01 -4.80748206e-01
1.54369438e+00 -1.66651583e+00 -5.60505748e-01 4.59433526e-01
3.35914910e-01 1.77059084e-01 3.63750100e-01 -1.14437699e-01
7.34567642e-02 7.24589348e-01 -3.02939922e-01 -3.19842309e-01
3.73745203e-01 7.38708004e-02 -1.39644533e-01 8.61497700e-01
-1.92412242e-01 4.91549343e-01 -5.59089363e-01 -5.76498151e-01
-1.89065561e-01 -1.74113631e-01 -9.52307642e-01 9.65184718e-03
5.32073736e-01 1.08008273e-01 -5.49869895e-01 6.08491719e-01
8.16819668e-01 3.90535928e-02 3.15965861e-01 4.63093340e-01
-3.02075177e-01 3.40974122e-01 -9.92600203e-01 5.10559201e-01
-1.76260695e-02 3.37909788e-01 3.92841011e-01 -1.42065084e+00
7.83280492e-01 -6.72547594e-02 5.23146152e-01 -4.13160264e-01
5.14513850e-01 3.50718230e-01 3.13945748e-02 -3.33145827e-01
5.78455865e-01 -7.33398557e-01 -3.53957355e-01 1.04616702e+00
-5.07528186e-01 4.00240682e-02 4.87249270e-02 2.78400481e-01
1.05709529e+00 -5.02646744e-01 6.30310714e-01 -8.02915037e-01
3.78661424e-01 -1.67616889e-01 1.01959717e+00 1.03481638e+00
-8.92315269e-01 2.73515195e-01 9.11281168e-01 -5.97782016e-01
-9.85707462e-01 -8.07722569e-01 -2.05035359e-01 1.35114944e+00
-8.72047469e-02 2.40878910e-01 -9.81238186e-01 -8.93956006e-01
3.58382344e-01 9.40901995e-01 -5.66198170e-01 -1.20789506e-01
-4.05126393e-01 -1.12173879e+00 6.42734468e-01 3.56882438e-02
5.76217949e-01 -6.49598539e-01 -8.70924950e-01 -1.39250591e-01
-3.57753426e-01 -5.55782855e-01 -4.81898248e-01 1.16479352e-01
-6.33105040e-01 -1.05229771e+00 -6.36385441e-01 -1.35015294e-01
6.34184599e-01 -6.29795715e-02 1.12570727e+00 4.61112738e-01
-2.36286167e-02 3.12215858e-03 -5.69875427e-02 -6.63511992e-01
-4.90477145e-01 -4.69322354e-02 3.16387445e-01 -8.26308802e-02
5.73915541e-01 -3.59483778e-01 -4.33185995e-01 2.45995864e-01
-4.66673315e-01 -1.32931799e-01 -3.42807114e-01 7.21961975e-01
1.43787980e-01 -1.56295821e-02 1.06463778e+00 -1.39280355e+00
8.61958265e-01 -6.27482355e-01 -3.58242720e-01 4.55273658e-01
-9.42743480e-01 -1.80988520e-01 4.75802600e-01 -1.76067159e-01
-1.24650383e+00 -3.29462469e-01 3.93418372e-02 2.95493364e-01
-1.07698916e-02 1.47110045e-01 -4.92713213e-01 1.39939874e-01
6.88085020e-01 -5.40883541e-01 -1.66461170e-01 -2.60148108e-01
1.27912546e-02 1.00084412e+00 3.39145422e-01 -1.09747100e+00
4.68862504e-01 4.59507734e-01 -2.00744644e-02 -3.75826538e-01
-9.55895424e-01 -3.57038379e-02 -3.01114351e-01 -3.66794109e-01
6.23178601e-01 -4.32191819e-01 -1.19880104e+00 2.57959425e-01
-1.08677876e+00 -3.25456053e-01 -4.13429886e-01 3.88528258e-01
-5.61394095e-01 4.02207166e-01 -4.13839728e-01 -1.63363993e+00
-3.99189234e-01 -8.60452175e-01 2.70455539e-01 3.47218782e-01
-4.66029882e-01 -7.84483135e-01 -2.25567654e-01 9.79695737e-01
2.29455456e-01 3.10677946e-01 1.04856730e+00 -8.51797938e-01
1.02588832e-01 -1.24727443e-01 -3.06229323e-01 2.33271316e-01
-1.58404902e-01 1.18284009e-01 -9.43521202e-01 -3.14546116e-02
9.53259915e-02 -2.23156020e-01 8.28765750e-01 5.91946185e-01
1.11686802e+00 -9.12696123e-01 -1.61915481e-01 3.92691046e-01
1.19233620e+00 1.60089090e-01 3.95700634e-01 2.24725172e-01
2.44677052e-01 1.11850870e+00 3.85530740e-01 9.08455968e-01
5.88377178e-01 5.09531617e-01 2.91267276e-01 -6.95763826e-02
4.97322738e-01 8.79550353e-02 2.32810915e-01 1.91418931e-01
-6.86153650e-01 -1.72041506e-01 -8.71261656e-01 4.07090575e-01
-1.86407721e+00 -1.44318771e+00 -1.56133488e-01 2.41927505e+00
1.06285071e+00 1.13595188e-01 5.45698345e-01 3.82331610e-01
9.51537430e-01 -1.85460240e-01 -5.28533459e-01 -1.05463886e+00
-2.55861223e-01 3.76212150e-01 7.83191621e-01 8.42302084e-01
-9.58490968e-01 7.09122956e-01 6.28489494e+00 1.06143236e+00
-5.97806215e-01 2.24462152e-01 1.53312349e+00 -3.18061233e-01
-6.41861737e-01 8.29015821e-02 -6.15439832e-01 5.74481308e-01
5.08733928e-01 -1.00223899e+00 2.43830234e-01 8.18015158e-01
5.13929307e-01 -2.86335170e-01 -1.09389949e+00 6.83259845e-01
-2.80237496e-01 -1.09381807e+00 -1.98645085e-01 5.48143029e-01
8.17449093e-01 -4.91126299e-01 4.89503331e-02 1.64182335e-01
7.79473543e-01 -1.20300257e+00 8.98343742e-01 5.93476832e-01
9.34703648e-01 -1.14538956e+00 5.48652947e-01 8.16537976e-01
-5.39418638e-01 -3.47121716e-01 -4.02635902e-01 -8.34987462e-01
1.33952899e-02 1.11219382e+00 -1.68615431e-01 3.34544152e-01
4.28469837e-01 1.13195971e-01 -4.18265956e-03 6.64695084e-01
-8.47404823e-02 4.27938223e-01 7.80237932e-03 7.49055520e-02
6.78470656e-02 -3.81230414e-01 3.02584440e-01 1.22568333e+00
2.21261263e-01 8.39009404e-01 7.36104622e-02 9.82285321e-01
-3.43483418e-01 3.85174870e-01 -6.21435523e-01 3.71922046e-01
5.12930691e-01 9.95912671e-01 -6.68204546e-01 -3.73114765e-01
-5.67768328e-02 4.12073314e-01 3.66666377e-01 2.10823845e-02
-4.83333468e-01 -1.81349516e-01 1.00908375e+00 3.18785518e-01
-7.88663208e-01 3.79860729e-01 -1.05256021e+00 -1.38483608e+00
-4.00661737e-01 -8.50042284e-01 7.82991529e-01 -2.70167291e-01
-1.43983233e+00 -4.65130843e-02 1.33395763e-02 -5.84964573e-01
4.93239760e-02 -4.42423284e-01 -7.71997631e-01 1.10590446e+00
-1.36217797e+00 -6.46006048e-01 2.46935591e-01 5.72663784e-01
2.32994128e-02 -3.03573877e-01 7.78890133e-01 3.46703649e-01
-6.51936173e-01 9.22765315e-01 -9.89514515e-02 1.96554989e-01
4.07334387e-01 -9.14273262e-01 -2.79297143e-01 1.01671410e+00
-3.99338514e-01 7.79242933e-01 6.77382946e-01 -6.74232841e-01
-5.97798228e-01 -8.58444452e-01 1.58632982e+00 -5.19597948e-01
2.46407136e-01 -1.21468894e-01 -5.07833421e-01 4.79338109e-01
-2.25049436e-01 -2.20648035e-01 9.40934002e-01 2.96526819e-01
-4.60231453e-01 -2.07865518e-02 -1.88381457e+00 5.40682435e-01
1.14846838e+00 -5.54474831e-01 -3.24979722e-01 3.12325448e-01
1.64394692e-01 2.61156797e-01 -5.48809946e-01 1.23787060e-01
8.76769781e-01 -1.37574077e+00 7.44518399e-01 -1.01847196e+00
6.35864139e-01 3.58870089e-01 -1.36179984e-01 -9.35930729e-01
-4.06881213e-01 -6.47343695e-01 4.38696980e-01 1.18507922e+00
4.13104862e-01 -7.90471256e-01 8.49184692e-01 1.65223229e+00
3.59991878e-01 -6.32297933e-01 -1.19636261e+00 -6.67411983e-01
8.71029317e-01 -4.70528513e-01 7.24867880e-01 1.62032139e+00
5.98991036e-01 -1.08854897e-01 -4.94371593e-01 -2.69978851e-01
9.62604880e-01 7.31491074e-02 5.02197862e-01 -1.48712063e+00
-1.59924284e-01 -9.33972061e-01 -5.58273755e-02 -3.93490344e-01
6.22550607e-01 -1.04354835e+00 -4.03973728e-01 -1.18273473e+00
9.33988869e-01 -9.68910873e-01 -1.49312511e-01 6.54801607e-01
-3.19969684e-01 -4.61696722e-02 4.61988449e-01 2.47259155e-01
-1.71285376e-01 -8.48677009e-02 9.00305331e-01 -2.04369545e-01
-2.31305622e-02 4.60049093e-01 -1.52281082e+00 8.61066520e-01
9.93734658e-01 -6.21301889e-01 2.19881013e-02 -2.09972128e-01
4.71533328e-01 2.67703772e-01 3.10675502e-01 -3.48357558e-01
1.32554010e-01 -9.51321781e-01 1.76724881e-01 5.06564714e-02
-1.63668215e-01 -4.88661975e-01 1.74965963e-01 7.01140642e-01
-7.03777254e-01 -5.25249720e-01 -5.05400002e-01 -5.10588996e-02
3.43777418e-01 -5.98279595e-01 1.16528320e+00 -1.41824916e-01
3.57694656e-01 3.32288206e-01 -4.70490038e-01 2.90026337e-01
9.00737882e-01 -2.24645108e-01 -6.89294636e-01 -5.88766694e-01
-5.50609231e-01 2.48478219e-01 5.63622892e-01 -3.00529927e-01
8.29447433e-02 -1.11248827e+00 -1.11148489e+00 -1.29784212e-01
-2.38347650e-01 -5.80797553e-01 -1.49332717e-01 6.95301771e-01
-1.11653693e-01 1.19570240e-01 -3.45277727e-01 3.07108045e-01
-1.38308728e+00 1.96798936e-01 5.02216518e-01 -5.75557828e-01
4.35833931e-02 9.25591469e-01 4.07263577e-01 -2.90864736e-01
1.13011124e-02 1.41805202e-01 -1.89895347e-01 3.53812993e-01
5.59141159e-01 9.66561556e-01 -2.19890416e-01 -8.55982900e-01
-4.70745027e-01 1.53034121e-01 1.33094743e-01 -2.90815026e-01
1.22137558e+00 -1.39245301e-01 -4.31539714e-01 -2.82365624e-02
8.10505271e-01 1.71135828e-01 -7.80001342e-01 -4.92617227e-02
3.76140140e-02 -8.77403259e-01 -1.73264191e-01 -7.39034891e-01
-1.07870281e+00 7.78329670e-01 -1.53182134e-01 6.90159678e-01
1.10391033e+00 -2.08040282e-01 1.68021664e-01 1.75707728e-01
6.19841814e-01 -1.57901013e+00 -6.34517848e-01 3.81597579e-01
5.82848012e-01 -1.13422906e+00 4.18671221e-01 -3.11606139e-01
-6.74527466e-01 8.39993119e-01 4.13014978e-01 -4.22318503e-02
6.69102252e-01 1.17169753e-01 -1.87724754e-01 1.21607415e-01
-3.84726852e-01 -3.25244516e-02 -1.69928461e-01 5.18132329e-01
4.68454719e-01 8.47351193e-01 -1.25192761e+00 1.17394102e+00
-5.59682369e-01 -1.10558420e-01 6.88515723e-01 5.78696847e-01
-7.48387277e-01 -1.14004958e+00 -5.17482400e-01 9.98774707e-01
-9.89962399e-01 -8.87771919e-02 -6.86127126e-01 4.39916879e-01
6.82781935e-01 1.34455550e+00 2.61595249e-02 -2.42873058e-02
1.43348992e-01 -3.72829922e-02 2.54022598e-01 -6.14322722e-01
-9.82016206e-01 -2.44013980e-01 4.65011865e-01 -3.48544985e-01
-5.28906465e-01 -9.24362421e-01 -1.11600780e+00 -1.27157450e+00
-4.24753189e-01 2.41554394e-01 1.22611269e-01 1.07841194e+00
-2.38630340e-01 -2.70405799e-01 8.95015836e-01 -3.90883923e-01
-6.28712773e-01 -4.39553887e-01 -1.03792858e+00 4.37747449e-01
5.65474778e-02 -3.39114428e-01 -6.84498250e-01 -1.04195803e-01]
|
[8.83655834197998, 5.3765740394592285]
|
48763f05-6eb8-4a06-8d95-d7d4b9c9c54a
|
4d-or-semantic-scene-graphs-for-or-domain
|
2203.11937
| null |
https://arxiv.org/abs/2203.11937v1
|
https://arxiv.org/pdf/2203.11937v1.pdf
|
4D-OR: Semantic Scene Graphs for OR Domain Modeling
|
Surgical procedures are conducted in highly complex operating rooms (OR), comprising different actors, devices, and interactions. To date, only medically trained human experts are capable of understanding all the links and interactions in such a demanding environment. This paper aims to bring the community one step closer to automated, holistic and semantic understanding and modeling of OR domain. Towards this goal, for the first time, we propose using semantic scene graphs (SSG) to describe and summarize the surgical scene. The nodes of the scene graphs represent different actors and objects in the room, such as medical staff, patients, and medical equipment, whereas edges are the relationships between them. To validate the possibilities of the proposed representation, we create the first publicly available 4D surgical SSG dataset, 4D-OR, containing ten simulated total knee replacement surgeries recorded with six RGB-D sensors in a realistic OR simulation center. 4D-OR includes 6734 frames and is richly annotated with SSGs, human and object poses, and clinical roles. We propose an end-to-end neural network-based SSG generation pipeline, with a rate of success of 0.75 macro F1, indeed being able to infer semantic reasoning in the OR. We further demonstrate the representation power of our scene graphs by using it for the problem of clinical role prediction, where we achieve 0.85 macro F1. The code and dataset will be made available upon acceptance.
|
['Nassir Navab', 'Federico Tombari', 'Tobias Czempiel', 'Ulrich Eck', 'Evin Pınar Örnek', 'Ege Özsoy']
|
2022-03-22
| null | null | null | null |
['scene-graph-generation']
|
['computer-vision']
|
[ 3.73967499e-01 7.39713192e-01 1.98699087e-01 -2.47254103e-01
-3.68355811e-01 -4.18349147e-01 3.32848281e-01 5.30743361e-01
-8.97638276e-02 2.50353098e-01 5.52652657e-01 -3.41375887e-01
-6.21845067e-01 -4.15248990e-01 -7.03151762e-01 -3.69965971e-01
-1.95112735e-01 6.35048449e-01 3.29558887e-02 -1.10768348e-01
-1.94359228e-01 7.09424078e-01 -1.44756091e+00 4.35543358e-01
1.55154482e-01 9.76874888e-01 2.42251083e-01 7.59673238e-01
4.35022771e-01 1.00964618e+00 -5.64865947e-01 -7.73088858e-02
2.99557358e-01 -2.88677096e-01 -8.05165231e-01 2.51937747e-01
1.49732381e-01 -2.00529099e-01 -5.52655697e-01 4.49403077e-01
6.05659783e-01 1.62726734e-02 2.13621810e-01 -1.01119149e+00
4.39976668e-03 5.11844218e-01 1.28510579e-01 -6.67520463e-02
7.04390645e-01 2.21939221e-01 5.41842699e-01 -7.27897525e-01
9.91101980e-01 1.09668922e+00 5.10325134e-01 6.33737266e-01
-8.31887126e-01 -3.42642903e-01 6.99375570e-02 -4.73224409e-02
-1.06715488e+00 -3.98538053e-01 7.63498843e-01 -4.41634774e-01
8.36657286e-01 3.71819377e-01 1.12432122e+00 1.32111692e+00
2.05901191e-01 7.19966650e-01 6.65514648e-01 -3.02387595e-01
3.54007125e-01 -6.46984577e-02 -8.94322246e-02 1.00711179e+00
3.98886561e-01 -3.93469594e-02 -6.77386582e-01 -1.18545666e-01
1.11181831e+00 3.37218642e-01 -3.74409258e-01 -9.11605299e-01
-1.75420547e+00 3.51012200e-01 7.03424215e-01 2.53268838e-01
-5.33579826e-01 4.01641130e-01 3.35008025e-01 -1.73215732e-01
1.77707504e-02 8.39585423e-01 -3.39112431e-01 -1.73893481e-01
-3.19379330e-01 6.37219250e-02 9.08907056e-01 1.18462265e+00
1.39453243e-02 -4.18669343e-01 -1.62069932e-01 4.13786739e-01
1.22409597e-01 -1.32293150e-01 2.08177388e-01 -9.57607508e-01
2.27219000e-01 9.33719754e-01 1.93334684e-01 -1.00459695e+00
-9.09308672e-01 -5.20180166e-01 -6.68265224e-01 -2.62831487e-02
-2.13375688e-03 2.28804667e-02 -1.03608048e+00 1.25981987e+00
4.75056708e-01 4.23513442e-01 2.96347104e-02 9.33199346e-01
1.26107037e+00 -4.78210263e-02 2.56537080e-01 -3.22643341e-03
1.52868640e+00 -9.35139894e-01 -7.36752808e-01 -4.24575597e-01
8.88167620e-01 -4.75554883e-01 8.61616194e-01 3.22785288e-01
-1.14914870e+00 -4.76890922e-01 -9.54244375e-01 7.11648315e-02
-2.26192579e-01 8.14583898e-02 8.16053152e-01 1.40379325e-01
-9.47926402e-01 4.86538887e-01 -1.26628768e+00 -4.81016785e-01
6.32108927e-01 4.69101071e-01 -6.58325016e-01 -3.71211946e-01
-8.11406493e-01 8.91759634e-01 3.42988640e-01 4.28593844e-01
-1.26138639e+00 -7.97372997e-01 -1.18074656e+00 -1.40140012e-01
6.53055787e-01 -1.17469776e+00 1.11762309e+00 -3.87333870e-01
-1.10980499e+00 1.16568780e+00 1.49809912e-01 -1.26768708e-01
5.80511689e-01 -3.02618861e-01 -2.64192909e-01 3.29031527e-01
-1.33677647e-01 4.79691982e-01 2.95592010e-01 -1.47044635e+00
-1.70550168e-01 -5.70795715e-01 4.98049945e-01 3.61832172e-01
1.01871639e-01 -1.46547318e-01 -6.22911692e-01 -4.94092464e-01
2.08404675e-01 -1.16165435e+00 -6.85381293e-01 2.75780648e-01
-5.57881892e-01 2.58182406e-01 4.31652427e-01 -5.33652604e-01
9.45571244e-01 -2.29333353e+00 3.75963807e-01 1.56753138e-01
5.23040712e-01 6.48716018e-02 1.44422770e-01 4.04262066e-01
-4.43973213e-01 -1.75300658e-01 -1.75062656e-01 -5.28848469e-01
-1.32560313e-01 4.44628924e-01 3.09736788e-01 5.97749233e-01
-1.19900756e-01 9.45909619e-01 -1.12756586e+00 -4.68856275e-01
6.81267858e-01 6.01434171e-01 -5.89411020e-01 3.79520535e-01
-1.90934762e-02 8.91014218e-01 -6.10284269e-01 7.30571330e-01
8.82614851e-02 -3.91180634e-01 4.05419499e-01 -4.28305477e-01
2.92710185e-01 1.58801422e-01 -1.06196237e+00 2.37916231e+00
-5.96840024e-01 4.04464275e-01 1.45907700e-01 -7.90173352e-01
6.46384954e-01 4.63063776e-01 9.56621587e-01 -4.37771261e-01
4.85022217e-01 2.73733083e-02 -8.04426894e-02 -7.91489959e-01
1.36868536e-01 -2.98294395e-01 -3.02713245e-01 -1.64015472e-01
-2.57188752e-02 -4.89084661e-01 -8.59876052e-02 1.70979485e-01
1.46343648e+00 8.10576826e-02 5.06037056e-01 -1.76879391e-02
2.38846406e-01 3.15345287e-01 1.40049383e-01 6.35805130e-01
-1.97302952e-01 6.52152181e-01 6.41662478e-01 -7.47142911e-01
-6.75125659e-01 -1.09717643e+00 5.27857170e-02 4.84391689e-01
5.50373495e-01 -4.80810165e-01 -5.37928045e-01 -6.04250491e-01
-7.82870948e-02 6.47901952e-01 -8.47968519e-01 -3.95994127e-01
-7.51455367e-01 -2.91023523e-01 9.24102738e-02 7.01257527e-01
-1.56211212e-01 -1.11079407e+00 -1.10551167e+00 2.07106516e-01
-2.93333866e-02 -1.60647154e+00 1.62591875e-01 5.22001050e-02
-9.50122833e-01 -1.56292403e+00 -3.58188838e-01 -6.09024704e-01
1.00939298e+00 4.31701019e-02 1.23210669e+00 7.68688247e-02
-9.86976385e-01 8.10751081e-01 -3.75794202e-01 -4.92729336e-01
-3.74179661e-01 -3.54931116e-01 -1.44794688e-01 -2.24030435e-01
-2.93927044e-01 -2.43414566e-01 -1.04756737e+00 3.05216491e-01
-7.77697861e-01 5.15182495e-01 5.96110582e-01 5.10679781e-01
6.44371867e-01 -1.13231748e-01 -2.38052532e-01 -9.96548235e-01
3.39209735e-01 -5.32393217e-01 -1.58258617e-01 5.88591509e-02
-4.59266081e-02 -3.20724905e-01 4.59645152e-01 -4.31255512e-02
-6.85535192e-01 4.24882412e-01 1.19489685e-01 -7.27464318e-01
-5.54225206e-01 5.17707944e-01 -6.61931857e-02 1.75012589e-01
6.24713778e-01 -1.98484480e-01 1.07726520e-02 -3.51730883e-01
3.20959508e-01 1.43414631e-01 7.22732306e-01 -4.38692778e-01
4.25802112e-01 6.52667165e-01 4.12981212e-01 -3.99606466e-01
-8.83153498e-01 -5.80600977e-01 -5.50066710e-01 -4.89816248e-01
9.62921262e-01 -9.47069883e-01 -1.02022958e+00 1.39688551e-01
-9.41334426e-01 -3.85944813e-01 -3.92158866e-01 7.10554779e-01
-7.96946764e-01 -9.13218930e-02 -4.39747393e-01 -5.41893542e-01
-3.45233232e-02 -1.37992775e+00 1.55911958e+00 -8.28975067e-02
-3.85726392e-01 -1.07735324e+00 -2.58228570e-01 6.72086954e-01
1.40891686e-01 1.08786666e+00 8.21259379e-01 -6.25988960e-01
-5.85394621e-01 -5.20154715e-01 6.18194751e-02 -5.54059148e-02
3.68998647e-01 -3.50902885e-01 -6.11925244e-01 -3.23786169e-01
-8.16477016e-02 1.06536068e-01 1.25810832e-01 4.49270874e-01
1.59221697e+00 -4.93470468e-02 -7.50076532e-01 4.85720038e-01
1.27392745e+00 2.75450587e-01 4.32481289e-01 1.55648112e-01
1.02158678e+00 8.18791568e-01 8.15133810e-01 6.66163266e-01
3.65590543e-01 7.53512502e-01 1.05352497e+00 -4.57005173e-01
-2.31297329e-01 -1.90247372e-01 -2.11958706e-01 5.47499001e-01
-3.06061476e-01 -2.51933694e-01 -1.38111794e+00 5.63725948e-01
-1.93540084e+00 -5.02838016e-01 -1.74341679e-01 1.95556259e+00
3.89659733e-01 8.51426497e-02 -2.66836494e-01 4.11176831e-02
3.16437989e-01 3.34311388e-02 -4.83909190e-01 3.08092143e-02
4.82049584e-01 2.74008691e-01 5.05839169e-01 2.78835416e-01
-1.03714967e+00 6.41148031e-01 5.85691500e+00 1.72784775e-01
-7.37904429e-01 -1.52879655e-01 5.12397528e-01 -5.17943621e-01
1.21824384e-01 -1.43260285e-01 -2.51023471e-01 -3.75784142e-03
8.00667703e-01 8.95812213e-02 1.93329871e-01 9.86275911e-01
3.51461828e-01 -2.79753339e-02 -1.57369757e+00 1.16666877e+00
2.54897803e-01 -1.47621059e+00 -8.72385129e-02 -1.73896104e-01
3.51469606e-01 -3.19405675e-01 -2.65753746e-01 -1.28848962e-02
2.67782807e-01 -1.25159883e+00 7.69859016e-01 7.05908954e-01
7.72456765e-01 -3.35916400e-01 7.94305801e-01 3.23402524e-01
-1.05930030e+00 -2.59965211e-01 3.06845635e-01 1.63801178e-01
2.44045973e-01 2.23892927e-01 -1.30204344e+00 7.91057348e-01
7.35434294e-01 8.87617826e-01 -3.42994899e-01 9.53281403e-01
-1.35895148e-01 -6.93452209e-02 -1.07070804e-01 2.33111113e-01
1.36082813e-01 2.94700503e-01 4.75626945e-01 9.71555054e-01
2.47034386e-01 3.82212281e-01 2.25127399e-01 3.18649411e-01
2.95757726e-02 -2.67403275e-01 -7.71707296e-01 -9.06881690e-03
9.78608355e-02 1.13690913e+00 -8.51022303e-01 -1.79051220e-01
-4.51718420e-02 8.35212409e-01 -9.97303128e-02 1.47656471e-01
-1.05599761e+00 2.63349377e-02 7.61249721e-01 4.87517029e-01
-3.12539309e-01 -3.64624619e-01 -3.12707633e-01 -8.10360372e-01
4.56366949e-02 -8.13710928e-01 4.89832520e-01 -1.15001917e+00
-6.88618720e-01 5.97916722e-01 2.96709180e-01 -1.48547804e+00
-9.33747292e-02 -1.05561829e+00 2.50364318e-02 4.87352371e-01
-1.05948615e+00 -1.33198023e+00 -1.08995330e+00 7.76010931e-01
6.35626078e-01 1.15363158e-01 1.03409338e+00 2.34663665e-01
-4.86129075e-01 -6.19583465e-02 -5.19299090e-01 8.49338621e-02
3.18703890e-01 -9.95748281e-01 1.14536203e-01 3.38201851e-01
-1.80908382e-01 7.58745193e-01 7.32901871e-01 -4.64646012e-01
-1.71367121e+00 -9.40278709e-01 3.86096954e-01 -9.05273855e-01
4.92432326e-01 -5.63549042e-01 -2.68785149e-01 9.10296321e-01
-3.76233846e-01 3.10812891e-01 6.96888089e-01 -2.57401764e-01
2.12556973e-01 1.32720411e-01 -1.13313544e+00 7.49419510e-01
1.80882835e+00 -3.30611706e-01 -5.31457663e-01 8.62770617e-01
8.85512769e-01 -1.22912300e+00 -1.14521039e+00 8.34625304e-01
3.30062389e-01 -1.03774941e+00 1.37507963e+00 -6.19183540e-01
6.63725674e-01 -2.47604162e-01 -3.69816124e-02 -1.08762801e+00
4.40723486e-02 -4.42227930e-01 -9.05721039e-02 2.32928112e-01
7.33605027e-02 -3.22600663e-01 8.31691861e-01 8.35853875e-01
-7.25497007e-01 -1.19661796e+00 -8.91746819e-01 -3.05551201e-01
-6.53122127e-01 -6.24259114e-01 5.42045295e-01 9.36252356e-01
-3.07123866e-02 -7.83182010e-02 -6.98377788e-02 4.17540342e-01
4.06595170e-01 3.41212451e-02 6.75504923e-01 -1.22094655e+00
-4.42864209e-01 -2.90214390e-01 -6.69564307e-01 -5.82630038e-01
-1.10498041e-01 -7.12647378e-01 1.88624195e-03 -2.24343634e+00
-1.11814449e-02 -6.32067502e-01 -2.28793085e-01 7.13535011e-01
1.07350260e-01 -3.98194455e-02 1.29034221e-01 1.46977723e-01
-6.11783206e-01 1.96095660e-01 1.66513681e+00 1.61052831e-02
-1.07694708e-01 -1.96717560e-01 -5.75086176e-01 9.79242027e-01
4.78882194e-01 -3.99749517e-01 -4.96403277e-01 -5.19969523e-01
9.75076333e-02 5.48987150e-01 6.99481308e-01 -1.06906915e+00
2.29722470e-01 -4.44228649e-02 3.67814749e-01 -2.00986147e-01
7.23896861e-01 -1.31639373e+00 6.92957044e-01 8.09997141e-01
-2.59314001e-01 -1.15136594e-01 3.65652025e-01 5.54235816e-01
-2.14516819e-01 1.90017164e-01 4.50510442e-01 -5.06566286e-01
-6.86464548e-01 3.44423920e-01 3.00684702e-02 -2.17848197e-01
1.54180980e+00 -4.18523192e-01 -2.22376555e-01 -1.10119373e-01
-1.33985245e+00 1.82166591e-01 4.20295805e-01 7.09904909e-01
8.77284050e-01 -9.28044438e-01 -5.14221191e-01 2.46147573e-01
4.34633225e-01 5.68862677e-01 6.21331513e-01 9.78643417e-01
-1.04866171e+00 3.98854464e-01 -9.44920555e-02 -6.05990529e-01
-1.30809486e+00 5.76504767e-01 4.96050954e-01 -2.26067007e-01
-9.66977835e-01 7.82979846e-01 4.31320816e-01 -4.87061828e-01
3.37387919e-01 -5.17187536e-01 -2.34504431e-01 -2.40980327e-01
1.53086230e-01 1.02688968e-01 1.99493200e-01 -4.83506829e-01
-6.98852003e-01 4.80334967e-01 3.55146021e-01 2.68335551e-01
1.43516219e+00 3.57628345e-01 7.17125600e-03 3.45603138e-01
9.38767672e-01 -2.23786935e-01 -9.49593961e-01 1.77732706e-01
-2.32950613e-01 -4.58893657e-01 -1.59948647e-01 -9.52785492e-01
-9.58085060e-01 6.05086982e-01 4.44166720e-01 1.60404071e-02
1.16962051e+00 4.56620365e-01 5.42048335e-01 1.35458469e-01
7.44122267e-01 -5.03619432e-01 3.29199165e-01 2.92867720e-02
1.22079253e+00 -9.46004868e-01 1.08628813e-02 -8.71335983e-01
-7.47265279e-01 8.65921676e-01 5.62142313e-01 1.78080574e-02
5.59450805e-01 2.08289206e-01 3.09259426e-02 -8.03231657e-01
-5.33197880e-01 1.05976425e-01 2.79312789e-01 4.40871567e-01
2.88762629e-01 2.91436404e-01 1.59504846e-01 3.75032246e-01
-2.52405047e-01 1.31835639e-01 4.13857073e-01 1.23051548e+00
1.81091815e-01 -6.22162044e-01 -2.27270916e-01 2.39884377e-01
-2.64830589e-01 1.15674920e-01 -2.58960217e-01 9.53066289e-01
2.33091012e-01 8.25252235e-01 -9.01194930e-04 -2.24258453e-01
1.11009264e+00 -3.18672031e-01 4.35101926e-01 -8.99111032e-01
-6.62407756e-01 -2.02072889e-01 3.47194999e-01 -1.20327687e+00
-5.26080847e-01 -3.26591283e-01 -1.63522744e+00 1.17072940e-01
2.63623685e-01 -2.61858404e-01 9.55071867e-01 5.59609294e-01
5.32707930e-01 1.30775559e+00 3.14978093e-01 -1.07315385e+00
4.22888026e-02 -5.81774533e-01 -3.03911209e-01 7.92608798e-01
3.48504066e-01 -7.72191226e-01 -2.02921703e-01 7.98707083e-02]
|
[14.012417793273926, -3.4129183292388916]
|
23a433ca-d1f0-455d-940b-a09cb1a76302
|
wavefront-sensor-for-millimeter-submillimeter
|
2102.09286
| null |
https://arxiv.org/abs/2102.09286v1
|
https://arxiv.org/pdf/2102.09286v1.pdf
|
Wavefront sensor for millimeter/submillimeter-wave adaptive optics based on aperture-plane interferometry
|
We present a concept of a millimeter wavefront sensor that allows real-time sensing of the surface of a ground-based millimeter/submillimeter telescope. It is becoming important for ground-based millimeter/submillimeter astronomy to make telescopes larger with keeping their surface accurate. To establish `millimetric adaptive optics (MAO)' that instantaneously corrects the wavefront degradation induced by deformation of telescope optics, our wavefront sensor based on radio interferometry measures changes in excess path lengths from characteristic positions on the primary mirror surface to the focal plane. This plays a fundamental role in planned 50-m class submillimeter telescopes such as LST and AtLAST.
|
['Kotaro Kohno', 'Toshikazu Onishi', 'Tai Oshima', 'Tatsuya Takekoshi', 'Mikio Kurita', 'Tomoko Nakamura', 'Sachiko Okumura', 'Keiichi Matsuda', 'Satoya Nakano', 'Masato Hagimoto', 'Yohei Togami', 'Nario Kuno', 'Noriyuki Kawaguchi', 'Tetsuhiro Minamidani', 'Ikumi Hashimoto', 'Hideo Ogawa', 'Nozomi Okada', 'Akio Taniguchi', 'Tetsutaro Ueda', 'Kimihiro Kimura', 'Yuhei Fukasaku', 'Ryohei Kawabe', 'Yoichi Tamura']
|
2021-02-18
| null | null | null | null |
['radio-interferometry']
|
['miscellaneous']
|
[ 2.52314895e-01 3.77258182e-01 8.61860812e-01 -3.38338315e-01
-3.22024465e-01 -6.50954604e-01 2.07962424e-01 -8.23058486e-01
-2.03217834e-01 5.15189648e-01 8.17748010e-02 -4.56381410e-01
-1.69481218e-01 -8.11298788e-01 -2.88919568e-01 -5.10367513e-01
-2.24264674e-02 9.48973715e-01 5.09418368e-01 -2.50460595e-01
4.65962082e-01 6.32804155e-01 -1.02285457e+00 3.30400951e-02
3.84202898e-01 1.21254468e+00 3.81143153e-01 9.41522419e-01
4.79766339e-01 1.39130473e-01 -4.43589061e-01 2.17074111e-01
6.48084104e-01 -4.91730511e-01 -2.90325582e-01 5.68523332e-02
7.92905807e-01 -3.20150167e-01 -2.43793294e-01 1.08460343e+00
-7.68830925e-02 -3.44526112e-01 5.96596897e-01 -4.66654181e-01
-1.97976574e-01 -4.36993778e-01 -1.03263997e-01 2.53313452e-01
2.90218592e-01 3.34505767e-01 4.45511192e-01 -5.72569728e-01
3.93777430e-01 7.66164422e-01 7.30181694e-01 1.52646661e-01
-5.69860697e-01 -4.70932126e-01 -7.52442598e-01 -2.81590998e-01
-8.64388704e-01 -6.33821428e-01 4.01578963e-01 -8.30803335e-01
8.40015173e-01 3.04192424e-01 5.00790775e-01 -5.21285795e-02
1.30842507e+00 -6.79321945e-01 1.20580506e+00 -6.93236053e-01
1.59807518e-01 1.24834217e-01 -3.66191447e-01 8.17416549e-01
8.07975590e-01 6.38989866e-01 -4.37903732e-01 1.04801446e-01
1.38663828e+00 -4.71725501e-02 -7.78363705e-01 1.67570621e-01
-1.15246916e+00 4.35029000e-01 5.23094594e-01 5.94358325e-01
-3.91155690e-01 3.84491354e-01 -9.49234545e-01 9.34546709e-01
5.87644935e-01 1.75111890e+00 -4.65559065e-01 -5.36809973e-02
-5.36761463e-01 -1.06785506e-01 3.08237523e-01 9.11076725e-01
8.40190709e-01 -1.84496909e-01 2.11686552e-01 -3.34992677e-01
1.07933342e+00 9.19008911e-01 3.78359169e-01 -1.03952610e+00
3.53485554e-01 7.57261992e-01 6.48048580e-01 -3.25649858e-01
-6.35486007e-01 -6.38136864e-01 -4.35584694e-01 9.05437350e-01
2.92982608e-01 -2.34948426e-01 -1.07813013e+00 8.49538565e-01
1.31818220e-01 3.73349451e-02 8.90477151e-02 1.02692008e+00
4.58702624e-01 1.04330033e-01 -1.09504688e+00 -5.05540490e-01
1.40756273e+00 -3.32378089e-01 -4.44967151e-01 -7.27843225e-01
6.39521420e-01 -1.33751976e+00 5.90625465e-01 4.31393236e-01
-7.93441534e-01 -1.15765497e-01 -1.09772873e+00 2.90084302e-01
3.90018940e-01 -1.61499143e-01 6.86735332e-01 6.59140170e-01
-7.59045780e-01 8.51174831e-01 -6.23406708e-01 -4.83742543e-02
-1.35678470e-01 2.06256613e-01 -5.62261567e-02 3.77531856e-01
-3.97614926e-01 7.58724988e-01 -1.67455040e-02 -2.87035346e-01
8.92804787e-02 -1.23531914e+00 9.78189185e-02 -1.31303325e-01
1.50972279e-02 -1.00589335e+00 1.47265875e+00 -3.26344818e-01
-2.08198118e+00 9.95646477e-01 -3.24900472e-03 -5.13826013e-01
-7.49108493e-02 -2.24196970e-01 -4.84479308e-01 1.70119837e-01
-3.32107931e-01 -4.06466573e-01 8.27080309e-01 -5.87835729e-01
-6.85034335e-01 -8.69534969e-01 -3.40448469e-01 -3.30613315e-01
3.33760500e-01 1.83173064e-02 4.70665991e-01 1.48258299e-01
1.31509018e+00 -1.14709806e+00 -2.79605627e-01 -2.09443137e-01
-2.82639086e-01 -5.23111820e-02 1.17089927e+00 -1.53289765e-01
5.89944422e-01 -2.02526093e+00 -2.98965007e-01 -3.11216980e-01
2.92173624e-01 6.37748316e-02 3.88196290e-01 6.80342615e-01
2.78697193e-01 -5.40794075e-01 -8.93044025e-02 -8.24660286e-02
-5.39496839e-01 -5.75991571e-01 -3.63468766e-01 6.49899423e-01
-6.73769414e-01 7.87631750e-01 -3.60152453e-01 6.08183801e-01
8.95606205e-02 -4.12267148e-01 -3.71392310e-01 3.46458584e-01
-7.33479500e-01 1.14219224e+00 -7.50053108e-01 7.11673021e-01
9.35954094e-01 -4.82351631e-02 -4.18378234e-01 2.03965940e-02
-9.89217937e-01 4.80149329e-01 -6.14496171e-01 1.31822681e+00
-3.50517392e-01 9.10942852e-01 2.14155823e-01 1.45327359e-01
1.23552859e+00 4.23182994e-01 3.02642643e-01 -8.37882578e-01
9.10941958e-02 5.36518395e-01 -2.29990363e-01 -7.08426416e-01
5.52148461e-01 -9.45349157e-01 2.26754658e-02 2.07223967e-01
-7.13939428e-01 -1.36453569e+00 -7.80353546e-01 -3.18117023e-01
1.32194316e+00 -2.57249653e-01 2.60508567e-01 -5.55799186e-01
2.67442137e-01 2.37607881e-01 3.11533660e-01 3.35841537e-01
2.25545362e-01 6.07162178e-01 -4.64983225e-01 -8.84872139e-01
-1.12306809e+00 -1.00393403e+00 -7.58662671e-02 1.26161957e-02
1.29576325e-01 6.01369068e-02 -2.64267743e-01 3.22613627e-01
4.41707782e-02 4.42924827e-01 1.15914240e-01 3.25616263e-03
-2.79122561e-01 -4.57881838e-01 -1.48690492e-01 -9.42058265e-02
8.20198357e-01 -5.18593788e-01 -1.00583434e+00 1.63604051e-01
1.90777302e-01 -1.10478389e+00 1.54978961e-01 -5.86279452e-01
-1.14425743e+00 -1.36187339e+00 -5.06669953e-02 -2.66529173e-01
3.84555221e-01 8.17679107e-01 9.97046709e-01 -4.27879035e-01
-2.40720645e-01 7.09656417e-01 -3.34670752e-01 -9.55267787e-01
-3.11095566e-01 -3.03115845e-01 4.70940292e-01 -1.09477498e-01
2.29833126e-01 -1.03494346e+00 -9.85972583e-01 3.54446083e-01
-2.99425066e-01 2.63117373e-01 9.00841236e-01 -2.43109435e-01
1.25883877e-01 -2.68749110e-02 3.98496576e-02 -7.05455780e-01
-8.52919891e-02 4.64768149e-02 -1.67602336e+00 -2.32331023e-01
-7.22326815e-01 -3.40745822e-02 5.27403653e-01 3.33167702e-01
-8.83876085e-01 -3.58090132e-01 -1.87285431e-02 1.41615078e-01
1.09798998e-01 1.48582429e-01 1.17171824e-01 -8.41971040e-01
1.07029021e+00 5.98118827e-02 -2.70100713e-01 -5.38527012e-01
-3.09298337e-01 7.46853352e-01 7.72868216e-01 3.13847393e-01
1.56777537e+00 1.06262255e+00 6.26625061e-01 -1.37438548e+00
-1.08181524e+00 -1.10744572e+00 -3.62643808e-01 -2.68315107e-01
9.55259442e-01 -7.89713323e-01 -9.74985063e-01 4.24840569e-01
-1.22113705e+00 -2.48039633e-01 -3.76577437e-01 1.36999249e+00
-3.64856869e-01 -1.61259472e-02 1.93238957e-03 -6.21610999e-01
-2.75032818e-01 -1.54086575e-01 1.21619725e+00 6.47751868e-01
-1.77685881e-03 -8.92959356e-01 7.57840812e-01 3.47241908e-01
9.18298781e-01 7.06321895e-02 4.72169966e-01 7.98452735e-01
-1.72267604e+00 -4.12361205e-01 -1.51809528e-02 -1.60649449e-01
3.70822430e-01 -3.79519999e-01 -1.08090627e+00 -7.49425232e-01
1.17463458e+00 1.70879334e-01 8.44125092e-01 8.54091108e-01
4.46003169e-01 -5.12851626e-02 -6.33584023e-01 1.26939809e+00
1.86695182e+00 1.82058915e-01 9.65366542e-01 3.53445232e-01
1.29788250e-01 1.33398205e-01 8.29526365e-01 3.11101884e-01
-1.94404468e-01 8.96969736e-01 6.66025639e-01 5.16532421e-01
-1.92499772e-01 4.77070332e-01 3.07101667e-01 7.85920620e-01
-6.17049158e-01 -5.85878752e-02 -4.93347824e-01 4.51543853e-02
-8.91248763e-01 -9.32400823e-01 -8.62649739e-01 2.81506133e+00
5.63659444e-02 -4.63559851e-02 -8.30337405e-01 -2.82470435e-01
-1.93942860e-01 -1.59339845e-01 -1.21196859e-01 3.02842762e-02
3.36715430e-01 3.70395303e-01 1.04072917e+00 1.37463021e+00
-6.97822392e-01 7.36511409e-01 6.17250061e+00 -3.64768416e-01
-1.41371882e+00 -8.47418085e-02 -4.53437626e-01 3.59325483e-02
-9.84050333e-01 3.23181540e-01 -1.07444906e+00 1.71151161e-01
7.35147893e-01 -1.48125440e-01 3.98726732e-01 3.84590805e-01
3.64178985e-01 -5.25782883e-01 -5.87437391e-01 9.10536468e-01
-4.90615457e-01 -1.45730746e+00 -4.32467967e-01 4.37101543e-01
7.67065048e-01 4.37685102e-01 2.35856846e-02 -3.59163880e-01
-9.45921168e-02 -5.33558369e-01 4.41046655e-02 9.53208983e-01
9.66478229e-01 -4.44669165e-02 2.91326195e-01 6.78658783e-01
-9.59504724e-01 1.87746227e-01 -8.47636104e-01 -6.56486869e-01
2.58762419e-01 1.35919952e+00 -1.58018041e+00 3.03653240e-01
4.00589228e-01 -2.51729161e-01 -1.31795615e-01 1.13139760e+00
-3.77219170e-01 6.04569733e-01 -4.07173514e-01 1.21696196e-01
1.55101448e-01 -6.63968742e-01 1.20735896e+00 9.36100930e-02
9.64408457e-01 6.50367558e-01 -2.83257455e-01 6.88549638e-01
4.67042858e-03 -5.68880737e-01 -7.24797130e-01 2.02277210e-02
2.97806114e-01 1.39695728e+00 -4.51240033e-01 1.03523985e-01
-1.06939189e-01 7.43535280e-01 -3.90374511e-01 2.04471529e-01
-2.77695596e-01 -2.93464005e-01 1.01961362e+00 8.74500275e-01
-2.32179537e-01 -1.16031706e+00 -2.86237717e-01 -8.55339766e-01
2.07050428e-01 2.26140037e-01 -1.47700027e-01 -9.60790932e-01
-4.46235597e-01 5.62676907e-01 -6.39616549e-01 -1.26644456e+00
-2.46436760e-01 -9.33772504e-01 -7.56467998e-01 8.82531345e-01
-1.51739335e+00 -9.47412848e-01 -6.31594479e-01 9.25523937e-02
1.37122884e-01 -5.87606914e-02 1.12490082e+00 -3.13970745e-01
4.13073868e-01 -1.93707183e-01 4.26405787e-01 -5.94782531e-01
7.87623286e-01 -1.09611392e+00 7.95091391e-01 9.75991368e-01
-1.16861902e-01 2.40904480e-01 1.28205681e+00 -8.78666759e-01
-1.91009963e+00 -1.22021413e+00 1.04736590e+00 -6.72606170e-01
2.32042402e-01 -2.51014531e-01 -4.11779851e-01 1.13569069e+00
-5.90038858e-02 -1.52032733e-01 3.10010076e-01 -6.50109276e-02
1.62255600e-01 -3.11820209e-01 -1.26091230e+00 9.47264805e-02
1.12039435e+00 -3.45811546e-01 -6.35647118e-01 9.92723584e-01
7.69628048e-01 -4.33080137e-01 -4.69731241e-01 6.43450320e-01
6.11548424e-01 -1.28361189e+00 4.00928587e-01 -2.11987093e-01
-1.03647739e-01 -2.75545418e-01 1.17949523e-01 -1.49360073e+00
-8.93165290e-01 -1.40636921e+00 4.31443185e-01 1.73652083e-01
8.18289444e-02 -1.62252331e+00 1.57879508e+00 5.57539724e-02
-9.03738439e-01 1.12735480e-01 -1.26317203e+00 -1.18471313e+00
-3.54747534e-01 -7.09562376e-02 9.64984223e-02 6.13178551e-01
-2.12373515e-03 4.19372767e-01 8.15148205e-02 1.05145371e+00
9.81499970e-01 6.15410924e-01 1.03892338e+00 -1.87616014e+00
-6.09757006e-01 3.48602422e-02 -6.06002808e-01 -1.38071489e+00
-8.57126474e-01 -5.68392813e-01 9.69953537e-02 -1.41490364e+00
-5.08909881e-01 -6.47691727e-01 5.71638465e-01 -6.23143315e-01
4.65563834e-01 3.91020209e-01 -1.90593049e-01 5.02638221e-01
1.22458644e-01 1.92451581e-01 1.42093003e+00 3.42569143e-01
-4.24672842e-01 1.06312919e+00 -2.75286764e-01 9.32864010e-01
8.80152583e-01 -4.85656530e-01 -1.98033690e-01 -5.65082431e-01
1.04997158e+00 1.90260172e-01 2.37421080e-01 -1.61242759e+00
5.29132366e-01 -5.02163827e-01 6.57566935e-02 -3.12232018e-01
5.55954754e-01 -4.94201988e-01 5.26143909e-01 8.19265127e-01
7.26482511e-01 -5.83080888e-01 -2.09660217e-01 2.97521144e-01
7.10194707e-02 -2.78966576e-01 1.35646534e+00 -5.99909902e-01
-2.65201777e-01 4.48293447e-01 -4.31169838e-01 -6.03254475e-02
8.82111311e-01 -1.05416998e-01 -7.49134600e-01 -2.30139524e-01
-1.09168433e-01 -4.29113925e-01 9.99019563e-01 -2.18850255e-01
7.72500575e-01 -1.07437217e+00 -6.03474021e-01 4.69451249e-01
2.16336310e-01 3.76523584e-02 1.69671252e-01 8.26024354e-01
-1.14138782e+00 1.23919964e+00 5.10869846e-02 -3.78676981e-01
-1.35498643e+00 -1.25115395e-01 9.28657115e-01 -6.40353039e-02
-9.31896031e-01 1.52328980e+00 1.55999765e-01 -1.07788824e-01
-7.18061924e-01 -6.86114669e-01 3.74431074e-01 -8.81949663e-01
7.59683132e-01 2.89954189e-02 2.98206687e-01 1.03570081e-01
-2.38537058e-01 8.39922369e-01 3.23770463e-01 -1.04248941e-01
1.22398436e+00 -1.15166165e-01 -1.66184500e-01 4.41758424e-01
5.22946894e-01 8.66951287e-01 -1.11425924e+00 -6.98739961e-02
-2.05332577e-01 -8.26779008e-01 1.85561135e-01 -4.28636312e-01
-1.43280581e-01 6.04118168e-01 4.31641221e-01 3.63307357e-01
8.89362752e-01 2.13320524e-01 9.57865715e-01 6.86838984e-01
9.03836012e-01 -7.74458051e-01 9.32491869e-02 5.78881562e-01
9.63054359e-01 -9.94505525e-01 3.66203099e-01 -7.11471736e-01
5.43771565e-01 1.30411148e+00 1.29153982e-01 -1.59328818e-01
7.91414499e-01 1.14367768e-01 2.55515099e-01 -9.17803347e-01
-6.20422363e-01 1.14125505e-01 4.16809916e-01 2.84690708e-01
2.91983545e-01 2.77672380e-01 -2.69771039e-01 -3.12556803e-01
-7.64054179e-01 -2.79130340e-01 1.01984119e+00 7.88826883e-01
-1.81625950e+00 -1.06227064e+00 -8.81708086e-01 4.60742742e-01
8.34066235e-03 1.20454468e-01 -1.66119367e-01 1.93703860e-01
-6.48128241e-02 7.89355218e-01 2.28956670e-01 -1.49665698e-01
3.22978437e-01 -6.30036741e-03 1.11824584e+00 -1.02993298e+00
9.05805305e-02 -3.32907379e-01 -7.26352620e-04 -4.51491803e-01
-9.19640660e-02 -3.36784422e-01 -1.13116705e+00 -3.28562781e-02
-3.24230582e-01 3.65306765e-01 1.06475949e+00 1.00898600e+00
3.42544079e-01 5.35860769e-02 1.02507222e+00 -6.40966654e-01
-6.17632151e-01 -1.34033763e+00 -9.87879634e-01 -2.76984096e-01
3.50313991e-01 -3.57167304e-01 -7.85797179e-01 -4.43514526e-01]
|
[9.757906913757324, -2.716773271560669]
|
2f087010-6210-412d-a170-10589dbcfc00
|
constructing-dreams-using-generative-ai
|
2305.12013
| null |
https://arxiv.org/abs/2305.12013v1
|
https://arxiv.org/pdf/2305.12013v1.pdf
|
Constructing Dreams using Generative AI
|
Generative AI tools introduce new and accessible forms of media creation for youth. They also raise ethical concerns about the generation of fake media, data protection, privacy and ownership of AI-generated art. Since generative AI is already being used in products used by youth, it is critical that they understand how these tools work and how they can be used or misused. In this work, we facilitated students' generative AI learning through expression of their imagined future identities. We designed a learning workshop - Dreaming with AI - where students learned about the inner workings of generative AI tools, used text-to-image generation algorithms to create their imaged future dreams, reflected on the potential benefits and harms of generative AI tools and voiced their opinions about policies for the use of these tools in classrooms. In this paper, we present the learning activities and experiences of 34 high school students who engaged in our workshops. Students reached creative learning objectives by using prompt engineering to create their future dreams, gained technical knowledge by learning the abilities, limitations, text-visual mappings and applications of generative AI, and identified most potential societal benefits and harms of generative AI.
|
['Cynthia Breazeal', 'Prerna Ravi', 'Randi Williams', 'Daniella DiPaola', 'Safinah Ali']
|
2023-05-19
| null | null | null | null |
['prompt-engineering']
|
['natural-language-processing']
|
[ 2.82125652e-01 1.02772117e+00 1.00187138e-01 7.89339095e-02
-1.08033791e-01 -7.88442552e-01 8.15974295e-01 -1.62377745e-01
1.24546610e-01 6.52016103e-01 6.37195468e-01 -3.09861720e-01
1.20831430e-01 -8.34722281e-01 -8.35268021e-01 -3.77567202e-01
4.96723801e-01 3.19035977e-01 -1.52756274e-02 -4.40778434e-01
5.96094906e-01 4.66852844e-01 -1.85014892e+00 4.00192261e-01
9.72858787e-01 2.92701930e-01 -3.76446515e-01 4.84533012e-01
-5.12901306e-01 1.46410584e+00 -1.30212140e+00 -8.98574710e-01
-1.56989530e-01 -8.04275572e-01 -8.25203598e-01 -2.70989649e-02
4.87141937e-01 -6.77334666e-01 2.13576928e-02 1.07710862e+00
2.85176992e-01 -8.92659798e-02 1.62959233e-01 -1.62989902e+00
-1.32061267e+00 8.39017689e-01 -1.64122194e-01 -1.07628815e-01
6.10622525e-01 4.76716489e-01 -1.55902833e-01 -4.22153801e-01
1.11350727e+00 1.16923845e+00 4.70684946e-01 1.07839668e+00
-1.13064587e+00 -1.45469677e+00 -4.30154830e-01 -2.06620887e-01
-1.21889412e+00 -4.57259923e-01 8.39613438e-01 -9.38683152e-01
6.73355103e-01 4.67574477e-01 1.91431379e+00 1.36425972e+00
2.20418692e-01 7.17238963e-01 1.16309655e+00 -4.24080938e-01
1.67787328e-01 1.08647764e+00 -3.87398601e-01 7.42787898e-01
2.55469322e-01 -2.71051109e-01 -8.45664144e-01 1.50480807e-01
6.80438459e-01 -4.36861604e-01 3.21388602e-01 8.04020986e-02
-9.22850370e-01 9.07333195e-01 -2.75257062e-02 6.73779368e-01
-1.03622749e-01 5.48109055e-01 -3.68376188e-02 3.68780643e-02
7.97303915e-01 5.64748585e-01 5.62308431e-01 -7.81450570e-01
-3.94590646e-01 2.02327251e-01 4.99555737e-01 8.32691789e-01
2.31084853e-01 3.47489148e-01 1.09922051e-01 3.60111326e-01
4.46282357e-01 3.31709892e-01 3.44415337e-01 -1.13494563e+00
-1.83882207e-01 8.56364429e-01 -2.59957433e-01 -1.07443058e+00
2.30619967e-01 -6.14864640e-02 1.49574243e-02 7.09400654e-01
2.43337005e-01 -4.51699615e-01 -7.46841669e-01 1.26962113e+00
5.68423986e-01 1.59584180e-01 7.58493915e-02 6.35196507e-01
1.27237153e+00 6.14779234e-01 2.52776742e-01 -7.36472756e-02
1.11985683e+00 -3.32279116e-01 -1.30570161e+00 -8.98858309e-02
7.16873229e-01 -9.85556841e-01 1.00513935e+00 4.28887397e-01
-1.41501427e+00 1.64218806e-02 -1.02402723e+00 -1.53052181e-01
-6.64998055e-01 -5.06231368e-01 7.90490985e-01 1.67908549e+00
-8.81204307e-01 5.89180827e-01 -5.09434760e-01 -5.68342924e-01
9.46088493e-01 2.43947446e-01 -7.19238371e-02 2.54691303e-01
-6.33076489e-01 8.37670743e-01 1.84229955e-01 -1.44372776e-01
-6.20702744e-01 -1.21352458e+00 -4.27921891e-01 -4.75677282e-01
1.80279613e-02 -5.44128835e-01 1.16135347e+00 -1.22190404e+00
-1.81431603e+00 1.02717125e+00 8.18498313e-01 -8.96127373e-02
6.66687131e-01 -4.46309924e-01 -4.79413360e-01 6.41093850e-02
2.64309049e-01 8.50223422e-01 4.76146847e-01 -1.24798954e+00
-2.19195798e-01 -2.16609985e-01 -1.03994720e-02 1.35945873e-02
-6.32489204e-01 1.05438545e-01 5.53860128e-01 -3.94717276e-01
-1.30484849e-01 -8.36085260e-01 2.67671824e-01 2.55272835e-01
-2.33757645e-01 6.05258457e-02 1.35419810e+00 -5.48621178e-01
7.05498874e-01 -2.31885076e+00 -6.51225567e-01 1.75288573e-01
4.56895292e-01 3.99008423e-01 4.66603667e-01 1.04980791e+00
5.35306692e-01 7.59933591e-01 5.65940738e-01 1.75487190e-01
5.23070283e-02 1.17518201e-01 -1.28589109e-01 3.56337249e-01
-3.43578100e-01 4.90045518e-01 -1.13525152e+00 -5.55179894e-01
3.19748938e-01 1.21407712e+00 -4.48546298e-02 3.42203602e-02
-4.30585407e-02 5.59255958e-01 -5.07736683e-01 3.82696211e-01
6.23040557e-01 2.97086537e-01 7.00835958e-02 3.47104967e-01
-6.12582147e-01 2.98294693e-01 -6.19615197e-01 1.41896987e+00
-4.04944211e-01 1.29831707e+00 -1.29063472e-01 -7.02591762e-02
9.88982499e-01 5.21049917e-01 2.30413035e-01 -7.74706185e-01
1.50396392e-01 6.19447492e-02 -2.43161649e-01 -1.02416575e+00
3.09279621e-01 -2.43758500e-01 2.01551244e-01 9.23083484e-01
-1.02230795e-01 -8.94524813e-01 -3.21546048e-01 2.96283782e-01
9.11237955e-01 5.45618176e-01 -2.36355811e-01 -8.19849819e-02
-2.63969660e-01 1.43781275e-01 -7.80040771e-02 2.74212033e-01
1.21417060e-01 3.19099694e-01 3.90826702e-01 -7.18245149e-01
-7.05347955e-01 -1.09125161e+00 3.34337950e-01 1.00620747e+00
-2.46664155e-02 -4.31397974e-01 -1.01681757e+00 -4.95425522e-01
-4.66365963e-01 1.46939766e+00 -3.85419607e-01 -4.82505471e-01
-1.81682825e-01 6.68211505e-02 6.66911066e-01 -1.73305154e-01
4.25014198e-01 -1.36355507e+00 -1.34097552e+00 1.92781821e-01
1.15922108e-01 -6.97237074e-01 -1.52932540e-01 -7.27368355e-01
-4.61081773e-01 -5.93791366e-01 -4.08804536e-01 -7.38818705e-01
7.65472829e-01 8.13227296e-02 6.05350792e-01 1.33561924e-01
-5.38331270e-01 7.11532176e-01 -2.93917418e-01 -1.11367559e+00
-1.02000380e+00 -5.54570973e-01 -4.41774875e-01 -3.24903280e-01
1.93193138e-01 -1.23106658e+00 -4.33244050e-01 -3.52125257e-01
-9.19292748e-01 9.02367711e-01 3.21124107e-01 3.35761458e-02
-2.01060504e-01 -1.44403409e-02 2.98390478e-01 -1.18194318e+00
8.89011204e-01 -2.93771535e-01 -1.39157355e-01 6.86993524e-02
-4.87874806e-01 -2.79985815e-01 7.24159330e-02 -8.27784002e-01
-1.16589904e+00 -2.27224529e-01 1.12474756e-02 -9.83742997e-02
-1.62083834e-01 -2.48522535e-01 -1.09344192e-01 -4.70823824e-01
8.02500665e-01 -5.79042472e-02 1.37337118e-01 7.52446102e-03
3.28840733e-01 6.57483578e-01 2.76709020e-01 -4.16124910e-01
7.60201752e-01 3.90930533e-01 -1.71928704e-01 -1.21623647e+00
-3.04170400e-01 4.77163494e-01 6.98403269e-02 -1.14582396e+00
9.89778578e-01 -6.33417368e-01 -1.18843901e+00 2.91761935e-01
-9.98953402e-01 -3.51959676e-01 -9.12016988e-01 4.70491648e-01
-3.09629858e-01 -4.70253974e-01 -2.72015750e-01 -1.25374711e+00
-3.46216172e-01 -6.18706048e-01 3.42441529e-01 6.01116717e-01
-6.37635469e-01 -8.60536218e-01 -3.06008775e-02 1.14964259e+00
5.49300611e-01 1.17104638e+00 7.07627118e-01 -2.31773749e-01
-5.21921456e-01 -3.99950966e-02 3.52498055e-01 8.59095622e-03
-4.60276417e-02 3.01162452e-01 -1.08313763e+00 3.62533182e-01
-1.63233832e-01 -5.39314568e-01 -1.71932712e-01 -4.04487938e-01
6.24114573e-01 -1.38830233e+00 -1.25000864e-01 -8.36731028e-03
1.29863119e+00 7.06506312e-01 7.57749379e-01 4.15102839e-01
6.89738035e-01 9.14874911e-01 1.96153075e-01 3.76816899e-01
2.52822965e-01 3.03288400e-02 2.12124556e-01 3.63361001e-01
-4.24038887e-01 -8.43564391e-01 5.97243190e-01 7.91535199e-01
-1.87399492e-01 8.71604383e-02 -1.01717615e+00 7.86978245e-01
-1.51690936e+00 -9.06594574e-01 -5.89259803e-01 1.73491621e+00
6.54045641e-01 2.72416532e-01 3.90416771e-01 -2.83672218e-03
3.97500247e-01 -1.77698553e-01 -6.02303445e-02 -1.15340948e+00
3.16526115e-01 4.94721770e-01 1.57741219e-01 2.02210352e-01
-3.52808744e-01 8.69682968e-01 6.22602224e+00 6.50366783e-01
-9.86191690e-01 4.87026006e-01 6.11775041e-01 -3.88442665e-01
-1.08745456e+00 7.78520182e-02 -2.22814023e-01 3.21625680e-01
6.97227716e-01 -4.56832737e-01 3.88401508e-01 5.51420987e-01
-6.10616319e-02 -1.15413181e-01 -6.40441656e-01 5.18073380e-01
2.53985465e-01 -1.80257285e+00 -1.10151805e-01 6.03146665e-02
9.73187506e-01 -6.38418138e-01 3.72041404e-01 -2.56642759e-01
7.14804351e-01 -1.20227778e+00 1.03803408e+00 3.48064750e-01
6.08942270e-01 -1.04418671e+00 2.64820810e-02 1.02434844e-01
-2.20932201e-01 2.53558546e-01 2.93382794e-01 -5.75541854e-01
-2.08037749e-01 -1.73475984e-02 -1.32136261e+00 -4.47567701e-01
5.46588242e-01 -6.57469630e-02 -4.66719478e-01 5.50638318e-01
-3.74207914e-01 6.23569727e-01 -4.34211016e-01 -6.02896273e-01
8.62751305e-02 -3.05097491e-01 4.82105553e-01 9.93057847e-01
5.44183969e-01 2.32756823e-01 -7.55836487e-01 1.29563892e+00
2.68521816e-01 4.11636271e-02 -1.04476202e+00 -9.04837251e-01
3.81782591e-01 1.07923937e+00 -9.41279829e-01 3.55978459e-01
-4.42890003e-02 6.42428398e-01 -4.56162691e-01 1.24584235e-01
-5.78004479e-01 -2.68248469e-01 6.82475567e-01 1.13249564e+00
-2.98577011e-01 4.49263752e-02 -3.62739295e-01 -3.34721416e-01
-4.17363733e-01 -7.96472669e-01 -4.02302563e-01 -1.13548028e+00
-8.71208906e-01 1.87650263e-01 2.36561269e-01 -6.87283218e-01
-3.32284719e-01 1.65330231e-01 -9.42053378e-01 1.62807688e-01
-4.35500443e-02 -1.76127946e+00 -1.78786427e-01 1.22137122e-01
2.98299819e-01 -5.75051224e-03 6.44416988e-01 -2.68067233e-02
-1.83585271e-01 5.06741345e-01 -1.35404184e-01 -2.15492547e-01
2.63838768e-02 -5.62438548e-01 2.47470856e-01 4.87175375e-01
3.17723393e-01 3.11564922e-01 1.07099819e+00 -8.33990753e-01
-1.70538092e+00 -3.80194038e-01 7.01669157e-01 -1.11084819e-01
6.25280261e-01 -5.18356025e-01 -3.42836767e-01 8.93853009e-01
7.63176799e-01 -5.74159563e-01 1.04074180e+00 -4.64268535e-01
9.31695625e-02 1.39389440e-01 -1.82542586e+00 8.30610096e-01
1.30240178e+00 -2.82942742e-01 -4.35662299e-01 6.05797291e-01
7.06913114e-01 -3.12303662e-01 -7.42918253e-01 -5.29288948e-01
9.45791364e-01 -1.06732106e+00 6.84987247e-01 -6.35075942e-02
6.14042819e-01 1.81690544e-01 7.17639387e-01 -9.88790214e-01
2.59033442e-01 -1.30647218e+00 3.47866744e-01 1.97840905e+00
2.03373179e-01 -5.78855217e-01 1.02775764e+00 1.17708254e+00
-1.98244229e-01 -1.59715891e-01 -9.33545709e-01 -2.04008222e-01
9.67427120e-02 -4.59301382e-01 6.70050144e-01 1.32983840e+00
3.73651743e-01 4.84245904e-02 -4.17989403e-01 -2.27713794e-01
5.23948431e-01 -5.21968067e-01 8.49975586e-01 -1.02201462e+00
1.30772084e-01 -4.12044674e-01 -7.05277205e-01 4.00624126e-01
-3.15329313e-01 -5.31976819e-01 -3.92141432e-01 -1.77026045e+00
-1.75699189e-01 1.67510450e-01 8.51099253e-01 4.37119246e-01
6.66481853e-01 6.19584978e-01 7.10716307e-01 -1.49614647e-01
-1.96480587e-01 -1.11940969e-02 1.81191826e+00 2.58724123e-01
-4.95706856e-01 -4.37798858e-01 -8.68306100e-01 8.60347569e-01
8.02921772e-01 -5.71065903e-01 -8.10176253e-01 -2.10428163e-01
9.88579869e-01 -2.56854564e-01 3.68285537e-01 -1.35068679e+00
1.24843873e-01 -5.81974447e-01 4.09527510e-01 -3.20288479e-01
5.20116270e-01 -9.68922496e-01 1.09936213e+00 5.62408864e-01
-2.39829525e-01 -5.09042025e-01 4.20474201e-01 -4.73101825e-01
4.03271675e-01 -3.64143908e-01 3.05565983e-01 -2.69692779e-01
3.36512886e-02 -6.70243502e-01 -1.09367943e+00 6.32603988e-02
1.64679396e+00 -7.09878922e-01 -4.79765236e-01 -8.75821352e-01
-7.49619722e-01 -2.57749587e-01 5.56177020e-01 4.81721967e-01
8.67025018e-01 -1.36352289e+00 -3.16415906e-01 3.72276902e-01
-1.84126213e-01 -1.04321964e-01 1.59467548e-01 2.61634886e-01
-1.06230891e+00 -3.34901273e-01 -8.93851101e-01 1.51151061e-01
-1.36946774e+00 3.86820793e-01 -1.30621240e-01 6.02741599e-01
-7.40858793e-01 9.80522811e-01 1.08396955e-01 9.20337662e-02
-7.91260153e-02 1.92294285e-01 -9.43481848e-02 6.49433509e-02
6.03221238e-01 9.62770879e-01 -6.46934450e-01 -5.91822684e-01
-9.53436270e-02 2.96638161e-01 2.19281271e-01 -3.91618311e-01
1.49467242e+00 3.13380696e-02 -4.03125919e-02 7.98642516e-01
8.39700818e-01 3.33925575e-01 -9.51260686e-01 7.26871967e-01
-4.55043912e-01 -8.85556459e-01 -2.34477848e-01 -1.10245466e+00
-1.02082133e+00 8.55073333e-01 6.75846875e-01 5.01220584e-01
8.74613702e-01 1.04856148e-01 8.70147824e-01 -3.44169945e-01
-1.26536518e-01 -1.21110654e+00 4.54813629e-01 -3.45873266e-01
1.08043826e+00 -4.26747531e-01 3.07613790e-01 -6.65062606e-01
-7.08289266e-01 1.09290528e+00 8.55644107e-01 1.78979248e-01
4.06046003e-01 4.14395303e-01 3.30842696e-02 -5.35539091e-01
-3.99000645e-01 2.33506843e-01 9.28594247e-02 1.21200895e+00
4.97433186e-01 1.49737984e-01 -8.15626621e-01 9.09313187e-02
-9.02864933e-01 5.42197645e-01 1.06103408e+00 1.14085400e+00
-2.49351993e-01 -9.31848228e-01 -7.59413600e-01 -4.10564207e-02
-7.40291655e-01 5.13300121e-01 -1.31200123e+00 1.08229768e+00
8.77208829e-01 9.25605953e-01 3.95434201e-02 -6.43160701e-01
-1.18904868e-02 2.11083218e-01 6.75943851e-01 -3.41125935e-01
-1.23032141e+00 -2.86749721e-01 5.20720959e-01 -9.51981768e-02
-1.07520007e-01 -6.18637621e-01 -1.10453212e+00 -1.05077171e+00
-7.35227615e-02 -6.84393197e-02 1.22647178e+00 5.73780179e-01
4.69684958e-01 3.23721677e-01 2.90532243e-02 -1.99106157e-01
5.47856152e-01 -7.47170091e-01 -1.13145731e-01 1.19786128e-01
3.92044485e-02 -3.80270779e-02 -5.03109433e-02 -9.69446376e-02]
|
[9.43839168548584, 6.417750358581543]
|
a762f08e-0ecc-4277-804e-d021c5919a5d
|
190600901
|
1906.00901
| null |
https://arxiv.org/abs/1906.00901v2
|
https://arxiv.org/pdf/1906.00901v2.pdf
|
The iMet Collection 2019 Challenge Dataset
|
Existing computer vision technologies in artwork recognition focus mainly on instance retrieval or coarse-grained attribute classification. In this work, we present a novel dataset for fine-grained artwork attribute recognition. The images in the dataset are professional photographs of classic artworks from the Metropolitan Museum of Art, and annotations are curated and verified by world-class museum experts. In addition, we also present the iMet Collection 2019 Challenge as part of the FGVC6 workshop. Through the competition, we aim to spur the enthusiasm of the fine-grained visual recognition research community and advance the state-of-the-art in digital curation of museum collections.
|
['Christine Kaeser-Chen', 'Serge Belongie', 'Jennie Choi', 'Chenyang Zhang', 'Maria Kessler', 'Grace Vesom']
|
2019-06-03
| null | null | null | null |
['fine-grained-visual-recognition']
|
['computer-vision']
|
[ 3.44539911e-01 -4.66102362e-01 -2.25270927e-01 -4.29501981e-01
-5.16743720e-01 -7.73316801e-01 9.03264701e-01 -2.20818013e-01
-3.28278929e-01 4.04973090e-01 3.09889466e-01 3.71417552e-01
-3.83871794e-01 -7.00680673e-01 -6.30436063e-01 -3.01899463e-01
4.55105603e-01 7.68632472e-01 2.66631860e-02 1.16382740e-01
4.98744518e-01 8.74362469e-01 -1.94152558e+00 7.93226898e-01
3.76655191e-01 1.28791332e+00 -2.77035590e-02 7.38899708e-01
-3.09566945e-01 8.22881937e-01 -4.89425540e-01 -8.38070929e-01
4.11914825e-01 1.98605835e-01 -1.30305445e+00 5.62365353e-01
1.61943328e+00 -2.53045201e-01 -5.61056733e-01 1.01416433e+00
2.82725275e-01 1.68953300e-01 9.10116315e-01 -1.29457450e+00
-9.93018091e-01 2.18499601e-01 -2.09536925e-01 2.43266046e-01
1.80265695e-01 1.88098535e-01 1.19686151e+00 -1.01862490e+00
1.00617874e+00 1.31889439e+00 7.86921918e-01 3.84116620e-01
-1.18272698e+00 -5.04164517e-01 3.09107691e-01 6.63307846e-01
-2.00835109e+00 -5.35362720e-01 7.29517817e-01 -6.41905367e-01
7.94577658e-01 3.79965752e-01 8.51957381e-01 1.06026781e+00
-5.25541604e-01 1.09179568e+00 1.05862582e+00 -3.93749863e-01
1.00635022e-01 -3.15551698e-01 8.78750607e-02 7.38793433e-01
-7.48257758e-03 -2.01512307e-01 -7.43704975e-01 -1.49889514e-01
1.24395192e+00 2.81117618e-01 1.35007324e-02 -2.76900232e-01
-1.41866791e+00 4.66471016e-01 3.78962547e-01 5.06673574e-01
-5.31573474e-01 3.24765325e-01 3.29114616e-01 2.31403142e-01
2.83971637e-01 3.92653018e-01 -8.06297958e-02 -4.14929330e-01
-1.08358133e+00 1.68150172e-01 6.79125965e-01 1.27085602e+00
7.61218131e-01 -3.15126106e-02 -3.10941130e-01 1.15170872e+00
3.21427226e-01 5.27293622e-01 1.86418332e-02 -1.13139927e+00
2.28066817e-01 9.45987225e-01 -3.33302945e-01 -6.59829855e-01
5.14684469e-02 -3.60418297e-02 -8.25716555e-01 2.22155437e-01
4.71948624e-01 8.87112796e-01 -1.29113877e+00 7.00482130e-01
7.25407004e-02 3.41490060e-01 -6.13981128e-01 9.48423922e-01
1.03429544e+00 1.19772470e-02 4.61651891e-01 4.65047956e-01
1.38407707e+00 -8.51505280e-01 -4.72410887e-01 -1.37290835e-01
-1.03145897e-01 -1.10271668e+00 1.28603458e+00 2.81183004e-01
-8.76038849e-01 -6.21282339e-01 -8.61625195e-01 -2.32727885e-01
-4.96956766e-01 2.05561131e-01 6.04543686e-01 3.64261031e-01
-8.42717528e-01 4.99436110e-01 -4.03055757e-01 -5.94464779e-01
1.06845570e+00 1.12178639e-01 -5.85658550e-01 -3.47414911e-01
-5.36866188e-01 6.62890613e-01 1.55156389e-01 7.30734840e-02
-1.00106180e+00 -9.95228946e-01 -5.04543304e-01 -2.68088102e-01
1.23224564e-01 -7.81789243e-01 1.00235701e+00 -6.92332268e-01
-8.16306889e-01 1.81450975e+00 1.84310377e-01 -1.06433630e-01
3.03728491e-01 -1.02062501e-01 -4.88664299e-01 2.05755621e-01
1.89838201e-01 5.76059997e-01 9.91963863e-01 -1.23016083e+00
-8.55693579e-01 -6.00084782e-01 8.60517174e-02 -2.12383941e-01
-3.54052305e-01 -5.33932354e-03 -6.82614386e-01 -9.94044602e-01
-6.51430637e-02 -1.16590393e+00 2.00249791e-01 4.18115526e-01
-3.73783618e-01 -4.50431496e-01 8.99238169e-01 -2.81375110e-01
8.16681683e-01 -2.32848859e+00 -4.37302478e-02 1.34366930e-01
1.98643595e-01 2.07768336e-01 -2.11917564e-01 2.00863779e-01
2.41498560e-01 1.97396725e-01 3.06818690e-02 -4.71279323e-01
4.26812798e-01 4.18035686e-01 -3.79023761e-01 5.26358426e-01
-4.79470789e-02 9.10622656e-01 -7.92154670e-01 -9.09792602e-01
6.65478885e-01 5.01514852e-01 4.76316549e-02 2.30794460e-01
-2.51925647e-01 1.90536633e-01 -7.18697488e-01 1.43610907e+00
4.91502196e-01 -2.44854137e-01 -2.68219560e-01 -6.12773418e-01
1.00113396e-02 -1.43662617e-01 -1.25721467e+00 2.02073669e+00
-3.48711729e-01 8.00727189e-01 7.98665136e-02 -2.98299640e-01
8.92032087e-01 8.45055878e-02 4.03205067e-01 -4.47967678e-01
-6.73268512e-02 3.61266881e-02 -7.21551359e-01 -2.47151241e-01
7.16856122e-01 -4.38439660e-02 -4.37385701e-02 3.73285413e-01
1.29518792e-01 -6.01503849e-01 2.95990527e-01 4.15250063e-02
9.53880787e-01 3.11887681e-01 1.81283101e-01 -1.27684236e-01
5.50598025e-01 2.33509406e-01 1.12098970e-01 7.74803519e-01
-3.92622709e-01 8.70890796e-01 -3.70870590e-01 -1.01541519e+00
-1.37254965e+00 -1.30086422e+00 -3.62163037e-01 1.59779465e+00
-8.88677463e-02 -4.17369723e-01 -5.49684465e-01 -7.60440946e-01
2.76014060e-01 2.09493503e-01 -7.30591714e-01 2.90949434e-01
-3.69948775e-01 -1.77435204e-02 1.08149433e+00 6.03699923e-01
8.32969725e-01 -1.39372265e+00 -7.40142316e-02 -2.56291181e-01
-1.62716717e-01 -1.39776480e+00 -7.28443027e-01 -2.88186461e-01
-5.95880151e-01 -1.37734759e+00 -7.42654264e-01 -9.64337528e-01
4.72927988e-01 6.25517890e-02 1.54653883e+00 3.98283064e-01
-9.61836398e-01 9.92590487e-01 -4.09280896e-01 -3.62293512e-01
-5.32861538e-02 2.31522605e-01 -7.91178900e-04 3.49498749e-01
7.45252490e-01 -6.42785370e-01 -4.80917156e-01 4.93673831e-01
-7.56467342e-01 -5.23620963e-01 6.98671222e-01 5.75235248e-01
1.19392705e+00 -3.07509810e-01 8.09163600e-02 -5.73866785e-01
3.23863983e-01 1.93633139e-01 -4.38294142e-01 4.71043080e-01
-5.36118805e-01 3.43115255e-02 1.11325458e-01 -4.30228978e-01
-1.00843310e+00 4.70393419e-01 -1.27648577e-01 -9.42195117e-01
-8.62157822e-01 -1.83188304e-01 -2.29024634e-01 -5.55423260e-01
5.89076102e-01 3.10790688e-01 -6.93043828e-01 -8.27883303e-01
5.43092251e-01 9.46092606e-01 1.11952877e+00 -8.95823956e-01
7.54001141e-01 5.81364393e-01 1.86378449e-01 -1.03243542e+00
-1.02648318e+00 -8.11089218e-01 -9.80300426e-01 -5.31113565e-01
9.58658218e-01 -8.55394602e-01 -1.04793167e+00 4.60569292e-01
-9.08411801e-01 -1.39717981e-01 -7.39008605e-01 1.68364123e-01
-7.30298102e-01 2.72800624e-01 -5.18765867e-01 -5.91037512e-01
-5.32654703e-01 -7.33745933e-01 1.75349522e+00 7.52050281e-02
-1.86864406e-01 -7.21611679e-01 -9.28114057e-02 7.32142270e-01
2.57090867e-01 1.62818715e-01 6.23042226e-01 -2.28692189e-01
-9.62872088e-01 -3.26184064e-01 -3.37773085e-01 1.99033141e-01
-3.89056616e-02 1.55798465e-01 -1.30450714e+00 2.74315357e-01
-6.55619621e-01 -5.89910865e-01 1.09780860e+00 6.52795061e-02
1.39957523e+00 2.62347534e-02 -1.71901584e-01 8.70499611e-01
1.34573567e+00 -3.26764703e-01 8.61806631e-01 9.01345015e-01
1.18517458e+00 3.48119348e-01 5.10871172e-01 7.14076996e-01
4.57684129e-01 7.87431657e-01 4.87357765e-01 2.25815326e-01
-7.26731598e-01 -3.40436727e-01 -5.07204652e-01 3.84662300e-01
-8.15420151e-01 1.00837380e-01 -1.06257129e+00 7.83179879e-01
-1.80953634e+00 -1.17890501e+00 -3.73114526e-01 1.99191415e+00
7.60082424e-01 -2.79856861e-01 7.92682841e-02 9.38816145e-02
7.11641788e-01 2.53046930e-01 -1.43205419e-01 -1.12128921e-01
-4.63311225e-01 2.23699644e-01 5.62627792e-01 7.79560730e-02
-1.55757558e+00 1.19374120e+00 6.97654867e+00 1.10232067e+00
-5.56218088e-01 -1.48653492e-01 3.18705142e-01 -2.05194473e-01
-5.38475439e-02 -3.17464441e-01 -8.48361433e-01 1.66895300e-01
2.81990290e-01 1.19522125e-01 6.10322297e-01 1.05235672e+00
-4.93825912e-01 3.17914397e-01 -1.31609976e+00 1.45338929e+00
1.64076522e-01 -1.90084255e+00 2.63071537e-01 2.01403499e-01
7.66773224e-01 3.40723284e-02 -3.26574743e-02 1.13818139e-01
3.53499413e-01 -1.24099815e+00 6.94882452e-01 1.00008953e+00
1.02878976e+00 -6.36243522e-01 4.80627179e-01 -1.80453390e-01
-1.54554129e+00 2.86014050e-01 -4.20570374e-01 4.93648909e-02
-1.25416130e-01 2.97645330e-01 -5.35135508e-01 4.06224653e-02
1.14557040e+00 8.45091820e-01 -8.38365495e-01 1.14798832e+00
-3.90418798e-01 4.83310372e-01 -2.07459629e-01 4.80065262e-03
2.69966632e-01 8.14681500e-02 2.71391243e-01 1.41816843e+00
-2.82296419e-01 9.97126997e-02 3.91286016e-01 6.69556201e-01
-6.29967391e-01 -2.39163309e-01 -4.10981417e-01 -3.58035028e-01
4.65963215e-01 1.43022048e+00 -5.07324576e-01 -2.47167811e-01
-2.00898737e-01 1.11060929e+00 2.83334345e-01 1.94919378e-01
-1.42749101e-01 -1.81092426e-01 8.53133678e-01 1.05723396e-01
6.18124962e-01 -1.70510590e-01 -5.18972993e-01 -8.21725249e-01
-4.54675481e-02 -6.98339820e-01 5.82754850e-01 -9.59645689e-01
-2.00632048e+00 6.68657720e-01 -2.72240311e-01 -1.19830871e+00
-2.06914376e-02 -6.33636951e-01 -2.13199973e-01 6.30218387e-01
-1.28469193e+00 -2.02880359e+00 -7.56394446e-01 1.17106259e+00
7.08197176e-01 -3.96383226e-01 1.10041916e+00 6.10118628e-01
-6.48687631e-02 6.02553427e-01 1.68843456e-02 4.63756412e-01
8.00318420e-01 -1.26510394e+00 4.34900075e-01 3.25550705e-01
7.02635050e-01 4.36883330e-01 4.25691456e-01 -6.35355890e-01
-1.63748837e+00 -1.24313486e+00 6.76218510e-01 -8.68364155e-01
9.40313160e-01 -2.25658864e-01 -7.19373226e-01 7.40703225e-01
-8.50000978e-02 3.70994210e-01 5.68179607e-01 3.55598837e-01
-7.42574811e-01 3.87172103e-02 -1.11119998e+00 1.65482402e-01
1.28660226e+00 -1.01506341e+00 -7.86629260e-01 5.73531270e-01
1.48320168e-01 -9.94605348e-02 -1.50708163e+00 2.01096505e-01
9.07930911e-01 -1.07354559e-01 1.46689653e+00 -5.69559932e-01
3.99057388e-01 -3.78681004e-01 -5.74844003e-01 -8.15082788e-01
-5.42488098e-01 -1.78353846e-01 -9.63806883e-02 1.38872623e+00
-1.19581394e-01 3.50333929e-01 1.20905066e+00 6.12445951e-01
-2.96096563e-01 -1.85767978e-01 -1.02988410e+00 -1.03656745e+00
-3.16499770e-01 -6.29357159e-01 6.57213569e-01 9.10355091e-01
-5.89895368e-01 3.30356538e-01 -4.39034641e-01 -1.79542378e-01
1.15223205e+00 4.93747085e-01 9.76320148e-01 -1.60812628e+00
2.40409877e-02 -7.29792178e-01 -1.16294837e+00 -6.26377881e-01
1.46427825e-01 -7.34482348e-01 5.37588522e-02 -1.81766224e+00
3.01472217e-01 -2.88915515e-01 -9.48017910e-02 6.61328852e-01
1.35500193e-01 1.18447363e+00 4.97894794e-01 6.56976998e-01
-1.04190826e+00 4.00475889e-01 1.27855968e+00 -9.39649105e-01
3.30973625e-01 -1.01376995e-01 -2.42850915e-01 8.46271396e-01
3.87002230e-01 -3.81681025e-01 1.13564506e-01 -4.76052999e-01
-1.51412830e-01 -9.03602123e-01 7.39452124e-01 -8.76156986e-01
3.03509206e-01 -2.82980591e-01 6.56172335e-01 -7.69357443e-01
7.85342455e-01 -1.11655176e+00 1.97598442e-01 -6.83777034e-02
-2.01471016e-01 -4.15818214e-01 1.35196716e-01 5.23890853e-01
-3.50575686e-01 -2.30898097e-01 6.37164831e-01 -3.66164088e-01
-1.45428872e+00 6.95163429e-01 -1.39680743e-01 5.64170033e-02
8.06229174e-01 -4.11052823e-01 -5.90664685e-01 4.63353004e-03
-8.46499622e-01 -4.99005876e-02 6.78265274e-01 5.66962481e-01
6.72879934e-01 -1.74785316e+00 -9.53845739e-01 -6.72942847e-02
9.01549757e-01 -4.01229978e-01 2.84804463e-01 4.53028142e-01
-5.38490295e-01 3.22978199e-01 -5.34653842e-01 -6.19003534e-01
-1.50482821e+00 6.08133852e-01 5.90463243e-02 9.26004052e-02
-7.32083321e-01 9.74251568e-01 -1.55196294e-01 -2.75723815e-01
3.65682989e-01 1.72471851e-01 -7.95962289e-02 -1.04027867e-01
7.10582554e-01 5.11152625e-01 1.27227634e-01 -8.78141463e-01
-7.63613760e-01 9.49510813e-01 1.28199514e-02 2.05828130e-01
1.42841816e+00 4.57846969e-02 -2.91023683e-02 4.88685727e-01
9.23306406e-01 -3.02533150e-01 -1.12830615e+00 -4.01808232e-01
1.97185740e-01 -1.03802109e+00 2.47954875e-01 -9.48490918e-01
-9.45999920e-01 7.47689843e-01 8.10912609e-01 1.24976873e-01
1.14073229e+00 5.50237894e-01 5.66750228e-01 8.55368137e-01
5.74108958e-01 -1.41595292e+00 -1.07339278e-01 4.65746701e-01
1.10125828e+00 -1.23449957e+00 4.33438420e-01 -3.16046119e-01
-5.33288896e-01 1.23326671e+00 3.65860850e-01 -2.73301393e-01
5.04956543e-01 1.83540732e-01 -6.85712439e-04 -5.72604358e-01
-1.63923651e-01 -5.64334035e-01 7.51704276e-01 8.82762909e-01
3.33655119e-01 1.84469804e-01 9.10461470e-02 3.41391355e-01
-1.02799814e-02 2.00815842e-01 7.69994557e-02 9.48053479e-01
-5.21955311e-01 -1.12703931e+00 -5.54006100e-01 5.04140496e-01
-1.87218323e-01 -2.52824455e-01 -1.07502365e+00 7.31347620e-01
2.73912460e-01 7.70500541e-01 2.02680707e-01 -3.44776869e-01
7.93024004e-01 -1.30509883e-01 8.27797711e-01 -3.67880851e-01
-7.30626643e-01 -2.10064054e-01 4.08871561e-01 -6.55590117e-01
-6.14052355e-01 -6.14695191e-01 -6.76147819e-01 -7.63365924e-01
1.76094979e-01 -1.28338754e-01 9.63034451e-01 9.58237171e-01
2.43765712e-01 3.82495642e-01 1.44865960e-01 -9.65147376e-01
1.47641227e-01 -9.94532287e-01 -9.26015913e-01 7.13988781e-01
-3.03254575e-02 -6.93208575e-01 -2.81916279e-02 7.16175854e-01]
|
[11.327999114990234, 0.5163640379905701]
|
2df11ddf-751d-4d36-b79a-75838360a670
|
a-convolutional-spiking-network-for-gesture
|
2304.11106
| null |
https://arxiv.org/abs/2304.11106v2
|
https://arxiv.org/pdf/2304.11106v2.pdf
|
A Convolutional Spiking Network for Gesture Recognition in Brain-Computer Interfaces
|
Brain-computer interfaces are being explored for a wide variety of therapeutic applications. Typically, this involves measuring and analyzing continuous-time electrical brain activity via techniques such as electrocorticogram (ECoG) or electroencephalography (EEG) to drive external devices. However, due to the inherent noise and variability in the measurements, the analysis of these signals is challenging and requires offline processing with significant computational resources. In this paper, we propose a simple yet efficient machine learning-based approach for the exemplary problem of hand gesture classification based on brain signals. We use a hybrid machine learning approach that uses a convolutional spiking neural network employing a bio-inspired event-driven synaptic plasticity rule for unsupervised feature learning of the measured analog signals encoded in the spike domain. We demonstrate that this approach generalizes to different subjects with both EEG and ECoG data and achieves superior accuracy in the range of 92.74-97.07% in identifying different hand gesture classes and motor imagery tasks.
|
['Bipin Rajendran', 'Yiming Ai']
|
2023-04-21
| null | null | null | null |
['gesture-recognition']
|
['computer-vision']
|
[ 8.33808005e-01 -5.38630784e-01 3.20178539e-01 -2.09096119e-01
-3.56526762e-01 -4.80457842e-01 3.94572288e-01 -1.62771180e-01
-6.61115289e-01 9.74653602e-01 -3.38129997e-01 -1.92890540e-01
-4.66990471e-01 -4.43921447e-01 -4.63054001e-01 -9.39824820e-01
-1.53822735e-01 8.96419808e-02 8.31291527e-02 2.12044343e-01
6.04479134e-01 6.12462997e-01 -1.51682281e+00 2.56862909e-01
6.48744524e-01 1.47275925e+00 1.46792606e-01 5.63671708e-01
-1.02356717e-01 3.07176024e-01 -8.09843242e-01 2.80494422e-01
4.63985205e-02 -6.89289987e-01 -4.20973629e-01 -2.55951077e-01
-3.35497916e-01 5.49453460e-02 -2.52882063e-01 9.53877926e-01
7.57644594e-01 -1.33982539e-01 9.29171920e-01 -1.25863636e+00
-2.32657149e-01 3.28146219e-01 -2.16655910e-01 5.53458989e-01
3.86187732e-01 1.65209889e-01 1.23389572e-01 -6.28369391e-01
3.58634055e-01 3.83743167e-01 6.75924778e-01 5.59220433e-01
-1.35051334e+00 -1.11464345e+00 -5.64149141e-01 3.64516377e-01
-1.44622099e+00 -1.01308785e-01 7.26133943e-01 -6.37542367e-01
1.27102244e+00 1.49049863e-01 9.87461030e-01 1.36376297e+00
7.75711715e-01 4.41033751e-01 1.67228830e+00 -3.61797661e-01
7.00471640e-01 -1.33320555e-01 1.28667131e-01 1.21082641e-01
8.17760155e-02 1.47961527e-01 -7.89095342e-01 -2.61935264e-01
9.92101967e-01 1.20925985e-01 -4.53646183e-01 6.11185618e-02
-1.09456563e+00 4.13449258e-01 2.42781222e-01 6.52420878e-01
-8.87555778e-01 2.15727895e-01 2.36239895e-01 3.96604031e-01
-1.28075797e-02 7.82758176e-01 -3.07399333e-01 -6.41162694e-01
-1.15725827e+00 -6.39441758e-02 8.60443830e-01 5.24755299e-01
1.18669577e-01 -1.85291506e-02 -4.21117507e-02 5.24154127e-01
1.34355165e-02 4.77040142e-01 9.82079744e-01 -6.91861689e-01
-8.30721762e-03 4.77536798e-01 4.20034491e-03 -6.14286125e-01
-7.28421390e-01 6.11853739e-03 -8.38678896e-01 6.21061265e-01
6.17727339e-01 -1.13838762e-01 -8.27412784e-01 1.36191034e+00
-2.01448604e-01 3.93460542e-01 -1.13985285e-01 8.67900372e-01
5.59330285e-01 2.16061398e-01 2.43232250e-01 -2.43295327e-01
1.15881383e+00 -5.48678748e-02 -6.76063597e-01 -3.31322141e-02
1.44536644e-01 -1.23366497e-01 9.75587964e-01 7.76354194e-01
-8.88620853e-01 4.55904240e-03 -1.24030876e+00 4.97536361e-01
-5.68302870e-01 -1.58897247e-02 4.58719015e-01 6.75797045e-01
-8.95056605e-01 8.67532670e-01 -1.10347402e+00 -5.46350002e-01
5.06147385e-01 8.92123163e-01 -3.32444817e-01 5.27828217e-01
-9.19510603e-01 8.83147538e-01 2.79867589e-01 -9.75397378e-02
-2.59388179e-01 -5.64974129e-01 -2.43108034e-01 8.34207684e-02
-4.26176399e-01 -8.92550051e-02 8.49844456e-01 -9.23245549e-01
-1.92264152e+00 8.80093157e-01 1.19956344e-01 -4.28355992e-01
2.01751083e-01 2.11427733e-01 -4.10630673e-01 2.55327463e-01
-3.14229161e-01 4.25358534e-01 9.52172399e-01 -4.28297997e-01
-1.62552297e-01 -6.59835279e-01 -5.30411243e-01 -3.21175426e-01
-4.44451600e-01 1.89380631e-01 3.12707365e-01 -6.11943960e-01
2.54726380e-01 -9.49461102e-01 3.00003529e-01 -6.26597032e-02
-4.27918620e-02 2.42855237e-03 5.65665781e-01 -5.18658936e-01
9.68731701e-01 -2.25958419e+00 9.77807790e-02 5.84753156e-01
1.35183036e-01 1.14505962e-01 7.15494454e-02 3.76965404e-01
-1.95600420e-01 -2.76152998e-01 -4.54531729e-01 4.30907041e-01
-2.76939422e-01 -8.59500915e-02 -2.12563112e-01 4.14893627e-01
2.89415956e-01 1.15538621e+00 -7.38842547e-01 1.48999337e-02
2.70274431e-01 6.03496313e-01 7.11245239e-02 1.10762887e-01
4.38875943e-01 1.07806921e+00 -3.32992554e-01 6.29120886e-01
3.29874218e-01 -1.75330058e-01 1.96081460e-01 -1.97400711e-02
-1.46161422e-01 2.57768154e-01 -8.44539523e-01 1.79216862e+00
-3.74223262e-01 9.79484677e-01 -1.63987368e-01 -1.37987947e+00
1.00552738e+00 4.84746069e-01 5.89490235e-01 -1.10719836e+00
6.86004639e-01 6.18139565e-01 3.86917830e-01 -7.55008638e-01
-4.96712089e-01 -2.18738288e-01 6.25981763e-02 7.01486766e-01
3.12244684e-01 -1.74927697e-01 -2.03728423e-01 -4.50428814e-01
1.59807563e+00 -3.34432186e-03 4.22127187e-01 -5.19549727e-01
3.51665825e-01 -1.50262177e-01 -2.04738416e-02 5.87494433e-01
-2.24993780e-01 3.95485580e-01 1.83840245e-01 -2.53353834e-01
-6.68694437e-01 -1.03662503e+00 -3.74120235e-01 6.12804830e-01
1.05267175e-01 1.84133127e-01 -9.11345661e-01 1.16629809e-01
5.44425622e-02 4.81482208e-01 -5.53037882e-01 -4.06263471e-01
-4.61582839e-01 -7.60898829e-01 7.90388107e-01 9.84680891e-01
3.48452479e-01 -1.56852365e+00 -1.41393793e+00 6.67350888e-01
1.65628612e-01 -1.07113707e+00 2.16632679e-01 7.82427490e-01
-1.06870365e+00 -9.26914394e-01 -8.86276782e-01 -6.48396254e-01
3.63978922e-01 -5.07076263e-01 5.12756765e-01 -2.88063020e-01
-9.77578163e-01 5.45298636e-01 -2.42719933e-01 -6.34067059e-01
1.03103735e-01 -1.08746082e-01 2.15278372e-01 1.82556570e-01
1.00418091e+00 -1.18442369e+00 -5.43410838e-01 5.38155176e-02
-6.42401636e-01 -2.41191044e-01 7.02770233e-01 7.90579796e-01
5.03301859e-01 -3.29299957e-01 8.60203445e-01 -1.64311454e-01
1.00626075e+00 -3.19905043e-01 -3.79329115e-01 1.72764897e-01
-4.38212425e-01 7.06748292e-02 6.53737247e-01 -1.00638306e+00
-5.95894516e-01 1.62385747e-01 1.76754296e-01 -2.36989528e-01
-3.73151392e-01 3.42867643e-01 1.54746696e-01 -5.72483659e-01
9.04786825e-01 6.21254683e-01 -5.34341950e-03 -1.21793464e-01
-4.44581121e-01 1.20816350e+00 8.99252176e-01 -2.79390633e-01
1.14151567e-01 3.94046754e-01 1.18761010e-01 -8.85497868e-01
2.25278065e-01 -3.06010813e-01 -6.29823983e-01 -3.96898836e-01
7.74055600e-01 -5.15701652e-01 -9.68408823e-01 9.39910710e-01
-9.98584211e-01 -5.89879811e-01 6.52120933e-02 1.05497062e+00
-8.89508545e-01 -1.13345250e-01 -7.48804867e-01 -1.00583875e+00
-6.09792769e-01 -9.76829946e-01 8.23629379e-01 3.54358375e-01
-6.24562442e-01 -5.06123602e-01 4.69136098e-03 -2.48831317e-01
6.47275567e-01 3.68837923e-01 7.77198970e-01 -6.38265550e-01
-1.17132314e-01 -4.66366410e-01 -1.39859945e-01 -1.11182127e-02
2.80215949e-01 -5.04123449e-01 -1.16855729e+00 -6.37513548e-02
3.09200108e-01 -3.73815566e-01 3.79714340e-01 6.29398704e-01
1.39781022e+00 3.03824157e-01 -4.32570457e-01 4.84890133e-01
1.33633006e+00 7.92585373e-01 9.71208632e-01 2.40596533e-01
2.97044199e-02 3.60723168e-01 -3.92338820e-02 3.46247971e-01
-3.93525630e-01 4.76487368e-01 -4.04873937e-02 4.08756286e-01
2.55848259e-01 2.02402636e-01 3.66868488e-02 4.04529512e-01
-5.49454153e-01 9.29048806e-02 -1.00144351e+00 3.37090582e-01
-1.57799280e+00 -8.60097885e-01 8.54700729e-02 2.30593133e+00
7.89595127e-01 -2.73669455e-02 -4.33887579e-02 6.53814852e-01
6.53822064e-01 -7.43676782e-01 -8.58846664e-01 -3.49103808e-01
2.11051777e-02 1.16361034e+00 4.99753565e-01 -3.06640446e-01
-6.02641046e-01 3.52447510e-01 6.43267488e+00 2.74807364e-01
-1.79580224e+00 1.35427549e-01 1.90467775e-01 -2.90447265e-01
4.55450356e-01 -7.68844426e-01 -3.01735214e-04 8.21197212e-01
1.19562006e+00 -2.99716722e-02 9.10973787e-01 2.87094891e-01
1.31590784e-01 -3.92863572e-01 -1.09694469e+00 1.76869142e+00
-5.99618480e-02 -1.11247599e+00 -5.44854283e-01 -8.94405693e-02
2.66792923e-01 5.20092435e-02 -1.44935668e-01 -8.47275332e-02
-5.89339256e-01 -1.32456899e+00 5.45011878e-01 6.36632621e-01
1.22950113e+00 -3.37988436e-01 3.74582380e-01 3.13144505e-01
-9.96397913e-01 -3.77787888e-01 1.99386012e-02 -4.28436041e-01
-6.45806864e-02 1.99157134e-01 -3.75043422e-01 -1.65948913e-01
9.69547510e-01 4.10643935e-01 -5.56760281e-02 1.31620812e+00
-6.65958822e-02 7.44837940e-01 -6.11583710e-01 -4.63970721e-01
3.11427526e-02 -1.22916549e-01 3.02495778e-01 1.12431622e+00
6.74764931e-01 6.00798011e-01 -5.61923862e-01 1.08930004e+00
1.24280922e-01 -1.80033430e-01 -5.81125557e-01 -5.00420749e-01
3.89878273e-01 1.14355981e+00 -1.15872931e+00 -6.49758279e-02
-2.76446432e-01 1.21393383e+00 -2.08830666e-02 4.85393167e-01
-6.37047350e-01 -9.46813583e-01 2.92519003e-01 -8.51572081e-02
2.45858803e-02 -3.50007355e-01 -7.88760006e-01 -1.01620197e+00
2.86887169e-01 -4.66570258e-01 -7.01517910e-02 -8.11536133e-01
-1.10276353e+00 6.10561371e-01 -1.53815523e-01 -1.27921093e+00
-4.76518095e-01 -1.04619932e+00 -9.24245536e-01 8.87192428e-01
-1.07542968e+00 -3.27535927e-01 -5.68309665e-01 9.26369250e-01
2.22921595e-01 -9.86050144e-02 1.00008440e+00 1.83077067e-01
-1.11400723e-01 3.18628609e-01 2.64978260e-01 1.03500277e-01
5.33524275e-01 -9.35689986e-01 9.50695127e-02 2.65689492e-01
4.07052785e-02 3.06264073e-01 2.79516399e-01 -3.95590007e-01
-1.50305116e+00 -5.62734962e-01 6.36315942e-01 -4.20808010e-02
6.86400592e-01 -5.15129924e-01 -9.09925818e-01 2.51735538e-01
4.39404137e-02 -4.15664688e-02 8.07537556e-01 -4.83758181e-01
-3.56731899e-02 1.06965102e-01 -1.42520225e+00 2.96992958e-01
1.09475231e+00 -7.25862503e-01 -8.15036416e-01 -6.87222481e-02
-4.65845793e-01 -1.11173622e-01 -9.24166799e-01 3.09684277e-01
1.01550460e+00 -5.57002842e-01 6.56248569e-01 -5.65242112e-01
9.92313847e-02 8.73825401e-02 2.87264902e-02 -1.33733070e+00
-1.48128405e-01 -5.08258700e-01 -3.14217992e-02 5.64670920e-01
2.66553432e-01 -1.03695321e+00 6.14166856e-01 8.37162614e-01
3.45830590e-01 -7.32142389e-01 -1.10125351e+00 -9.97389436e-01
-5.92782199e-02 -5.26752234e-01 3.77158344e-01 5.84405422e-01
8.49236250e-01 -1.21811308e-01 1.45961240e-01 -2.09617183e-01
5.41760087e-01 7.57933259e-02 -2.18909867e-02 -1.40883195e+00
-1.33367777e-01 -5.47899425e-01 -1.07275510e+00 -4.28828597e-01
1.17468223e-01 -9.03570473e-01 2.93058097e-01 -1.20527649e+00
1.23304706e-02 -7.02275783e-02 -7.99167752e-01 3.75409722e-01
3.43319774e-01 6.34718537e-01 -1.22962289e-01 2.21143275e-01
6.56881779e-02 1.28475755e-01 5.55641472e-01 -1.62833512e-01
-4.36224371e-01 -1.96512803e-01 -6.89355806e-02 4.63653415e-01
1.00340784e+00 -5.00747919e-01 -1.70860961e-01 -1.20596707e-01
-3.63358073e-02 7.94034973e-02 6.53849542e-01 -1.52666724e+00
5.94663024e-01 2.45835036e-01 9.00053382e-01 -1.02561116e-01
2.81045049e-01 -6.56862915e-01 2.51338273e-01 5.96965790e-01
-3.50426316e-01 -1.78147614e-01 3.10215890e-01 4.86115634e-01
-1.60504594e-01 -5.75965233e-02 6.46143436e-01 -1.73836015e-02
-6.04733288e-01 -5.51950000e-02 -8.77977252e-01 -2.35082075e-01
1.12550640e+00 -7.91945100e-01 1.70797314e-02 -6.13994002e-02
-8.92256856e-01 -3.86783063e-01 8.63255560e-02 1.86240956e-01
6.52141690e-01 -1.20982635e+00 -3.17437679e-01 7.02250242e-01
1.49695426e-01 -8.72167706e-01 1.12803653e-01 1.35283470e+00
-2.82514066e-01 4.89421993e-01 -1.12819123e+00 -7.97744811e-01
-9.34524298e-01 1.08707525e-01 3.57385218e-01 2.68591374e-01
-7.13169754e-01 6.39945626e-01 -3.11489403e-01 1.43789843e-01
2.66347557e-01 -4.33398992e-01 -2.15795785e-01 -1.00669451e-01
7.60145068e-01 3.04205805e-01 4.39624310e-01 -2.49259591e-01
-5.80410361e-01 5.24584353e-01 5.28608561e-01 -3.24932873e-01
1.35576952e+00 2.88771957e-01 6.91823587e-02 5.91573417e-01
1.16113627e+00 -8.82427454e-01 -1.14157796e+00 2.49445826e-01
2.05257908e-01 -1.46890357e-01 3.35589126e-02 -1.08165109e+00
-8.97355795e-01 1.16702533e+00 1.05587649e+00 4.26107459e-02
1.53214300e+00 -2.99044460e-01 6.91708863e-01 5.79043865e-01
9.63522553e-01 -1.19351697e+00 -2.20649913e-01 7.38110468e-02
7.81622469e-01 -7.69261420e-01 -4.49571490e-01 -3.17763463e-02
-2.80690402e-01 1.47916460e+00 2.47926563e-01 -4.48933542e-01
8.07334960e-01 7.15176463e-01 -1.42883230e-02 -2.07384810e-01
-3.26498806e-01 9.16723981e-02 1.12608433e-01 7.16532528e-01
3.70774031e-01 1.47073209e-01 -7.76828706e-01 9.53160942e-01
1.34127568e-02 9.03988183e-01 2.97605932e-01 1.31502044e+00
-1.62649378e-01 -7.40668178e-01 -3.37459654e-01 9.36514497e-01
-4.09560382e-01 -6.53886944e-02 -4.63856310e-01 5.68515301e-01
-3.13974440e-01 7.81553507e-01 2.27613747e-01 -4.31734443e-01
1.61219910e-01 6.12238407e-01 1.01012468e+00 -3.07364017e-01
-6.14894390e-01 -1.32819384e-01 -5.35740435e-01 -7.89131522e-01
-5.55811107e-01 -4.45426583e-01 -1.65593922e+00 2.77081579e-01
-1.09955169e-01 -2.92728215e-01 1.12478542e+00 1.24540639e+00
5.50980270e-01 3.80983055e-01 1.97371766e-01 -1.17071950e+00
-3.92812639e-01 -1.17825055e+00 -1.07006657e+00 4.16462749e-01
-2.78940927e-02 -1.00251245e+00 -2.11432621e-01 8.55811387e-02]
|
[12.954687118530273, 3.3770713806152344]
|
9b884d6e-ce84-4c08-b6b2-e4d75020bae9
|
faceforensics-learning-to-detect-manipulated
|
1901.08971
| null |
https://arxiv.org/abs/1901.08971v3
|
https://arxiv.org/pdf/1901.08971v3.pdf
|
FaceForensics++: Learning to Detect Manipulated Facial Images
|
The rapid progress in synthetic image generation and manipulation has now come to a point where it raises significant concerns for the implications towards society. At best, this leads to a loss of trust in digital content, but could potentially cause further harm by spreading false information or fake news. This paper examines the realism of state-of-the-art image manipulations, and how difficult it is to detect them, either automatically or by humans. To standardize the evaluation of detection methods, we propose an automated benchmark for facial manipulation detection. In particular, the benchmark is based on DeepFakes, Face2Face, FaceSwap and NeuralTextures as prominent representatives for facial manipulations at random compression level and size. The benchmark is publicly available and contains a hidden test set as well as a database of over 1.8 million manipulated images. This dataset is over an order of magnitude larger than comparable, publicly available, forgery datasets. Based on this data, we performed a thorough analysis of data-driven forgery detectors. We show that the use of additional domainspecific knowledge improves forgery detection to unprecedented accuracy, even in the presence of strong compression, and clearly outperforms human observers.
|
['Matthias Nießner', 'Andreas Rössler', 'Luisa Verdoliva', 'Justus Thies', 'Davide Cozzolino', 'Christian Riess']
|
2019-01-25
| null | null | null | null |
['fake-image-detection']
|
['computer-vision']
|
[ 4.97994542e-01 1.09381117e-01 3.12372539e-02 -1.84890971e-01
-6.52963221e-01 -5.95656097e-01 8.84941518e-01 -8.93930718e-02
-2.57129222e-01 6.11027181e-01 2.80239820e-01 -4.31596972e-02
6.49781749e-02 -6.80638969e-01 -8.13929319e-01 -5.51002204e-01
-3.30658406e-02 1.03531107e-01 4.29479312e-03 -4.38728750e-01
5.42613328e-01 6.03996217e-01 -1.82969832e+00 5.94981670e-01
4.72798020e-01 1.14990282e+00 -3.14360797e-01 4.35837299e-01
3.67569596e-01 8.13431859e-01 -1.00900328e+00 -1.27068937e+00
3.49187464e-01 -2.77820557e-01 -6.98256731e-01 3.67767155e-01
8.56708169e-01 -9.26762044e-01 -4.72986072e-01 1.19811106e+00
6.04636014e-01 -3.88048559e-01 3.16267639e-01 -1.26415551e+00
-7.93418527e-01 4.24703062e-01 -6.44289196e-01 2.98398465e-01
4.95166063e-01 6.26629531e-01 5.77450156e-01 -1.04715323e+00
8.79478812e-01 1.52672577e+00 6.70347810e-01 4.01139170e-01
-1.02345526e+00 -9.88056064e-01 -5.69140971e-01 1.14948303e-01
-1.44123626e+00 -1.17025924e+00 6.02763057e-01 -4.07509476e-01
7.03263462e-01 2.76461035e-01 5.65957963e-01 1.68980539e+00
1.38265537e-02 4.23753500e-01 1.10356534e+00 -5.58051765e-01
5.06757423e-02 1.46992519e-01 -3.83310139e-01 8.80956233e-01
6.33207321e-01 3.99863064e-01 -8.21639419e-01 -5.18233418e-01
6.39052629e-01 -5.19919217e-01 -5.85323155e-01 -9.41503271e-02
-7.58686543e-01 8.14490020e-01 1.06764980e-01 1.65809780e-01
-8.44942555e-02 1.82089359e-01 4.24732655e-01 5.33935845e-01
4.05441910e-01 5.44860303e-01 5.74493222e-02 -1.88590840e-01
-1.01451695e+00 5.45918882e-01 6.96338177e-01 5.86133480e-01
4.05397296e-01 2.00534001e-01 1.76787768e-02 6.40002370e-01
-7.74724931e-02 4.43866253e-01 4.66704816e-01 -1.06794727e+00
4.60919589e-01 2.41456836e-01 2.83681422e-01 -1.88293552e+00
7.62822255e-02 -8.57646391e-02 -5.95641553e-01 5.42326570e-01
5.73720157e-01 1.26125067e-01 -6.83485925e-01 1.54739964e+00
1.17034279e-01 -2.37396900e-02 -3.58989894e-01 9.72819030e-01
5.56241035e-01 1.43453807e-01 -1.84283689e-01 -4.75580171e-02
1.39210093e+00 -5.07737041e-01 -7.65940011e-01 -1.66838706e-01
5.64105332e-01 -1.01301527e+00 7.78377473e-01 9.08401012e-01
-1.26609540e+00 -1.41315788e-01 -1.12367785e+00 3.13308090e-02
-3.21028948e-01 4.35626097e-02 4.45876360e-01 1.27422726e+00
-9.66941476e-01 8.10759604e-01 -3.43584687e-01 -2.55862683e-01
1.01621068e+00 7.76114464e-02 -7.10142732e-01 -2.01033726e-01
-1.12296367e+00 1.06782568e+00 1.43843889e-01 3.44994999e-02
-1.10424972e+00 -5.80302119e-01 -6.10330701e-01 -6.82588741e-02
4.05766636e-01 -1.50122538e-01 1.15067959e+00 -1.29906642e+00
-1.11378312e+00 1.34288895e+00 2.69690067e-01 -4.55703050e-01
1.01452529e+00 -2.27607593e-01 -6.50465906e-01 5.17621458e-01
-1.70987070e-01 5.36811531e-01 1.61174786e+00 -1.36383200e+00
-1.66092247e-01 -3.88599902e-01 -2.96974704e-02 -2.34731033e-01
-5.62554598e-01 5.06305218e-01 -1.37644306e-01 -1.03347039e+00
-4.02098387e-01 -6.23048663e-01 3.27229172e-01 5.12452424e-01
-4.39585686e-01 4.01010245e-01 9.32056010e-01 -8.76603007e-01
1.11198926e+00 -2.15073204e+00 -3.07238042e-01 5.72965331e-02
4.54631835e-01 4.88464177e-01 -8.69459435e-02 4.36982393e-01
-1.84364900e-01 4.32428122e-01 -2.21406043e-01 -3.75229686e-01
-1.31603867e-01 -1.39175272e-02 -4.61231709e-01 9.36347127e-01
4.46627170e-01 8.25436234e-01 -7.76961684e-01 -2.95869648e-01
6.18110746e-02 3.24244142e-01 -5.46777070e-01 -1.52046606e-01
-1.14524201e-01 -9.56253335e-02 2.19877362e-02 8.63913953e-01
8.74937654e-01 -6.41922504e-02 -8.03892091e-02 -2.41698936e-01
2.75287539e-01 -6.27187267e-02 -8.75760376e-01 1.30954635e+00
6.26096204e-02 8.86391819e-01 2.61297971e-01 -7.79189229e-01
6.47234380e-01 2.34670967e-01 9.84441414e-02 -5.17735958e-01
5.48342705e-01 1.53566509e-01 -1.32027432e-01 -5.65404832e-01
6.73103273e-01 -7.60404766e-02 9.19979289e-02 5.79680562e-01
1.51826769e-01 -9.11808386e-02 4.01431788e-03 1.47222996e-01
1.40350127e+00 -3.05674553e-01 1.55412763e-01 -1.24251895e-01
1.82294726e-01 -7.23769739e-02 7.01721162e-02 7.37186193e-01
-3.24255586e-01 6.24468088e-01 5.95442653e-01 -3.71406525e-01
-1.09248269e+00 -8.63749981e-01 -9.17597935e-02 7.76028037e-01
6.33853972e-02 -5.64024329e-01 -9.53392982e-01 -6.54724002e-01
3.49883378e-01 7.01371729e-01 -7.39327550e-01 -5.69595397e-01
-2.94241786e-01 -6.57842577e-01 1.13478386e+00 -5.88308461e-02
7.18193471e-01 -9.95322645e-01 -6.94435239e-01 -2.20623657e-01
-1.47553965e-01 -1.39702904e+00 -2.77450323e-01 -5.76781690e-01
-3.39323431e-01 -1.24764526e+00 -5.07795513e-01 -4.05818105e-01
5.48756897e-01 4.06145304e-01 1.20817161e+00 5.76385260e-01
-7.92874277e-01 3.05226654e-01 -4.80926454e-01 -3.60826015e-01
-8.70399535e-01 -5.77260077e-01 4.71317116e-03 1.95236668e-01
2.42537022e-01 -3.29643548e-01 -5.03385723e-01 5.18760979e-01
-1.03178811e+00 -3.72820675e-01 4.66423452e-01 7.01507688e-01
-3.09271991e-01 2.82010227e-01 3.49989653e-01 -6.25465989e-01
8.29968989e-01 -4.69678253e-01 -5.48121154e-01 8.35990086e-02
-3.93126011e-01 -1.53566509e-01 3.51594567e-01 -3.39741647e-01
-1.11993361e+00 -3.02130908e-01 7.16474503e-02 -7.12071180e-01
-2.00387746e-01 2.00429242e-02 -1.01523310e-01 -6.62227869e-01
1.13993633e+00 8.16625655e-02 2.31704444e-01 -3.81632924e-01
4.06567991e-01 7.15273201e-01 6.78205311e-01 -3.35250854e-01
8.95704448e-01 7.42747784e-01 -1.33868605e-01 -1.04975021e+00
-3.74415487e-01 1.10765144e-01 -2.51860619e-01 -4.47084874e-01
2.98641354e-01 -7.82087505e-01 -5.90743959e-01 9.75868165e-01
-1.25600338e+00 -7.59132355e-02 9.77031738e-02 -7.44905248e-02
-4.14879382e-01 8.18833232e-01 -6.71475410e-01 -6.87230349e-01
-8.83331299e-02 -1.07462776e+00 9.87321675e-01 -4.18839455e-01
-2.29944929e-01 -4.86372858e-01 -5.43102384e-01 6.56995535e-01
4.44374144e-01 6.64247990e-01 6.40211523e-01 -3.51954937e-01
-4.27255273e-01 -4.48719591e-01 -3.25831681e-01 6.61923885e-01
1.95111945e-01 2.00128630e-01 -1.14115059e+00 -3.34795177e-01
1.68243200e-01 -7.99782515e-01 8.65532041e-01 -1.90066397e-01
1.22897482e+00 -6.80496335e-01 -1.61104545e-01 3.15167964e-01
1.18388987e+00 -2.54792392e-01 9.81799722e-01 1.88612118e-01
2.94082582e-01 7.27413654e-01 3.80090117e-01 9.04886484e-01
-1.25594974e-01 7.29833305e-01 7.50703633e-01 1.24368049e-01
-2.73994863e-01 -1.88650399e-01 3.23563248e-01 4.74032536e-02
-5.68482541e-02 -4.25847262e-01 -7.68947840e-01 3.52341861e-01
-1.30732131e+00 -1.31186891e+00 1.00017354e-01 2.24321556e+00
7.93532491e-01 1.74340516e-01 1.38736874e-01 4.19315904e-01
9.11680222e-01 1.71234682e-01 -2.10841641e-01 -3.98649454e-01
-3.58314067e-01 1.72033235e-01 6.32259846e-01 5.92690334e-02
-1.05514252e+00 9.64345336e-01 7.07301855e+00 1.03963971e+00
-1.04085135e+00 9.95820612e-02 9.03896570e-01 -3.09986413e-01
8.33480954e-02 -3.85137826e-01 -4.67987150e-01 7.80982912e-01
8.60283911e-01 -9.00564417e-02 5.79176128e-01 6.58177435e-01
2.25388050e-01 -2.36236930e-01 -8.76986623e-01 1.16548789e+00
6.42776012e-01 -1.38960171e+00 2.36225814e-01 2.01414421e-01
5.56696475e-01 -3.49527091e-01 3.52380037e-01 -2.38698184e-01
3.10475230e-01 -1.30119586e+00 1.01321197e+00 2.63125032e-01
9.18895066e-01 -7.45078206e-01 4.42392230e-01 2.23779023e-01
-3.44211042e-01 -2.34762102e-01 -2.94026107e-01 -1.66140627e-02
2.01996583e-02 4.73028779e-01 -6.67776883e-01 1.39375806e-01
7.28516161e-01 2.94964373e-01 -6.66955233e-01 8.07781339e-01
-2.21154511e-01 4.87913668e-01 -2.72748739e-01 2.94348806e-01
1.83083843e-02 2.61012614e-01 4.47786570e-01 1.11014116e+00
3.21802557e-01 9.32325199e-02 -3.96614105e-01 7.97086835e-01
-5.84539294e-01 -1.13564655e-01 -7.77335763e-01 -2.97761261e-01
4.33091164e-01 1.08879304e+00 -3.59450012e-01 -2.29230434e-01
-3.22646648e-01 1.21258831e+00 3.54282498e-01 -1.82137489e-02
-8.68534505e-01 -8.97152796e-02 8.28704000e-01 4.16548342e-01
2.59616852e-01 -3.90571393e-02 4.02855081e-03 -1.08177471e+00
1.90327749e-01 -1.34986103e+00 1.77468240e-01 -8.07598114e-01
-1.26702046e+00 2.74893880e-01 -2.59217452e-02 -8.25871348e-01
-5.21081984e-02 -7.79337645e-01 -3.28581512e-01 4.78369623e-01
-1.25054276e+00 -8.98991585e-01 -5.69551647e-01 5.12695789e-01
2.75332093e-01 -3.35302413e-01 6.64475143e-01 3.69155258e-01
-3.71313274e-01 9.55394983e-01 -9.25701484e-02 3.27976346e-01
9.21875656e-01 -4.74535018e-01 6.05894446e-01 7.40940630e-01
4.28452455e-02 3.22406888e-01 8.37362051e-01 -7.35405684e-01
-1.38288522e+00 -9.70564306e-01 5.18288374e-01 -6.65218949e-01
7.87587345e-01 -4.68908519e-01 -9.43394423e-01 3.31417352e-01
1.27592519e-01 1.28433838e-01 3.31776738e-01 -4.84595895e-01
-7.48509228e-01 2.43978828e-01 -1.73865497e+00 4.58723694e-01
1.23109257e+00 -4.25767362e-01 -3.98009956e-01 6.05283976e-01
2.57020533e-01 -2.59133697e-01 -6.12019598e-01 3.48316759e-01
6.71274483e-01 -1.51187098e+00 9.23254550e-01 -5.25039852e-01
9.54789162e-01 2.35803530e-01 -8.60495865e-02 -1.16210163e+00
-1.68744132e-01 -9.70059335e-01 -1.81732461e-01 1.02698779e+00
1.87694840e-02 -3.85029972e-01 8.36301804e-01 5.35611391e-01
2.17748433e-01 -5.01089096e-01 -1.04011595e+00 -9.29754853e-01
-1.46986404e-02 -3.23820859e-01 6.01338089e-01 1.11441720e+00
-9.50596780e-02 -4.00435448e-01 -6.79293931e-01 -1.55467642e-02
8.36016119e-01 -3.36139441e-01 9.24424469e-01 -7.47771144e-01
-2.36264870e-01 -6.35502815e-01 -1.02267122e+00 -5.85393608e-01
7.00654462e-02 -5.06149828e-01 -2.64446974e-01 -6.32603228e-01
6.25878945e-02 -2.85125803e-02 3.23521167e-01 5.08578002e-01
-2.59406865e-03 7.80241489e-01 4.57872525e-02 2.03440771e-01
-1.05530329e-01 3.73645723e-01 1.16997778e+00 -4.76406440e-02
5.99631608e-01 -4.44072545e-01 -7.45444775e-01 7.29210973e-01
8.16856980e-01 -3.53794008e-01 -2.71942541e-02 -4.64114428e-01
1.81497917e-01 -2.16502830e-01 9.01370227e-01 -1.05565906e+00
-1.40045494e-01 1.20347599e-02 4.97562945e-01 7.01617170e-03
7.06648648e-01 -5.60620427e-01 6.91845492e-02 4.29510087e-01
-2.64854401e-01 -9.51017067e-02 2.15905502e-01 5.98011136e-01
3.32253128e-02 -3.22249204e-01 1.15416062e+00 -3.39440137e-01
-4.85669434e-01 -4.12304550e-02 -2.89048761e-01 3.52882966e-02
1.05693090e+00 -3.33176523e-01 -6.77064657e-01 -7.87950099e-01
-3.18667322e-01 -3.66042376e-01 8.40993047e-01 5.62060773e-01
7.43535817e-01 -1.17193449e+00 -8.99257362e-01 3.26537341e-01
2.54018992e-01 -8.54984820e-01 7.40918666e-02 4.65620160e-01
-6.86330676e-01 -1.24172390e-01 -4.09875691e-01 -2.26373464e-01
-1.47132826e+00 7.91420460e-01 2.26119280e-01 3.68980706e-01
-4.25519854e-01 8.95049393e-01 -1.33612812e-01 2.96208948e-01
1.06984630e-01 4.31412645e-02 2.67548323e-01 -8.10850412e-02
8.90281260e-01 5.21153450e-01 2.62795210e-01 -1.02780545e+00
-1.50685281e-01 1.20866532e-02 -2.25341171e-01 6.59320951e-02
9.10673261e-01 5.33417426e-02 -6.10969961e-02 -2.25777134e-01
1.15457523e+00 1.24202132e-01 -9.90821123e-01 1.32805062e-02
-1.02958776e-01 -1.38381898e+00 1.33381322e-01 -9.39239264e-01
-1.21002829e+00 6.02402687e-01 3.96512389e-01 1.76448613e-01
1.08095360e+00 -7.36069009e-02 7.20407128e-01 2.57229596e-01
5.38517714e-01 -9.30007279e-01 4.09530282e-01 -3.69169116e-02
1.30936062e+00 -1.27217889e+00 3.38825345e-01 -6.52689219e-01
-4.85316545e-01 1.05461574e+00 4.31833506e-01 -8.00606459e-02
4.82770741e-01 2.41430193e-01 -1.57077029e-01 -3.17998916e-01
-6.00483775e-01 3.32090557e-01 -1.77889407e-01 4.62418944e-01
1.39847919e-02 -1.59259066e-01 -9.11231935e-02 2.36885145e-01
-4.65461284e-01 1.89723670e-02 7.47912645e-01 9.88767326e-01
-4.74802256e-01 -8.84712756e-01 -7.40336359e-01 5.65522552e-01
-6.65120721e-01 5.37182502e-02 -8.60933304e-01 8.02558482e-01
1.69208094e-01 1.17446434e+00 -1.55438125e-01 -2.84361303e-01
4.35940586e-02 -1.98907599e-01 8.35618913e-01 -1.21126652e-01
-6.09058082e-01 -4.37433749e-01 3.88787150e-01 -8.19421113e-01
-1.44820705e-01 -6.59079254e-01 -4.05940503e-01 -9.12841022e-01
-5.21228075e-01 -4.56431657e-01 5.86238444e-01 7.51083195e-01
4.70948309e-01 -1.96456432e-01 5.76909006e-01 -9.43295002e-01
-8.55751157e-01 -7.67462850e-01 -4.79575485e-01 9.31333244e-01
1.88230649e-01 -9.25005794e-01 -6.61026001e-01 1.64552107e-01]
|
[12.570416450500488, 1.0974845886230469]
|
12b16584-eaf9-49c3-ab2b-f453b8c76b5a
|
joint-detection-and-identification-feature
|
1604.01850
| null |
http://arxiv.org/abs/1604.01850v3
|
http://arxiv.org/pdf/1604.01850v3.pdf
|
Joint Detection and Identification Feature Learning for Person Search
|
Existing person re-identification benchmarks and methods mainly focus on
matching cropped pedestrian images between queries and candidates. However, it
is different from real-world scenarios where the annotations of pedestrian
bounding boxes are unavailable and the target person needs to be searched from
a gallery of whole scene images. To close the gap, we propose a new deep
learning framework for person search. Instead of breaking it down into two
separate tasks---pedestrian detection and person re-identification, we jointly
handle both aspects in a single convolutional neural network. An Online
Instance Matching (OIM) loss function is proposed to train the network
effectively, which is scalable to datasets with numerous identities. To
validate our approach, we collect and annotate a large-scale benchmark dataset
for person search. It contains 18,184 images, 8,432 identities, and 96,143
pedestrian bounding boxes. Experiments show that our framework outperforms
other separate approaches, and the proposed OIM loss function converges much
faster and better than the conventional Softmax loss.
|
['Shuang Li', 'Tong Xiao', 'Liang Lin', 'Bochao Wang', 'Xiaogang Wang']
|
2016-04-07
|
joint-detection-and-identification-feature-1
|
http://openaccess.thecvf.com/content_cvpr_2017/html/Xiao_Joint_Detection_and_CVPR_2017_paper.html
|
http://openaccess.thecvf.com/content_cvpr_2017/papers/Xiao_Joint_Detection_and_CVPR_2017_paper.pdf
|
cvpr-2017-7
|
['person-search']
|
['computer-vision']
|
[-1.52777255e-01 -6.04512393e-01 -1.16653815e-02 -6.21274173e-01
-7.91136205e-01 -5.52387714e-01 5.09504139e-01 -6.75743958e-03
-1.04187620e+00 7.49112129e-01 1.30963847e-01 7.85171911e-02
3.41665357e-01 -7.66062558e-01 -8.68316650e-01 -4.20591265e-01
1.90824583e-01 5.98043799e-01 3.68172258e-01 2.13552058e-01
-1.13714427e-01 3.69902492e-01 -1.57528746e+00 1.78893253e-01
6.07421041e-01 1.10511887e+00 -1.07557707e-01 5.63973308e-01
3.51054549e-01 4.56306696e-01 -8.03437650e-01 -1.06188869e+00
5.54583549e-01 9.45423692e-02 -7.84990788e-01 4.87756953e-02
9.64137554e-01 -6.89928889e-01 -8.86889577e-01 1.13589609e+00
1.02415216e+00 3.50874513e-01 3.12268794e-01 -1.38719869e+00
-8.53457272e-01 2.50612587e-01 -6.55023336e-01 2.25536108e-01
4.34423417e-01 3.15101534e-01 7.56698787e-01 -1.00572109e+00
2.48692513e-01 1.62650669e+00 9.82211232e-01 7.35409379e-01
-1.12091291e+00 -9.46910203e-01 3.07239145e-01 3.99607629e-01
-1.71096921e+00 -5.17660022e-01 2.39514008e-01 -3.58894795e-01
7.32454181e-01 2.57653058e-01 4.77277011e-01 1.26726842e+00
-7.19539702e-01 1.04874969e+00 6.50002122e-01 -1.96579874e-01
-2.62711346e-01 4.05469865e-01 4.00861323e-01 5.25322914e-01
6.07910812e-01 2.50024229e-01 -3.55578363e-01 -2.25903019e-01
6.16552830e-01 1.53970405e-01 2.13825554e-02 -3.04207176e-01
-1.07491374e+00 6.05142951e-01 5.80782115e-01 -2.36463770e-01
-2.23299749e-02 1.68462530e-01 6.59318984e-01 9.65636130e-03
1.71319515e-01 -1.98915422e-01 -1.43058389e-01 1.12846270e-01
-7.07661211e-01 6.83075786e-01 6.76965237e-01 1.16025996e+00
6.20964289e-01 -3.73038858e-01 -6.54851437e-01 1.13839591e+00
2.70064175e-01 5.76060653e-01 1.95184186e-01 -6.67583704e-01
5.86530209e-01 6.62788093e-01 5.90165854e-01 -8.86152983e-01
-2.14614823e-01 -2.86796480e-01 -7.69807398e-01 -1.37833551e-01
8.22408855e-01 -1.78888589e-01 -8.16959441e-01 1.73194110e+00
4.74608362e-01 1.33711085e-01 -3.42309356e-01 1.31538057e+00
1.28901911e+00 3.09429348e-01 3.18649739e-01 5.41206360e-01
1.69444513e+00 -1.35129797e+00 -2.36045167e-01 -3.08679253e-01
6.98247552e-02 -6.32472813e-01 8.34376931e-01 -1.01302363e-01
-1.14279830e+00 -9.59594071e-01 -8.83501232e-01 -3.78915071e-01
-6.85314357e-01 4.63426262e-01 3.72640729e-01 8.05989444e-01
-9.89053011e-01 5.87738529e-02 -2.60241598e-01 -5.94993114e-01
5.97496748e-01 5.43957293e-01 -4.66524899e-01 -1.52642459e-01
-1.18326581e+00 7.00125456e-01 4.76261139e-01 2.50299186e-01
-7.55667984e-01 -6.20441914e-01 -8.69110465e-01 1.88976139e-01
2.12184235e-01 -9.16072607e-01 1.13687134e+00 -6.89862669e-01
-1.02757299e+00 1.41967773e+00 -4.21835929e-01 -5.24339736e-01
9.98783350e-01 -4.00555283e-01 -3.13293785e-01 -2.87353873e-01
5.02595186e-01 8.12364459e-01 4.14681971e-01 -1.31992614e+00
-1.07433987e+00 -4.76914883e-01 2.38713130e-01 -7.93071836e-02
-4.57763910e-01 6.44384325e-01 -1.07070565e+00 -6.41283095e-01
-4.87093002e-01 -7.31778264e-01 -1.13959648e-01 2.03263402e-01
-5.98637521e-01 -4.34341311e-01 4.76288855e-01 -7.86081731e-01
9.95664537e-01 -1.84981549e+00 -3.05783510e-01 7.49474242e-02
2.24036828e-01 5.38822532e-01 -1.40009206e-02 -5.94835670e-04
-3.25601511e-02 -1.06859729e-01 1.42690346e-01 -7.31874228e-01
4.70220387e-01 -1.24873742e-01 -8.78613964e-02 4.77196366e-01
3.45189543e-03 1.11043060e+00 -7.61480570e-01 -6.68335259e-01
1.78981707e-01 4.74974394e-01 -3.10820878e-01 1.25146359e-01
4.12688434e-01 2.69829959e-01 -2.23648444e-01 9.60089862e-01
9.51820374e-01 -3.41341496e-01 -2.72182584e-01 -2.03761563e-01
2.39085052e-02 -1.81639403e-01 -1.35864568e+00 1.41434777e+00
-5.37109748e-02 5.65838039e-01 -3.74911800e-02 -9.89044964e-01
8.11348915e-01 -1.04044490e-02 4.11903560e-01 -8.35940063e-01
3.17475796e-02 1.24167383e-01 -5.07639706e-01 -2.97307819e-01
5.43215275e-01 5.90755224e-01 -2.57645428e-01 2.99536377e-01
-7.35797957e-02 9.45325375e-01 3.34686816e-01 -7.70092234e-02
8.01820815e-01 -3.09390966e-02 2.82371156e-02 9.64802876e-02
9.90757644e-01 -7.47422650e-02 7.11610377e-01 1.18748677e+00
-7.92210758e-01 4.60116506e-01 1.07279653e-02 -1.04378557e+00
-1.25738680e+00 -1.14795899e+00 -1.74991101e-01 1.32485151e+00
6.72391236e-01 -1.84046462e-01 -7.79245138e-01 -8.04793775e-01
4.57270503e-01 -7.77620450e-03 -4.68374074e-01 2.40388829e-02
-8.74502301e-01 -9.92398500e-01 8.17258418e-01 7.49445438e-01
1.09705889e+00 -7.86834896e-01 -2.27647096e-01 6.57247081e-02
-4.01602030e-01 -1.47464144e+00 -1.03498435e+00 -5.71564972e-01
-7.83206895e-03 -1.32004774e+00 -1.17078805e+00 -1.07925260e+00
6.21818542e-01 4.77002770e-01 1.34932899e+00 8.32382068e-02
-7.01079428e-01 4.28178459e-01 1.20427385e-02 -3.63455445e-01
1.01819180e-01 -7.77216628e-02 2.29509726e-01 2.98956245e-01
1.04669881e+00 -4.79583070e-02 -9.86262500e-01 7.04730928e-01
-2.41451278e-01 -3.08159918e-01 3.18320781e-01 9.06023204e-01
5.20220578e-01 -7.07929805e-02 3.76190722e-01 -2.82190859e-01
4.41761762e-01 -1.85401782e-01 -7.48128831e-01 6.65338457e-01
-2.89354414e-01 -4.80376601e-01 1.41822398e-01 -5.70181549e-01
-9.85964596e-01 3.34620148e-01 -8.74851644e-02 -1.66127488e-01
-4.59524155e-01 -3.17010611e-01 -3.21279258e-01 -4.15661156e-01
4.04678732e-01 2.72278965e-01 -4.03746337e-01 -6.44458175e-01
1.94777176e-01 7.86265492e-01 9.77596283e-01 -6.53467357e-01
9.13932264e-01 5.88971078e-01 -4.26758647e-01 -5.45363605e-01
-7.19681382e-01 -7.95441628e-01 -7.42677629e-01 -2.64384329e-01
8.60797942e-01 -1.29132175e+00 -1.43695676e+00 7.39302337e-01
-1.38450754e+00 -4.82292734e-02 4.10086970e-04 2.39921495e-01
-1.72970235e-01 6.23837948e-01 -6.53438330e-01 -9.39886510e-01
-5.56667686e-01 -1.05541646e+00 1.35547984e+00 6.41838551e-01
1.67149261e-01 -5.32113612e-01 -1.40956640e-01 5.52776575e-01
1.90142468e-01 1.97915286e-01 2.08447412e-01 -7.42343903e-01
-7.19204187e-01 -7.87080646e-01 -1.03732753e+00 5.93144298e-02
-8.46550539e-02 -4.15286928e-01 -1.08674550e+00 -5.48598886e-01
-5.95077515e-01 -4.74004775e-01 1.20408630e+00 2.45881364e-01
1.29864502e+00 -2.52926588e-01 -6.57193184e-01 7.85203576e-01
1.26553822e+00 2.23239679e-02 3.81575942e-01 7.82750428e-01
7.05002308e-01 5.23882568e-01 4.06742245e-01 5.17162502e-01
9.30671215e-01 1.05056751e+00 2.55688757e-01 -3.32893759e-01
-1.49286315e-01 -3.40353131e-01 -1.10948784e-02 -1.35425732e-01
-2.86238045e-01 -1.24974772e-01 -8.22333097e-01 6.06350005e-01
-2.17457509e+00 -1.12703621e+00 1.73210204e-01 2.37807012e+00
6.36768579e-01 -8.98840576e-02 8.10260534e-01 -2.78586864e-01
1.25433743e+00 -1.41022682e-01 -7.51045763e-01 2.81896442e-01
-1.86832368e-01 -1.90238506e-01 7.43064106e-01 3.81661057e-01
-1.80293667e+00 9.64163065e-01 6.28487396e+00 6.42836094e-01
-5.41601360e-01 2.12014779e-01 6.83960736e-01 -1.96169779e-01
4.42197025e-01 -4.00904000e-01 -1.42038119e+00 9.24660325e-01
3.59587759e-01 -6.78419396e-02 2.34374329e-01 9.10764039e-01
4.48759459e-02 1.53214023e-01 -1.16915286e+00 1.62468815e+00
1.51618004e-01 -1.00397503e+00 -5.93613349e-02 -1.28240794e-01
5.51756501e-01 -2.97772914e-01 6.84222728e-02 4.80650216e-01
5.64372957e-01 -1.08629298e+00 8.08179200e-01 6.39327407e-01
6.21219695e-01 -7.76166439e-01 8.73656631e-01 1.72926143e-01
-1.65523589e+00 -2.97044158e-01 -6.37089312e-01 2.14858845e-01
2.35868514e-01 5.00003397e-02 -3.70184273e-01 4.63940561e-01
1.33570707e+00 4.15633857e-01 -8.70262444e-01 1.68418729e+00
1.33148640e-01 -2.21939273e-02 -2.54195929e-01 4.38534059e-02
3.93527485e-02 8.41392055e-02 3.05822074e-01 1.54222560e+00
-1.44111924e-02 -2.18069315e-01 5.88615358e-01 1.08243978e+00
-1.71552598e-01 -7.97329247e-02 -2.22160563e-01 4.47710007e-01
5.84113717e-01 1.12395442e+00 -3.03751826e-01 -6.11516893e-01
-6.78273201e-01 1.25721228e+00 4.57778066e-01 5.73510408e-01
-1.21564674e+00 -3.39172900e-01 9.20446575e-01 -1.36888057e-01
2.50669092e-01 -1.69144832e-02 3.74820898e-03 -1.24357677e+00
3.35956752e-01 -7.53811538e-01 6.43475711e-01 -3.37963581e-01
-1.87273633e+00 4.09067422e-01 -8.42161626e-02 -9.48364258e-01
4.38416041e-02 -7.36024976e-01 -5.74984133e-01 1.26710987e+00
-1.56134856e+00 -1.74103153e+00 -8.04236472e-01 8.39270949e-01
5.73997319e-01 -3.80569190e-01 5.36238551e-01 1.01539707e+00
-1.01826990e+00 1.32937527e+00 -1.19547583e-01 8.75173390e-01
1.10470927e+00 -1.20100188e+00 9.23831344e-01 9.52781677e-01
-3.09219033e-01 6.43605471e-01 2.68815726e-01 -4.77035642e-01
-9.48352814e-01 -1.34184301e+00 1.01577497e+00 -6.96455359e-01
3.80874246e-01 -6.20628119e-01 -6.44039154e-01 5.87131262e-01
-8.79424810e-02 3.32819849e-01 5.76114118e-01 -8.86024311e-02
-4.56419528e-01 -2.54208654e-01 -1.15606654e+00 4.68789965e-01
1.41756928e+00 -4.86304402e-01 -1.26124442e-01 5.93336284e-01
6.18493080e-01 -5.02707779e-01 -6.97300792e-01 3.79990220e-01
8.88548374e-01 -7.60964036e-01 1.90282094e+00 -9.09733593e-01
-1.86866194e-01 -2.78264701e-01 -1.37095615e-01 -5.61278343e-01
-4.11050379e-01 -3.22232783e-01 -1.24934666e-01 1.29338014e+00
1.36666149e-01 -7.29985774e-01 1.04780567e+00 1.10664523e+00
3.46840173e-01 -4.89203364e-01 -8.55806172e-01 -9.08354282e-01
-8.97091478e-02 4.52359281e-02 9.55060601e-01 5.64190686e-01
-6.41242146e-01 -1.16284929e-01 -6.61849380e-01 4.25906092e-01
1.28884077e+00 8.58302601e-03 1.23922193e+00 -1.13756275e+00
-2.36428693e-01 -6.13860607e-01 -5.15649855e-01 -1.30832124e+00
2.58903176e-01 -6.06794536e-01 -1.95733402e-02 -1.37074316e+00
7.52962828e-01 -5.03117263e-01 -3.94062132e-01 5.19465685e-01
-5.55873692e-01 4.78690475e-01 1.59812644e-01 3.93079877e-01
-9.21135902e-01 3.37999672e-01 7.52537966e-01 -6.64202750e-01
4.04561162e-02 3.97585392e-01 -5.92701733e-01 6.68001235e-01
4.16299194e-01 -1.39575768e-02 8.93865675e-02 -6.52530432e-01
-3.21266264e-01 -4.75636989e-01 1.19974399e+00 -1.14162898e+00
7.10623026e-01 1.70582205e-01 9.87300634e-01 -7.56003976e-01
4.53356326e-01 -6.68559372e-01 -6.94374442e-02 2.69571692e-01
-2.62290239e-01 1.40536716e-02 1.91526219e-01 6.47014558e-01
-1.44384369e-01 -1.84133686e-02 6.98832214e-01 -3.30400199e-01
-1.07372034e+00 8.76842141e-01 2.47707888e-01 1.54634546e-02
9.54920888e-01 -5.08539259e-01 -4.49793696e-01 -1.39077246e-01
-6.52657151e-01 6.71434581e-01 3.84272695e-01 6.05381191e-01
5.62335789e-01 -1.65973854e+00 -8.27809513e-01 1.50073737e-01
2.40603074e-01 -1.48811698e-01 2.96341926e-01 3.55924010e-01
-3.65443498e-01 6.10710323e-01 -1.23209693e-01 -5.75828075e-01
-1.54771996e+00 7.71127582e-01 6.84075356e-01 1.82376001e-02
-5.67298055e-01 1.10333002e+00 3.31485152e-01 -5.66973388e-01
9.12772179e-01 3.76405239e-01 -2.13536248e-01 -2.01437935e-01
8.89284790e-01 5.10881007e-01 -2.98581004e-01 -9.54190373e-01
-5.07372081e-01 6.26305997e-01 -1.72714651e-01 2.02336103e-01
9.19232905e-01 -2.93726653e-01 -1.63010806e-02 -2.36628219e-01
1.07584715e+00 -3.56251687e-01 -1.28394723e+00 -6.34038270e-01
3.46142128e-02 -8.31270874e-01 -6.22511148e-01 -6.26195550e-01
-8.76411140e-01 5.53144336e-01 1.08327889e+00 -6.71175718e-02
8.08614850e-01 -4.29567182e-04 1.14081061e+00 5.42606056e-01
4.84039783e-01 -1.28654587e+00 4.86846641e-02 3.16924453e-01
6.64839327e-01 -1.78826809e+00 -1.27173811e-01 -2.79468000e-01
-2.55671680e-01 8.23507249e-01 1.08089936e+00 -3.86671983e-02
4.34926242e-01 -7.55543113e-02 -2.03068599e-01 1.94506720e-01
-1.23217799e-01 -4.93879408e-01 4.19379562e-01 7.16255009e-01
7.80167654e-02 -2.85762157e-02 1.36343688e-01 9.16757166e-01
-1.00811288e-01 -1.40703619e-01 -2.87885845e-01 6.39068365e-01
-3.36861968e-01 -1.13800764e+00 -6.21110916e-01 9.60938707e-02
-3.36618602e-01 6.63513839e-02 -5.31090856e-01 7.43747950e-01
4.47950333e-01 9.20086801e-01 1.34760186e-01 -2.95739859e-01
5.89801431e-01 -1.13361247e-01 3.47692937e-01 -2.20207304e-01
-4.58913326e-01 -4.45299119e-01 7.78879076e-02 -5.77726543e-01
-4.28880066e-01 -4.69498068e-01 -4.95815575e-01 -7.10148036e-01
-2.82186776e-01 -2.31251135e-01 3.39713603e-01 7.32470214e-01
2.56332219e-01 1.07726000e-01 4.05146122e-01 -8.79065871e-01
-5.84992290e-01 -7.40942836e-01 -1.87546968e-01 6.95915878e-01
2.52685726e-01 -8.23492587e-01 1.09992608e-01 9.60130468e-02]
|
[14.80420970916748, 0.8566433787345886]
|
9c2ffedf-ec54-4d98-b116-af8d8c677475
|
sea-a-spatially-explicit-architecture-for
|
2304.12532
| null |
https://arxiv.org/abs/2304.12532v1
|
https://arxiv.org/pdf/2304.12532v1.pdf
|
SEA: A Spatially Explicit Architecture for Multi-Agent Reinforcement Learning
|
Spatial information is essential in various fields. How to explicitly model according to the spatial location of agents is also very important for the multi-agent problem, especially when the number of agents is changing and the scale is enormous. Inspired by the point cloud task in computer vision, we propose a spatial information extraction structure for multi-agent reinforcement learning in this paper. Agents can effectively share the neighborhood and global information through a spatially encoder-decoder structure. Our method follows the centralized training with decentralized execution (CTDE) paradigm. In addition, our structure can be applied to various existing mainstream reinforcement learning algorithms with minor modifications and can deal with the problem with a variable number of agents. The experiments in several multi-agent scenarios show that the existing methods can get convincing results by adding our spatially explicit architecture.
|
['Guoliang Fan', 'Bin Zhang', 'Zhiwei Xu', 'Dapeng Li']
|
2023-04-25
| null | null | null | null |
['multi-agent-reinforcement-learning']
|
['methodology']
|
[-6.30218506e-01 -3.35124969e-01 -1.53966904e-01 9.02307779e-03
-2.61981100e-01 -5.37903011e-01 9.74760056e-01 2.08048269e-01
-7.72187412e-01 9.64088619e-01 1.04857564e-01 6.43123221e-03
-3.31979752e-01 -1.12271154e+00 -6.80871308e-01 -8.61055672e-01
-3.28020394e-01 7.74487853e-01 8.39716434e-01 -4.67191130e-01
5.21320045e-01 5.61726391e-01 -1.27394938e+00 1.03914939e-01
7.35110223e-01 7.28154063e-01 7.67772079e-01 6.96082234e-01
4.43441793e-02 1.34699845e+00 -8.10513854e-01 2.60170043e-01
3.81702930e-01 -3.23408842e-01 -6.15945160e-01 1.31791428e-01
-2.45840639e-01 -7.26216972e-01 -4.40535456e-01 8.62311721e-01
4.20354664e-01 2.45393723e-01 5.22671223e-01 -1.40027773e+00
-7.09526241e-01 4.27072853e-01 -7.93144584e-01 2.21602887e-01
-4.20846865e-02 2.76090175e-01 2.94072539e-01 -4.83472437e-01
4.41935152e-01 1.19272518e+00 4.03444111e-01 2.85739273e-01
-4.58776772e-01 -2.87101656e-01 4.99399185e-01 7.71060109e-01
-1.38895190e+00 -1.01332590e-01 7.66916871e-01 -2.49860138e-01
1.01183319e+00 -1.46856830e-01 8.22100997e-01 5.69988430e-01
2.91589588e-01 7.57118702e-01 1.01813304e+00 -2.87752450e-01
5.69306254e-01 -2.28534311e-01 -4.99184519e-01 8.48958313e-01
-5.40710315e-02 5.02037406e-01 -2.33689740e-01 8.77548903e-02
1.39015937e+00 2.95748293e-01 2.48697996e-01 -4.71859753e-01
-1.49977517e+00 9.60235238e-01 8.99222016e-01 3.94587010e-01
-4.60840225e-01 5.78572989e-01 2.53651708e-01 4.08715814e-01
1.61945194e-01 2.83067554e-01 -3.32254469e-01 -4.94666807e-02
-4.73958403e-01 4.79770839e-01 5.87740242e-01 1.16853523e+00
9.33104336e-01 1.66209251e-01 2.34662011e-01 6.14417911e-01
3.97035778e-01 5.15613258e-01 6.82988048e-01 -1.27687383e+00
4.22276199e-01 6.29354477e-01 3.60997230e-01 -8.90450776e-01
-6.50447309e-01 -2.82067716e-01 -9.87634540e-01 7.63923764e-01
1.92116231e-01 -4.87463772e-01 -4.85912859e-01 1.54833746e+00
6.18840396e-01 5.19434571e-01 1.64977193e-01 9.70462441e-01
5.15106559e-01 7.42189050e-01 -7.59602413e-02 -8.21506083e-02
1.10495949e+00 -1.57091618e+00 -5.72760463e-01 -3.19866419e-01
5.68511844e-01 -2.98669636e-01 5.08381486e-01 1.13130406e-01
-9.09887314e-01 -4.98657793e-01 -1.01743495e+00 2.25118592e-01
-6.71275854e-01 5.32204993e-02 7.54103065e-01 8.24571699e-02
-1.27621245e+00 3.60043347e-01 -9.73662019e-01 -2.14028284e-01
3.58659804e-01 4.36476141e-01 -1.71016648e-01 1.56809717e-01
-9.81834769e-01 1.29027629e+00 6.65573835e-01 -6.41154647e-02
-1.06863856e+00 -1.88204139e-01 -7.09938824e-01 -1.81810006e-01
4.28114325e-01 -7.67048657e-01 1.47346997e+00 -9.20922577e-01
-1.77662086e+00 -8.88560414e-02 1.82346672e-01 -3.89637142e-01
5.21586061e-01 1.05853602e-01 -7.89354295e-02 2.45858073e-01
2.39120930e-01 7.32976615e-01 6.48559928e-01 -1.18206525e+00
-1.21654069e+00 -4.52072203e-01 3.85729969e-01 7.72838593e-01
-1.44147752e-02 -1.16875991e-01 -3.05963576e-01 -5.37711024e-01
-2.07712457e-01 -9.34079528e-01 -7.56217837e-01 1.45685121e-01
2.20709994e-01 -4.43846405e-01 1.19464397e+00 -2.09556624e-01
6.46208584e-01 -1.98851907e+00 4.49699581e-01 -7.63718635e-02
3.70587021e-01 9.59167071e-03 -2.89882898e-01 7.79764295e-01
5.26162624e-01 -3.51627111e-01 1.04638943e-02 5.65683246e-02
1.34503692e-01 4.38050985e-01 -9.62365121e-02 5.02740741e-01
-1.85644180e-01 9.09666836e-01 -1.11759269e+00 -7.30183780e-01
5.31288445e-01 1.80474117e-01 -4.40272987e-01 9.23960060e-02
-3.99146140e-01 5.41028321e-01 -9.20851052e-01 2.81248748e-01
5.24837434e-01 -4.75644886e-01 3.68984267e-02 3.19417804e-01
-5.16286731e-01 -3.15218747e-01 -1.33688354e+00 2.03405714e+00
-3.58871639e-01 3.73070955e-01 2.93229461e-01 -9.10500348e-01
6.60001397e-01 3.08124632e-01 6.84080958e-01 -9.96913314e-01
-2.02579454e-01 5.60481735e-02 -2.12844275e-02 -3.85007769e-01
5.84608972e-01 7.88816661e-02 -1.42162487e-01 5.51653087e-01
-1.97805375e-01 -1.57121971e-01 3.21697652e-01 7.20177442e-02
1.29654384e+00 1.24947622e-01 5.93928933e-01 -4.64876518e-02
3.22486997e-01 4.29302603e-01 5.12489438e-01 6.87085629e-01
-3.83907199e-01 8.68877172e-02 5.64372055e-02 -7.82915533e-01
-1.15073121e+00 -8.30304742e-01 3.14384788e-01 1.11724198e+00
6.05568051e-01 -2.35056877e-01 -5.76826215e-01 -6.79238319e-01
-1.31602362e-01 1.82144180e-01 -3.84764850e-01 2.85630822e-01
-8.26716125e-01 -6.52380943e-01 2.10707814e-01 6.35845304e-01
1.12618732e+00 -1.29926348e+00 -1.09859502e+00 4.53496546e-01
1.44015044e-01 -1.02087092e+00 -2.55423129e-01 -5.24092428e-02
-5.20244598e-01 -1.08215368e+00 -7.19770491e-01 -1.14846301e+00
4.93331730e-01 5.15077233e-01 6.64444864e-01 2.34689310e-01
9.74727422e-02 3.63500923e-01 -5.84432304e-01 -5.06245136e-01
-1.74842745e-01 2.31802866e-01 9.42834932e-03 -4.44083840e-01
1.37291625e-01 -6.92810059e-01 -7.66295254e-01 3.47918779e-01
-1.00767827e+00 3.05482090e-01 6.95515156e-01 6.47529423e-01
2.08016679e-01 4.69762415e-01 8.68774295e-01 -5.02752960e-01
6.62777424e-01 -5.46808660e-01 -9.33717608e-01 1.29678518e-01
-2.61394680e-01 -4.07289788e-02 9.54714954e-01 -4.00879860e-01
-7.73324370e-01 1.04319192e-01 7.54253417e-02 -1.94155440e-01
-3.91787648e-01 4.44502831e-01 1.53122887e-01 -3.67773414e-01
4.17508125e-01 2.85167009e-01 1.37549788e-01 -2.50360612e-02
5.48661113e-01 5.57341754e-01 1.79454356e-01 -3.62500906e-01
6.13322735e-01 6.45646691e-01 1.56638354e-01 -6.72072113e-01
2.94432491e-02 -2.04009458e-01 -8.51531029e-01 -2.26380065e-01
6.45209968e-01 -9.19827461e-01 -1.17937946e+00 6.64356112e-01
-1.43006802e+00 -7.60317206e-01 -2.13741139e-01 5.91982722e-01
-9.53699470e-01 3.43433291e-01 -8.04674506e-01 -4.33063984e-01
1.79325670e-01 -1.34906793e+00 8.82894695e-01 3.58875662e-01
4.09440935e-01 -1.28409684e+00 5.76081514e-01 -2.09012359e-01
6.83943927e-01 -1.28395170e-01 6.62343323e-01 -4.57064807e-01
-1.10491526e+00 2.82928884e-01 -1.20745085e-01 -2.80725479e-01
1.57132611e-01 -4.59921896e-01 -3.63832474e-01 -3.01573396e-01
3.46428677e-02 -2.37061828e-01 5.45741558e-01 4.25849617e-01
9.67700005e-01 -5.26851833e-01 -5.03754020e-01 2.94176340e-01
1.63199556e+00 4.45361912e-01 4.23447758e-01 7.31325507e-01
5.99275768e-01 5.12573779e-01 5.78115165e-01 8.00045013e-01
9.23553765e-01 8.43798459e-01 9.58395600e-01 -5.53673618e-02
5.48346266e-02 -4.53142822e-02 2.95153946e-01 8.30945373e-01
-1.22956768e-01 -2.13888451e-01 -9.39450860e-01 5.71720719e-01
-2.60841537e+00 -1.35742784e+00 1.41440824e-01 1.60123670e+00
4.83277708e-01 -6.05795741e-01 2.26458952e-01 -3.20499957e-01
6.15381837e-01 4.47017133e-01 -7.00870156e-01 2.15885453e-02
-3.48440409e-02 -3.99991810e-01 7.30997980e-01 5.80286622e-01
-1.31764281e+00 1.13525927e+00 6.87386608e+00 8.61362100e-01
-8.62284482e-01 3.01044345e-01 1.63509756e-01 -1.99758068e-01
2.70703137e-01 -1.61857247e-01 -5.53602219e-01 6.57708704e-01
7.43645847e-01 -1.16249267e-02 8.61009240e-01 8.45716119e-01
2.59755820e-01 -2.97980160e-01 -8.39777827e-01 1.12239826e+00
-1.39691783e-02 -1.71272314e+00 -1.64527655e-01 1.50369674e-01
8.88652325e-01 3.96026760e-01 1.16960324e-01 3.29980433e-01
9.59003568e-01 -1.02454031e+00 5.86787820e-01 3.76452833e-01
2.24758342e-01 -7.58884490e-01 5.78874588e-01 1.01915860e+00
-1.50148070e+00 -4.46008354e-01 -5.84640384e-01 -4.63950098e-01
1.81580260e-01 -1.59083366e-01 -7.81057775e-01 6.05601430e-01
6.12865806e-01 9.09188092e-01 -4.94105637e-01 1.22998238e+00
-1.70747072e-01 -1.67762250e-01 -4.43520069e-01 -4.79379654e-01
7.12451637e-01 -4.46853817e-01 2.74351358e-01 7.31554151e-01
2.93581784e-01 2.11024642e-01 5.66526055e-01 4.54218060e-01
2.77570724e-01 -9.01170522e-02 -1.00177252e+00 4.84954745e-01
6.89540982e-01 1.17337275e+00 -7.39265919e-01 -3.93089831e-01
-6.98921144e-01 9.15441871e-01 6.92187190e-01 4.91350949e-01
-8.25760841e-01 -1.25085190e-01 2.79467165e-01 -1.46834180e-01
5.37053764e-01 -6.30575359e-01 1.63199693e-01 -9.19536948e-01
-1.24258615e-01 -8.11861873e-01 1.65470749e-01 -7.69378304e-01
-1.24392056e+00 4.90787327e-01 -4.73733284e-02 -1.44931591e+00
-6.12641037e-01 -6.56141877e-01 -5.96092582e-01 3.58947814e-01
-1.85670662e+00 -1.44724321e+00 -1.77630544e-01 9.78657126e-01
6.54140711e-01 -7.46922851e-01 7.48479545e-01 2.23950058e-01
-2.33928591e-01 -1.10297799e-01 5.12569785e-01 3.05960238e-01
2.04403087e-01 -1.38358247e+00 -6.58812001e-02 5.50850213e-01
1.73111349e-01 1.62742972e-01 3.38980079e-01 -5.42345405e-01
-1.52698207e+00 -1.23035038e+00 2.73780555e-01 -2.75687844e-01
7.79352248e-01 6.73353150e-02 -5.05026519e-01 1.01705670e+00
8.69072258e-01 5.26361182e-05 3.67965281e-01 -1.11685894e-01
8.45010728e-02 -9.18074921e-02 -1.00387132e+00 8.04714561e-01
8.50525677e-01 -1.12294316e-01 -5.54890513e-01 6.17223740e-01
5.57649493e-01 -5.31214595e-01 -6.63904965e-01 3.88925597e-02
6.14578351e-02 -9.24989879e-01 9.39485371e-01 -4.37661827e-01
2.20079154e-01 -6.76765740e-01 -1.46958068e-01 -1.88840520e+00
-6.11601353e-01 -3.59089196e-01 -1.54361919e-01 6.51051342e-01
-3.76471691e-02 -7.21860051e-01 8.37367058e-01 2.24575952e-01
-1.59433171e-01 -4.31288570e-01 -1.19219267e+00 -6.39469147e-01
2.26094246e-01 5.34329899e-02 1.09904706e+00 8.82942319e-01
2.27205426e-01 3.76644373e-01 -4.55962270e-01 4.29576337e-01
4.97827709e-01 1.64549634e-01 9.82072473e-01 -9.57015038e-01
-3.92911732e-01 -6.12549961e-01 -4.93757427e-01 -1.47309184e+00
1.48455456e-01 -6.20979190e-01 -4.17242683e-02 -1.86783755e+00
1.63153797e-01 -5.83514571e-01 -8.50641951e-02 3.87168348e-01
2.81208456e-01 -3.22343349e-01 3.60540807e-01 4.13655609e-01
-1.26770139e+00 9.52519059e-01 1.45539701e+00 -2.54904121e-01
-6.58189505e-02 -4.74955216e-02 -2.00845867e-01 8.35587025e-01
1.01596880e+00 -1.55683100e-01 -6.16216958e-01 -9.56287205e-01
1.98111057e-01 3.18311989e-01 5.96953750e-01 -1.20486939e+00
1.02960658e+00 -7.20893383e-01 3.39564085e-01 -6.45457625e-01
5.29104412e-01 -1.38412654e+00 -2.19209474e-02 5.42841136e-01
-8.62804279e-02 8.75833511e-01 -2.89552212e-02 7.05389857e-01
-3.37489754e-01 -3.60899746e-01 5.72324216e-01 -5.87234676e-01
-1.23142552e+00 5.47667384e-01 -4.72447574e-01 -2.70506263e-01
1.68430388e+00 -3.64478980e-03 -8.20259094e-01 -4.09510314e-01
-3.70384276e-01 6.63935959e-01 5.51993132e-01 1.68703645e-01
8.45306218e-01 -1.54895139e+00 -6.81159317e-01 1.30403996e-01
-1.13743402e-01 2.69514143e-01 2.62764364e-01 5.76049685e-01
-8.62932205e-01 4.30315584e-01 -4.61804390e-01 -4.89421457e-01
-7.60968387e-01 7.67149329e-01 4.61262614e-01 -3.89120609e-01
-6.61147356e-01 1.87416360e-01 2.86856651e-01 -4.76865590e-01
2.04639480e-01 -1.02758005e-01 -4.13424104e-01 -3.02226722e-01
6.55507684e-01 4.44029182e-01 -3.54363590e-01 -4.88871008e-01
-1.76496133e-01 5.62660813e-01 -8.62553269e-02 -4.46841329e-01
1.80793834e+00 -1.79319248e-01 -2.59922743e-01 3.36138666e-01
7.54207850e-01 -2.94618309e-01 -1.62994599e+00 -3.75658572e-01
-1.02017209e-01 -2.78861344e-01 -7.34635219e-02 -5.61952353e-01
-1.06010556e+00 7.60397792e-01 4.30483520e-01 4.30270106e-01
8.28462601e-01 -5.36986452e-04 3.93983573e-01 5.10138750e-01
9.87731814e-01 -1.40420020e+00 4.22456890e-01 8.80858064e-01
9.51929867e-01 -1.31407773e+00 -8.96312646e-04 2.17232704e-01
-8.99716258e-01 9.58332658e-01 8.60204875e-01 -4.87777472e-01
8.34506691e-01 5.38813233e-01 3.30423228e-02 -2.75768280e-01
-9.79370117e-01 -2.28864864e-01 -3.88621837e-01 1.10001004e+00
-2.72911131e-01 4.79170568e-02 6.74773678e-02 5.71580753e-02
4.65897284e-02 -1.35108471e-01 4.12725985e-01 1.05885231e+00
-8.55884731e-01 -1.08492446e+00 -2.70212740e-01 3.84236798e-02
1.84708804e-01 1.49153799e-01 -6.77892119e-02 9.81545746e-01
1.20407484e-01 8.87962043e-01 2.34421581e-01 -1.16192006e-01
6.89341649e-02 -4.89861965e-01 4.85060304e-01 -5.24748862e-01
-6.22937083e-01 9.41309631e-02 -3.99567276e-01 -4.17125046e-01
-6.47950768e-01 -5.50410867e-01 -1.62208557e+00 -6.02996528e-01
1.25545546e-01 1.93441406e-01 6.31651402e-01 1.01962447e+00
5.08412063e-01 6.62265599e-01 6.55183494e-01 -9.74110961e-01
-5.52063346e-01 -9.76919532e-01 -6.51690304e-01 1.62882265e-02
5.02228737e-01 -7.56006837e-01 2.54404753e-01 -2.63828993e-01]
|
[3.797667980194092, 1.9734996557235718]
|
00b64a4a-08cb-4e24-90cb-8d9fa8fbc5ac
|
locking-on-leveraging-dynamic-vehicle-imposed
|
2306.17529
| null |
https://arxiv.org/abs/2306.17529v1
|
https://arxiv.org/pdf/2306.17529v1.pdf
|
Locking On: Leveraging Dynamic Vehicle-Imposed Motion Constraints to Improve Visual Localization
|
Most 6-DoF localization and SLAM systems use static landmarks but ignore dynamic objects because they cannot be usefully incorporated into a typical pipeline. Where dynamic objects have been incorporated, typical approaches have attempted relatively sophisticated identification and localization of these objects, limiting their robustness or general utility. In this research, we propose a middle ground, demonstrated in the context of autonomous vehicles, using dynamic vehicles to provide limited pose constraint information in a 6-DoF frame-by-frame PnP-RANSAC localization pipeline. We refine initial pose estimates with a motion model and propose a method for calculating the predicted quality of future pose estimates, triggered based on whether or not the autonomous vehicle's motion is constrained by the relative frame-to-frame location of dynamic vehicles in the environment. Our approach detects and identifies suitable dynamic vehicles to define these pose constraints to modify a pose filter, resulting in improved recall across a range of localization tolerances from $0.25m$ to $5m$, compared to a state-of-the-art baseline single image PnP method and its vanilla pose filtering. Our constraint detection system is active for approximately $35\%$ of the time on the Ford AV dataset and localization is particularly improved when the constraint detection is active.
|
['Michael Milford', 'Ankit Vora', 'Shubham Shrivastava', 'Punarjay Chakravarty', 'Sourav Garg', 'Stephen Hausler']
|
2023-06-30
| null | null | null | null |
['visual-localization', 'autonomous-vehicles']
|
['computer-vision', 'computer-vision']
|
[-1.57661363e-01 -2.04040617e-01 -3.07710245e-02 -5.59518516e-01
-7.15382040e-01 -1.06814837e+00 7.62531579e-01 2.45591253e-02
-9.14843976e-01 4.08871859e-01 -3.37125421e-01 -2.08435327e-01
4.25570086e-02 -4.85669196e-01 -8.67517769e-01 -3.65029812e-01
-1.39771134e-01 9.28543925e-01 8.50274980e-01 -1.92106262e-01
3.71694863e-01 1.02895427e+00 -1.55718184e+00 -3.12098563e-01
3.01729947e-01 9.73351896e-01 4.36255515e-01 7.36234307e-01
1.93842113e-01 2.18029127e-01 -5.63354194e-01 1.06443306e-02
6.83729112e-01 1.64053246e-01 -2.55983442e-01 5.89461764e-04
8.27538669e-01 -3.90416086e-01 -1.84386864e-01 9.09826338e-01
1.70172870e-01 2.98362553e-01 2.47014478e-01 -1.44853377e+00
3.47892821e-01 -5.24352342e-02 -4.77800190e-01 2.29204282e-01
6.12656891e-01 2.99928486e-01 5.26522994e-01 -1.01144707e+00
1.10100722e+00 1.17458045e+00 9.37930465e-01 3.50528210e-01
-1.21701264e+00 -7.92563379e-01 3.17705035e-01 2.45962173e-01
-1.79898143e+00 -7.16814458e-01 4.30769026e-01 -4.92235869e-01
1.20641172e+00 2.15299092e-02 5.21172464e-01 6.78057432e-01
2.72242904e-01 1.97421834e-01 4.35595185e-01 -1.09774925e-01
2.48541191e-01 6.95393011e-02 -2.94386923e-01 7.99498141e-01
3.44773382e-01 1.95600286e-01 -6.97272956e-01 -1.07538655e-01
3.90237451e-01 -2.76200324e-01 1.27566829e-01 -1.00046813e+00
-1.40237761e+00 6.81152225e-01 4.00503665e-01 -2.25138232e-01
-2.26116866e-01 5.23593962e-01 2.72193998e-01 6.77136779e-02
2.49460295e-01 3.42412382e-01 -5.31720102e-01 -3.13536525e-01
-1.05802798e+00 3.05928260e-01 4.41079378e-01 1.36016750e+00
1.00704968e+00 -6.06781729e-02 2.22620189e-01 1.74428269e-01
5.81445038e-01 9.24727976e-01 -1.82815075e-01 -1.41850197e+00
3.31795663e-01 5.34989595e-01 5.25019050e-01 -1.04568243e+00
-5.77938139e-01 -8.69127586e-02 2.01680571e-01 4.87075806e-01
2.89984554e-01 -1.43147334e-01 -1.00789988e+00 1.44928789e+00
7.61376441e-01 3.71054500e-01 3.29487175e-02 9.76401925e-01
3.90985936e-01 4.62886721e-01 -2.21616775e-01 -7.73433000e-02
1.07954919e+00 -7.12252855e-01 -5.11683524e-01 -6.52925253e-01
6.31533802e-01 -9.54291701e-01 2.34545127e-01 1.88762277e-01
-5.92334509e-01 -5.45521200e-01 -1.23222351e+00 6.27163127e-02
-4.41347688e-01 1.01590686e-01 4.03951496e-01 4.57449377e-01
-1.25964069e+00 1.63708910e-01 -1.14161265e+00 -6.91191256e-01
2.07776055e-02 6.40227079e-01 -5.07032931e-01 -7.18006939e-02
-7.87848651e-01 1.41557002e+00 2.45092988e-01 1.71193600e-01
-1.07113552e+00 -4.43479776e-01 -1.04198480e+00 -4.82704431e-01
5.15204728e-01 -3.40645254e-01 1.08052015e+00 -3.31373543e-01
-1.04994631e+00 5.67564130e-01 -5.98387897e-01 -6.62311733e-01
9.14475441e-01 -3.30665082e-01 -1.98238567e-01 6.22422658e-02
4.95493084e-01 1.09890580e+00 7.83026993e-01 -1.48229992e+00
-1.07087958e+00 -2.96870828e-01 4.05263677e-02 3.57991844e-01
4.94252115e-01 -4.00618315e-02 -1.16784191e+00 1.20870307e-01
6.95167482e-01 -1.33210671e+00 -5.41553974e-01 4.06360090e-01
1.85529903e-01 6.22472800e-02 1.40359235e+00 -2.00090200e-01
7.09616840e-01 -2.14309311e+00 -2.00582206e-01 5.09528518e-01
-1.01668969e-01 -6.73445463e-02 -1.34738669e-01 2.45837092e-01
4.71435487e-01 -1.59030199e-01 1.81250498e-01 -7.62160182e-01
-9.18690639e-05 5.21119893e-01 -9.18326005e-02 9.77763951e-01
-1.12064749e-01 5.53796768e-01 -1.04453075e+00 -3.67752492e-01
7.47360408e-01 7.33238637e-01 -3.94236952e-01 -1.13995910e-01
-2.03553826e-01 3.75182450e-01 -3.47011000e-01 8.19591105e-01
8.75642061e-01 3.92177612e-01 2.18944140e-02 -1.28893420e-01
-5.65406144e-01 8.19916427e-02 -1.45608974e+00 1.72077012e+00
-2.62195081e-01 9.73279715e-01 2.77618080e-01 -2.13503376e-01
9.42542791e-01 -8.95548239e-02 6.21698141e-01 -4.16501373e-01
6.07218035e-02 2.39647359e-01 -3.23125981e-02 -1.96098119e-01
9.94130433e-01 3.74657393e-01 -1.61435768e-01 -1.85834005e-01
-2.14422673e-01 -3.47115040e-01 4.78170633e-01 1.80462882e-01
1.31614435e+00 4.30031776e-01 -9.44260322e-03 -2.15579972e-01
4.08147544e-01 5.30488789e-01 6.76404297e-01 8.48514318e-01
-5.45637310e-01 6.37632370e-01 2.23506708e-03 -3.17560554e-01
-9.30080891e-01 -9.02940273e-01 -3.86587292e-01 8.92771840e-01
8.49308133e-01 -3.67976099e-01 -3.76360148e-01 -6.80278361e-01
2.71035254e-01 5.40341258e-01 -5.03717065e-01 1.52822062e-01
-7.57403791e-01 -1.45461097e-01 4.56022859e-01 5.52891314e-01
3.92077118e-01 -5.78761578e-01 -1.19761908e+00 1.87020376e-01
8.94710943e-02 -1.34883440e+00 -3.56052995e-01 1.65555164e-01
-4.05634969e-01 -1.03557432e+00 -7.67282471e-02 -5.34949243e-01
9.03241217e-01 5.28889179e-01 6.40938044e-01 8.26734398e-03
-1.99382305e-02 4.68307287e-01 -3.62103432e-01 -2.99900502e-01
-2.89013982e-01 -2.75453001e-01 4.64968532e-01 -2.40103930e-01
4.46692586e-01 1.07521258e-01 -6.18729889e-01 7.68632352e-01
-1.16632432e-01 -1.81013361e-01 3.46263647e-01 3.95107955e-01
7.93418348e-01 -1.18794277e-01 -1.90009862e-01 -3.71955037e-01
-8.12791511e-02 -2.55682886e-01 -1.24756849e+00 -2.03989372e-01
-5.91468036e-01 -1.07615985e-01 7.50756115e-02 -4.22798872e-01
-6.72234356e-01 7.76497245e-01 1.88790318e-02 -7.86686957e-01
-7.48077482e-02 3.94895971e-01 -1.23371063e-02 -6.35274172e-01
7.17345357e-01 -1.96459144e-01 -1.89099219e-02 -1.49065867e-01
2.93921888e-01 8.81790072e-02 6.30733669e-01 -3.15074593e-01
1.11934125e+00 7.39196718e-01 1.23477980e-01 -3.96330923e-01
-2.64931917e-01 -8.66212010e-01 -8.44799280e-01 -5.16158700e-01
7.51415193e-01 -1.05970109e+00 -7.40697563e-01 5.23021864e-03
-1.22584879e+00 -1.62580162e-01 1.55305956e-02 6.40197337e-01
-4.23907280e-01 2.88890451e-01 -9.34656651e-04 -8.11367333e-01
2.38713637e-01 -1.58575952e+00 1.34022033e+00 4.72125225e-02
-1.95750147e-01 -5.09019256e-01 -4.15548906e-02 8.47282708e-02
4.63290066e-01 2.26948619e-01 -8.20799693e-02 -2.93376654e-01
-9.15376186e-01 -5.82272947e-01 -1.16252981e-01 -3.25265527e-01
-3.13235104e-01 1.45254299e-01 -6.45031452e-01 -5.87094069e-01
-3.94355744e-01 2.78173983e-01 7.47744501e-01 2.79657245e-01
3.76914382e-01 1.14847422e-01 -8.81212831e-01 5.79970658e-01
1.41391742e+00 4.16071802e-01 3.39937896e-01 6.12708569e-01
5.69043815e-01 3.74884009e-01 1.21257353e+00 3.66894901e-01
5.81385672e-01 1.04618526e+00 9.20581341e-01 1.60743430e-01
2.66092084e-02 -2.36287624e-01 4.08371657e-01 -7.83887729e-02
5.36045507e-02 -7.77905881e-02 -1.12036788e+00 6.05380118e-01
-1.93108678e+00 -6.99358940e-01 -2.18380272e-01 2.23855662e+00
2.15574130e-01 4.21409309e-01 -2.68346578e-01 -4.01423872e-01
6.21130824e-01 1.80861592e-01 -5.62878013e-01 -4.69565541e-02
2.44576126e-01 -4.10654008e-01 1.33982909e+00 9.75625694e-01
-1.13184655e+00 1.29343843e+00 6.08218718e+00 2.34721825e-01
-1.12902415e+00 4.35218588e-02 -1.32827044e-01 -1.89153850e-01
4.98747490e-02 4.26217079e-01 -1.31367385e+00 2.82083720e-01
7.93325663e-01 1.72514811e-01 2.22236872e-01 1.25163829e+00
3.46415430e-01 -7.15264380e-01 -1.22999108e+00 9.59278286e-01
1.46507949e-01 -1.27716088e+00 -4.88698274e-01 2.40930840e-01
5.09512603e-01 6.47553802e-01 -8.53657797e-02 1.88333467e-01
1.68572485e-01 -6.60230100e-01 1.19872618e+00 3.50305974e-01
6.65977359e-01 -7.81542480e-01 8.52257550e-01 3.98747802e-01
-1.54882634e+00 2.84153186e-02 -3.00603658e-01 9.00125888e-04
4.16755825e-01 -1.35954632e-03 -1.34827542e+00 2.68879920e-01
7.70330727e-01 3.45067292e-01 -7.70223498e-01 1.13076544e+00
-1.19273679e-03 1.68489471e-01 -9.15477216e-01 1.07929662e-01
3.28768760e-01 1.10509373e-01 7.80754983e-01 1.27115071e+00
5.21859348e-01 -1.27067000e-01 5.49943089e-01 5.10720670e-01
2.70877331e-01 -3.68944198e-01 -6.52994812e-01 5.96525550e-01
9.77999568e-01 1.19049287e+00 -1.08727908e+00 -2.64368773e-01
-2.07061350e-01 6.33099020e-01 -1.17302150e-01 2.36716703e-01
-1.10854912e+00 2.88963150e-02 8.62234533e-01 2.90791214e-01
4.55400616e-01 -8.08276653e-01 -1.82512075e-01 -9.09589529e-01
8.19987431e-02 -2.82518893e-01 -5.63950576e-02 -7.07293272e-01
-4.31134254e-01 5.43628812e-01 2.00015128e-01 -1.48084605e+00
-5.74804723e-01 -3.57965291e-01 -1.04884133e-01 6.70273125e-01
-1.17560327e+00 -1.22120440e+00 -4.35978621e-01 4.30363029e-01
6.72545910e-01 8.63796175e-02 4.79041874e-01 2.67057240e-01
-7.96726719e-02 4.05678689e-01 -1.35820791e-01 -1.93342827e-02
6.92225635e-01 -8.46100867e-01 4.92608935e-01 1.33117139e+00
1.11277454e-01 6.22018278e-01 1.12409246e+00 -8.99747491e-01
-1.76594329e+00 -1.08220541e+00 8.81325722e-01 -1.00079918e+00
4.62048441e-01 -6.71484232e-01 -4.01373744e-01 8.31178844e-01
-2.47641921e-01 4.31023777e-01 -1.69575915e-01 -1.43406570e-01
-1.43975705e-01 -2.22724229e-01 -1.40606081e+00 5.25674641e-01
1.09249210e+00 -2.89015114e-01 -2.55684614e-01 1.66672274e-01
6.21810019e-01 -1.06196547e+00 -4.11667407e-01 7.35910833e-01
6.70659304e-01 -7.45862186e-01 9.19482172e-01 2.36083418e-01
-6.48100376e-01 -1.24201977e+00 -5.02375185e-01 -7.38079667e-01
-1.14869423e-01 -4.28216070e-01 6.36043847e-02 9.56561446e-01
4.86396998e-01 -4.16901082e-01 9.06941831e-01 7.51730502e-01
-3.76415312e-01 -3.17498982e-01 -1.50424790e+00 -5.83675861e-01
-8.52192461e-01 -8.34652901e-01 2.60964215e-01 6.71272159e-01
-4.11067039e-01 -1.63826421e-01 -1.13584869e-01 7.94719636e-01
4.95501190e-01 -2.31004000e-01 1.22559071e+00 -9.44040775e-01
1.85680300e-01 -2.07087502e-01 -1.01480353e+00 -1.15798712e+00
1.47111550e-01 -3.87364239e-01 6.97546124e-01 -1.46373081e+00
-2.37705857e-01 -6.47747576e-01 1.07862696e-01 4.21043098e-01
4.10119385e-01 5.55651188e-01 3.60570163e-01 1.68603003e-01
-9.33888912e-01 3.82422954e-02 4.75091964e-01 -2.98324656e-02
-2.29296565e-01 -1.24061055e-01 5.74312061e-02 8.47029448e-01
4.67637420e-01 -4.15888548e-01 -2.91684300e-01 -6.00151002e-01
1.31790444e-01 -4.23025340e-02 4.36545819e-01 -1.35148537e+00
5.74354053e-01 -2.22320497e-01 5.46548367e-01 -1.17066479e+00
7.69371152e-01 -1.21881390e+00 5.43737829e-01 3.88468415e-01
1.69794932e-01 3.34407717e-01 4.56962794e-01 4.78692144e-01
7.21785501e-02 -1.22569516e-01 6.56673789e-01 -5.91642456e-03
-1.46012926e+00 2.53389955e-01 -4.71385866e-01 -5.70018351e-01
1.27988398e+00 -5.96059263e-01 -1.42399088e-01 -1.70994014e-01
-5.29423892e-01 5.36269069e-01 9.65931356e-01 6.49641931e-01
7.03303754e-01 -9.47025359e-01 -3.28200996e-01 3.03998858e-01
2.36882821e-01 2.92653054e-01 -1.33877724e-01 8.29581439e-01
-9.54375505e-01 3.31247896e-01 1.00646012e-01 -1.19762075e+00
-1.45296502e+00 4.63784635e-01 3.58516365e-01 3.24472934e-01
-5.09187818e-01 8.12378645e-01 -1.67546242e-01 -2.42238179e-01
3.77312958e-01 -3.30269039e-01 6.49321973e-02 8.54561701e-02
3.57635409e-01 2.98784196e-01 1.46659017e-01 -1.33947933e+00
-8.92804027e-01 8.53777647e-01 7.07628876e-02 -3.56696963e-01
9.46749449e-01 -5.66574216e-01 1.67203784e-01 8.29436108e-02
9.60107863e-01 2.52395749e-01 -1.80536127e+00 1.37707725e-01
1.51412860e-01 -6.78726077e-01 -7.06396066e-03 -5.68390787e-01
-7.38046348e-01 2.40797967e-01 1.02435005e+00 -3.14434558e-01
4.45100367e-01 3.01875621e-01 3.44932616e-01 5.28163970e-01
1.06477106e+00 -1.12078214e+00 -3.61457050e-01 9.13157225e-01
6.62153959e-01 -1.37408662e+00 2.39474222e-01 -3.36114854e-01
-3.71988803e-01 9.01368499e-01 7.90344357e-01 -1.23038918e-01
3.24189156e-01 5.91200709e-01 3.90447795e-01 -1.05049372e-01
-3.96276116e-01 -3.58279288e-01 1.03179723e-01 7.83743024e-01
-2.02576831e-01 -1.99352503e-01 -2.04916254e-01 -1.99379116e-01
-2.56944120e-01 -4.85711396e-01 2.72020787e-01 1.04300153e+00
-7.18070090e-01 -7.19261646e-01 -5.63608825e-01 1.48299197e-02
8.58895853e-02 2.39474192e-01 -1.51970461e-01 9.63307917e-01
6.00490630e-01 1.04099500e+00 2.62337208e-01 -4.05121148e-01
3.60820502e-01 -1.23989604e-01 2.60099918e-01 -4.57993716e-01
-2.53959775e-01 2.59549797e-01 3.17550600e-01 -1.03846729e+00
-5.39593697e-01 -1.07459021e+00 -1.67800832e+00 -1.98041633e-01
-5.30767143e-01 -1.24250084e-01 1.11938131e+00 8.87715399e-01
4.94123518e-01 -4.85499809e-03 3.52552980e-01 -1.55900598e+00
2.13676058e-02 -6.79194987e-01 4.67549339e-02 -5.52057661e-02
5.34559309e-01 -1.01830649e+00 -2.31689438e-01 -3.89325581e-02]
|
[7.373980522155762, -2.1359565258026123]
|
ad34b65a-da3c-420f-bfee-ce6422e6ae57
|
removing-supervision-in-semantic-segmentation
|
2303.17410
| null |
https://arxiv.org/abs/2303.17410v1
|
https://arxiv.org/pdf/2303.17410v1.pdf
|
Removing supervision in semantic segmentation with local-global matching and area balancing
|
Removing supervision in semantic segmentation is still tricky. Current approaches can deal with common categorical patterns yet resort to multi-stage architectures. We design a novel end-to-end model leveraging local-global patch matching to predict categories, good localization, area and shape of objects for semantic segmentation. The local-global matching is, in turn, compelled by optimal transport plans fulfilling area constraints nearing a solution for exact shape prediction. Our model attains state-of-the-art in Weakly Supervised Semantic Segmentation, only image-level labels, with 75% mIoU on PascalVOC2012 val set and 46% on MS-COCO2014 val set. Dropping the image-level labels and clustering self-supervised learned features to yield pseudo-multi-level labels, we obtain an unsupervised model for semantic segmentation. We also attain state-of-the-art on Unsupervised Semantic Segmentation with 43.6% mIoU on PascalVOC2012 val set and 19.4% on MS-COCO2014 val set. The code is available at https://github.com/deepplants/PC2M.
|
['Fiora Pirri', 'Nico Samà', 'Simone Rossetti']
|
2023-03-30
| null | null | null | null |
['unsupervised-semantic-segmentation', 'patch-matching']
|
['computer-vision', 'computer-vision']
|
[ 2.40964651e-01 6.47662699e-01 -3.26398790e-01 -5.22198081e-01
-1.16292930e+00 -8.77851546e-01 3.46891820e-01 1.69001520e-01
-3.51790845e-01 2.45344296e-01 -3.71943384e-01 3.75703461e-02
1.85445771e-01 -6.76207483e-01 -1.02222061e+00 -4.32474405e-01
1.92315113e-02 9.19712722e-01 5.94511926e-01 3.65971588e-02
2.23262981e-01 -2.01214869e-02 -1.46740198e+00 5.15337706e-01
8.89761508e-01 1.36713004e+00 5.62767267e-01 8.20201218e-01
-1.93671033e-01 2.55718708e-01 -1.83301717e-01 -3.41317475e-01
4.58920658e-01 -2.80793399e-01 -1.13231397e+00 2.31393486e-01
7.75086105e-01 3.58346879e-01 5.68293445e-02 1.27307737e+00
9.17362049e-02 -1.60924494e-01 9.52300906e-01 -1.11080432e+00
-5.53272486e-01 5.44942260e-01 -5.70581019e-01 -3.64657760e-01
-9.07562599e-02 2.14780912e-01 1.34340358e+00 -8.92630696e-01
6.81120753e-01 1.22061598e+00 7.29934514e-01 5.93855560e-01
-1.31114018e+00 -2.66376913e-01 3.61038893e-01 1.63542922e-03
-1.39545143e+00 -1.99975997e-01 5.38460016e-01 -4.31241214e-01
6.76622152e-01 2.10704044e-01 6.18274450e-01 6.46022916e-01
-1.44650802e-01 1.28210068e+00 1.16372526e+00 -3.08194309e-01
3.15532833e-01 -3.27685289e-02 1.31694824e-01 9.33464170e-01
-3.50565575e-02 -2.76047826e-01 -3.10031772e-01 3.63294989e-01
6.38680875e-01 -1.47443458e-01 1.39269635e-01 -4.95024860e-01
-1.03189278e+00 6.42844379e-01 7.33326972e-01 1.94144338e-01
-7.33302310e-02 3.62114847e-01 3.27490717e-01 1.14213377e-02
6.77015662e-01 4.50170517e-01 -9.80639040e-01 1.89070925e-01
-1.21400797e+00 2.21707270e-01 5.68759561e-01 1.28843415e+00
9.64963019e-01 -3.12364072e-01 -1.62009880e-01 1.01831448e+00
3.23133081e-01 6.11685038e-01 2.06892267e-01 -1.30276871e+00
3.60657185e-01 7.50042975e-01 -8.24448988e-02 -5.52569807e-01
-5.02996325e-01 -7.29873538e-01 -5.97418606e-01 -1.23472676e-01
6.73821270e-01 1.62776515e-01 -1.57433915e+00 1.50341976e+00
3.06282878e-01 6.88840896e-02 -1.48427874e-01 8.62903595e-01
7.63551831e-01 5.53859949e-01 1.88817948e-01 2.94284642e-01
1.57974875e+00 -1.47504950e+00 -2.36931041e-01 -6.11897290e-01
7.22034752e-01 -7.63384044e-01 1.19372630e+00 3.35552543e-01
-9.86887753e-01 -5.30868769e-01 -6.97443962e-01 -3.06909293e-01
-4.36268121e-01 3.69500935e-01 7.17957377e-01 3.50264639e-01
-1.07799196e+00 5.79410195e-01 -1.06112874e+00 -3.91777098e-01
1.06416821e+00 3.30492288e-01 -2.06501111e-02 -1.40420213e-01
-5.34802020e-01 4.59153295e-01 4.86026883e-01 -8.59223604e-02
-1.12849963e+00 -8.73050809e-01 -8.83970618e-01 -1.18970066e-01
5.76651573e-01 -3.76069754e-01 1.20896590e+00 -9.56838310e-01
-1.11171508e+00 1.51035810e+00 -1.16480686e-01 -4.89843816e-01
5.46707928e-01 2.03096922e-02 2.62349676e-02 3.25694412e-01
6.21264458e-01 1.64320707e+00 4.93904442e-01 -1.51672184e+00
-9.36593115e-01 -4.93179411e-01 -2.10500553e-01 2.17657119e-01
7.15964809e-02 -5.71668744e-01 -1.04903877e+00 -5.52356005e-01
5.73110759e-01 -1.11602485e+00 -4.31760818e-01 1.43084690e-01
-6.70110583e-01 -4.04760391e-01 7.75544465e-01 -6.00940883e-01
5.91652691e-01 -1.88636470e+00 2.02874199e-01 -2.55487245e-02
8.36357847e-03 1.65817276e-01 -3.56760919e-01 -1.96738631e-01
1.82221025e-01 1.64862663e-01 -6.71143115e-01 -8.74636948e-01
1.24853469e-01 1.91168174e-01 9.25097317e-02 5.07531047e-01
3.83190930e-01 1.23238599e+00 -7.88134336e-01 -6.78946555e-01
3.30023587e-01 6.64799958e-02 -7.32513785e-01 6.36353940e-02
-8.20308506e-01 6.09807312e-01 -4.92989808e-01 1.22509837e+00
6.53006732e-01 -4.14613038e-01 -2.53138300e-02 -1.88157603e-01
1.42377585e-01 3.11257830e-03 -7.58409739e-01 2.27841330e+00
-3.39428812e-01 3.72274846e-01 4.51062411e-01 -1.34794462e+00
8.26118171e-01 -5.62802926e-02 6.31971002e-01 -7.13396728e-01
2.70038933e-01 4.35739815e-01 -3.31659883e-01 -1.19433537e-01
3.20364803e-01 2.36067265e-01 -3.15326184e-01 -9.79076847e-02
4.13953871e-01 -6.87025785e-01 3.26592028e-01 1.84577435e-01
7.65807331e-01 5.24053037e-01 -2.25036755e-01 -6.99899733e-01
2.94125587e-01 4.36492890e-01 6.47888422e-01 6.06363893e-01
-3.09627950e-01 1.10857975e+00 6.40159786e-01 -2.00303689e-01
-1.02146876e+00 -1.10537970e+00 -3.29703629e-01 1.15827787e+00
3.95750582e-01 -4.24196124e-01 -1.30288625e+00 -1.03283787e+00
-2.25015562e-02 6.34346843e-01 -5.92271984e-01 2.43796602e-01
-3.35627168e-01 -6.79668307e-01 4.80214000e-01 4.51993376e-01
4.39095616e-01 -1.06177378e+00 -2.40832612e-01 8.97281468e-02
-2.15702340e-01 -1.47639763e+00 -5.59928298e-01 4.62941229e-01
-8.13494623e-01 -1.08938396e+00 -8.92901957e-01 -1.09353232e+00
8.38137627e-01 9.87768266e-03 1.28321695e+00 1.29765421e-01
-4.43496883e-01 2.39843339e-01 -3.68915707e-01 -2.21284956e-01
-1.61257833e-01 4.72054660e-01 -5.02169371e-01 -1.90271378e-01
1.71215206e-01 -3.43619078e-01 -7.66755283e-01 5.55017114e-01
-7.54388392e-01 4.32305217e-01 4.45811838e-01 6.38105273e-01
1.34031129e+00 -2.60790735e-01 4.42588866e-01 -1.02092552e+00
-4.36769694e-01 -4.68494147e-01 -7.12010324e-01 1.79675162e-01
-7.06337333e-01 -1.89818129e-01 5.73148489e-01 -1.85767323e-01
-6.34779751e-01 6.06139123e-01 -3.25876743e-01 -4.50324833e-01
-5.24716735e-01 -1.79682355e-02 -4.36785907e-01 1.12785034e-01
5.03965139e-01 7.04826787e-02 -7.43759200e-02 -6.85254395e-01
6.18811727e-01 6.22626424e-01 8.18880916e-01 -8.18462133e-01
4.43524867e-01 6.16058350e-01 -1.93592504e-01 -5.93543947e-01
-1.33688915e+00 -7.55624533e-01 -8.77298892e-01 -9.56491977e-02
1.32441854e+00 -1.17790830e+00 -4.30822343e-01 7.04506695e-01
-8.66890073e-01 -9.49144244e-01 -3.84829581e-01 -1.99788101e-02
-8.32083881e-01 2.94856150e-02 -6.18558228e-01 -3.94243807e-01
-3.43928307e-01 -1.27660441e+00 1.66457593e+00 1.92676067e-01
-5.61367497e-02 -9.89755988e-01 -4.04086560e-01 9.24692035e-01
5.12390099e-02 1.39499426e-01 5.82056940e-01 -5.06342828e-01
-7.08021402e-01 -9.71959457e-02 -5.28550506e-01 3.74130815e-01
-1.64359301e-01 -2.87213773e-01 -1.05621338e+00 -1.62319407e-01
-5.99349082e-01 -5.03699243e-01 1.27883422e+00 6.28823876e-01
1.62046945e+00 4.77470178e-03 -5.16404629e-01 9.19038713e-01
1.50263214e+00 -2.30320618e-01 4.01342511e-01 3.41576226e-02
1.20508265e+00 7.62237191e-01 8.07312012e-01 -1.97296944e-02
8.24986160e-01 6.62388623e-01 6.88382983e-01 -2.57202417e-01
-5.50242603e-01 -4.35558915e-01 2.50494350e-02 6.30019248e-01
4.14212912e-01 -2.62895077e-01 -1.05397856e+00 1.06799150e+00
-1.95558548e+00 -1.82446033e-01 -4.50812161e-01 1.75440025e+00
8.09513927e-01 3.39597076e-01 8.03743377e-02 -2.85239905e-01
6.59304976e-01 -2.74268980e-03 -6.89393759e-01 -1.48353666e-01
-9.09120217e-02 1.89240515e-01 1.05046332e+00 6.04487121e-01
-1.48127520e+00 1.62860823e+00 5.13727236e+00 1.46526539e+00
-8.00400376e-01 3.64027232e-01 1.17040074e+00 8.57733414e-02
-4.32973862e-01 1.58825532e-01 -9.86366332e-01 5.74169934e-01
5.47958493e-01 7.11877108e-01 2.64405280e-01 8.19226027e-01
-7.79210180e-02 -2.95387387e-01 -9.85815942e-01 9.06175733e-01
-1.47974402e-01 -1.32418132e+00 -2.75229335e-01 1.24982722e-01
1.18765414e+00 4.32543695e-01 -3.61884423e-02 3.16816777e-01
3.49978238e-01 -1.12425566e+00 1.06961513e+00 1.82126269e-01
1.05580723e+00 -5.11880577e-01 4.69620585e-01 3.54256898e-01
-1.25925338e+00 1.65247306e-01 -3.60495269e-01 2.89444745e-01
7.87595734e-02 6.41033053e-01 -5.38459778e-01 4.35732424e-01
9.29240108e-01 9.97073889e-01 -8.24722648e-01 6.62251413e-01
-3.44577223e-01 6.58510327e-01 -6.08396411e-01 1.98776886e-01
6.10000253e-01 -2.21794412e-01 2.22290352e-01 1.27444339e+00
9.97414663e-02 -1.94268093e-01 5.10016859e-01 9.23957705e-01
-2.67864347e-01 9.42023620e-02 -1.75324664e-01 2.92027481e-02
1.92712396e-01 1.40670693e+00 -1.48934758e+00 -4.03896630e-01
-1.07727006e-01 1.08251631e+00 3.33701402e-01 1.85235083e-01
-7.21582115e-01 3.43502685e-02 3.48623544e-01 5.74045815e-02
5.11300623e-01 -8.58998448e-02 -8.48199844e-01 -1.06218064e+00
-6.12787232e-02 -6.32102311e-01 3.93275946e-01 -5.77352762e-01
-1.20987713e+00 3.38294208e-01 -1.30378887e-01 -9.00500119e-01
1.79850802e-01 -7.39892185e-01 -4.23046887e-01 5.14079273e-01
-1.40249765e+00 -1.59401500e+00 -1.36800215e-01 2.78273076e-01
9.20305908e-01 -1.01355657e-01 6.86001480e-01 2.39909738e-01
-4.22968745e-01 4.80696946e-01 7.21542118e-03 9.50384587e-02
5.89182556e-01 -1.66041791e+00 5.43644845e-01 6.57797396e-01
2.56944984e-01 -9.97899547e-02 5.23649037e-01 -5.62794030e-01
-1.03835118e+00 -1.46255410e+00 6.85268283e-01 -5.64292550e-01
4.44834828e-01 -7.33078361e-01 -6.90415561e-01 5.77147305e-01
-6.20878227e-02 3.17698359e-01 1.02494359e-01 5.09719402e-02
-2.11029693e-01 -5.95714785e-02 -1.24259436e+00 4.72148329e-01
1.29010439e+00 -4.02089268e-01 -9.96468887e-02 7.20280647e-01
1.02361119e+00 -5.85228980e-01 -9.08033729e-01 4.31883246e-01
2.55594105e-01 -6.42017007e-01 9.61874247e-01 -2.41907120e-01
6.13917649e-01 -3.70524436e-01 -3.54016870e-01 -1.08965361e+00
2.30873395e-02 -2.17295721e-01 1.60888195e-01 1.21131384e+00
6.77552283e-01 -2.66598195e-01 1.25152588e+00 2.37792403e-01
-5.22901535e-01 -1.03736389e+00 -7.15880573e-01 -7.02498376e-01
3.86913687e-01 -6.98225737e-01 1.73546150e-01 7.49003112e-01
-6.23506129e-01 2.72530615e-01 3.62394080e-02 9.90912616e-02
9.16380882e-01 2.74009079e-01 5.72408319e-01 -1.01782715e+00
-2.15146914e-01 -6.20497704e-01 -2.62359530e-01 -1.29747343e+00
3.76428485e-01 -1.34212053e+00 3.45838189e-01 -1.72797501e+00
2.38190487e-01 -7.39353895e-01 -1.13348164e-01 7.67974257e-01
-2.67288778e-02 8.27925026e-01 2.80102164e-01 2.86249280e-01
-1.03468370e+00 4.15959775e-01 1.48713088e+00 -3.08959395e-01
-6.63214102e-02 -1.06119461e-01 -7.11377025e-01 8.69919181e-01
9.53185678e-01 -6.51395798e-01 -2.65744269e-01 -4.95315999e-01
-5.82959205e-02 -1.56874210e-01 5.11546016e-01 -9.10656095e-01
5.72246574e-02 -8.64949310e-04 3.17851067e-01 -8.32535803e-01
3.25352192e-01 -5.48688889e-01 -3.20838422e-01 3.68382514e-01
-3.62207621e-01 -5.10309458e-01 1.13432743e-01 4.94135886e-01
-1.20246693e-01 -2.73125499e-01 9.71377194e-01 -2.80468643e-01
-9.53818202e-01 5.42458415e-01 -3.30240503e-02 4.74235564e-01
9.39152658e-01 -1.44514307e-01 -1.15746498e-01 1.61848888e-02
-9.97631431e-01 8.60402346e-01 7.28060305e-01 4.65815097e-01
1.80652216e-01 -9.67481732e-01 -6.64114416e-01 -1.03591852e-01
2.49579936e-01 7.80382276e-01 2.46331066e-01 8.36462677e-01
-6.26010537e-01 4.36541378e-01 3.24232355e-02 -1.19635177e+00
-1.04258811e+00 1.66896790e-01 3.06579202e-01 -2.39744529e-01
-4.05674577e-01 1.17675972e+00 5.35450757e-01 -9.69860613e-01
1.50157988e-01 -2.79987872e-01 1.29642993e-01 6.09041238e-03
-1.26398459e-01 7.53597543e-02 -7.13455165e-03 -6.15055442e-01
-5.02955258e-01 9.15505230e-01 1.46436729e-02 4.30655032e-02
1.19332278e+00 -2.53645867e-01 -1.08703934e-01 1.50821716e-01
1.45168614e+00 -3.65136951e-01 -1.79045320e+00 -1.73380464e-01
1.18198283e-01 -7.01578856e-02 1.74162328e-01 -1.16312706e+00
-1.38734388e+00 8.54313433e-01 6.75578177e-01 -1.02578275e-01
8.98599505e-01 5.54321885e-01 9.54069972e-01 1.31743133e-01
4.40338492e-01 -1.32627535e+00 4.38499898e-02 6.47468805e-01
5.66645563e-01 -1.51145291e+00 -1.11633316e-01 -8.12643886e-01
-6.53235018e-01 6.15778863e-01 6.40228331e-01 -3.23224187e-01
6.62523568e-01 1.53723165e-01 2.00440288e-02 -2.55966783e-01
-5.11525750e-01 -5.64833641e-01 5.25707304e-01 5.33403039e-01
2.41765067e-01 5.93509674e-01 -4.84276600e-02 7.00980008e-01
-2.23749205e-01 -4.69868839e-01 1.50307849e-01 5.35356462e-01
-4.67858493e-01 -1.03214991e+00 -1.99878559e-01 4.93108630e-01
-4.96976227e-01 -9.44253057e-02 -3.68133992e-01 6.32952631e-01
4.78511900e-01 8.77548158e-01 3.69730681e-01 -1.08075403e-02
1.44715369e-01 -4.29791287e-02 4.85853642e-01 -8.90892923e-01
-3.46494943e-01 5.20876288e-01 -3.12405024e-02 -6.70954764e-01
-1.44842118e-01 -5.86228490e-01 -1.70883203e+00 4.90895547e-02
-1.15516424e-01 -9.16158706e-02 7.71913886e-01 9.09789503e-01
3.59867215e-01 5.15387237e-01 2.96895981e-01 -1.00640845e+00
-1.29632905e-01 -9.78797376e-01 -4.79525119e-01 4.96573538e-01
2.80257557e-02 -4.33354825e-01 -1.48250327e-01 3.47374976e-01]
|
[9.582404136657715, 0.6460556387901306]
|
96dced80-cb8c-443f-8c4f-3568737da69e
|
henet-forcing-a-network-to-think-more-for
|
2110.10872
| null |
https://arxiv.org/abs/2110.10872v1
|
https://arxiv.org/pdf/2110.10872v1.pdf
|
HENet: Forcing a Network to Think More for Font Recognition
|
Although lots of progress were made in Text Recognition/OCR in recent years, the task of font recognition is remaining challenging. The main challenge lies in the subtle difference between these similar fonts, which is hard to distinguish. This paper proposes a novel font recognizer with a pluggable module solving the font recognition task. The pluggable module hides the most discriminative accessible features and forces the network to consider other complicated features to solve the hard examples of similar fonts, called HE Block. Compared with the available public font recognition systems, our proposed method does not require any interactions at the inference stage. Extensive experiments demonstrate that HENet achieves encouraging performance, including on character-level dataset Explor_all and word-level dataset AdobeVFR
|
['Youdong Ding', 'Shugong Xu', 'Shiyi Mu', 'Jingchao Chen']
|
2021-10-21
| null | null | null | null |
['font-recognition']
|
['computer-vision']
|
[ 1.41456157e-01 -5.29887140e-01 1.08451039e-01 -5.87678671e-01
-3.14862669e-01 -8.78464937e-01 2.73266733e-01 -2.55207777e-01
-9.43250060e-02 4.82469052e-01 -1.38158724e-01 -6.52705073e-01
1.31376535e-01 -4.73130375e-01 -4.73004699e-01 -8.20966244e-01
3.80557925e-01 1.28069609e-01 2.81400979e-01 -2.70612150e-01
3.80989462e-01 5.52093208e-01 -1.47502923e+00 6.52347982e-01
1.14078879e+00 9.22766864e-01 3.75741571e-01 9.42831516e-01
-5.24925709e-01 5.92248142e-01 -9.01652992e-01 -5.31751037e-01
4.78149205e-01 5.31745702e-02 -3.52109402e-01 3.43702465e-01
1.04574525e+00 -5.54319859e-01 -2.63746470e-01 1.10052919e+00
6.87112629e-01 -1.91323265e-01 7.32834935e-01 -8.23524356e-01
-1.17421150e+00 5.12284756e-01 -6.62685573e-01 1.39410093e-01
2.97059655e-01 2.42041335e-01 9.17660773e-01 -9.14779842e-01
2.52500534e-01 1.21647930e+00 5.24967015e-01 3.98847997e-01
-9.80774522e-01 -6.72186971e-01 4.11476493e-01 1.24768443e-01
-1.29870713e+00 -1.24366857e-01 8.08592200e-01 -4.46374089e-01
9.87156928e-01 5.91958284e-01 3.44587088e-01 1.21646261e+00
-2.11016506e-01 1.21621442e+00 1.54470670e+00 -5.72823107e-01
-8.34997073e-02 4.72281396e-01 7.62556374e-01 6.83742046e-01
1.11511825e-02 -3.75223070e-01 -2.97508806e-01 2.39561677e-01
7.38288879e-01 1.20654583e-01 -3.96755308e-01 -1.17060661e-01
-7.08673775e-01 5.18032789e-01 4.66451161e-02 1.67729467e-01
2.81680763e-01 -3.63466114e-01 2.52201557e-01 5.14291048e-01
2.48930335e-01 2.25896537e-01 -4.33917403e-01 -2.98840284e-01
-9.41160500e-01 -1.00168817e-01 9.77250874e-01 1.27685392e+00
3.63451511e-01 8.79305974e-02 -1.67617068e-01 1.18193889e+00
7.59038702e-02 7.02876270e-01 4.37655061e-01 4.95019928e-02
1.00221157e+00 6.20292306e-01 -1.73581645e-01 -7.82301843e-01
-6.48426786e-02 -2.59769678e-01 -8.77793252e-01 2.61667043e-01
6.91352367e-01 1.18300132e-02 -1.05257499e+00 9.71984088e-01
-9.86597538e-02 -3.22303981e-01 -2.72619992e-01 9.89176273e-01
8.17385077e-01 6.79306209e-01 -2.46844113e-01 3.23231518e-01
1.51486301e+00 -1.26362598e+00 -6.49145305e-01 -2.61571825e-01
4.05927569e-01 -1.08446729e+00 1.45527112e+00 6.82321727e-01
-6.75428867e-01 -7.29043067e-01 -1.30986845e+00 -3.80790263e-01
-9.98408794e-01 7.67415106e-01 8.17566991e-01 1.25741053e+00
-8.04245591e-01 3.82685870e-01 4.19314057e-02 -2.89313167e-01
3.68386418e-01 2.89850473e-01 -2.28015304e-01 -1.83040164e-02
-8.76767993e-01 7.53610313e-01 2.22182721e-01 4.35502708e-01
-2.60043889e-01 -2.60336369e-01 -5.68063676e-01 2.18301818e-01
3.79389048e-01 1.05004273e-01 8.38988721e-01 -9.61642444e-01
-1.65008640e+00 6.04524791e-01 -7.27038383e-02 -1.01541229e-01
1.02403891e+00 -3.78776819e-01 -8.66653860e-01 -3.41610789e-01
-5.02683640e-01 4.44909543e-01 1.41570914e+00 -8.33256960e-01
-3.76344949e-01 -7.52276406e-02 -1.37881562e-01 1.09173596e-01
-9.39978838e-01 2.88424045e-01 -6.24170482e-01 -8.57273996e-01
-7.00377487e-03 -5.90033054e-01 4.66572195e-01 3.37559789e-01
-6.81218863e-01 -4.29228187e-01 1.03460872e+00 -8.57059479e-01
1.30171871e+00 -2.17795753e+00 -2.31854081e-01 2.79103041e-01
1.40441656e-01 5.18518448e-01 -4.02267933e-01 4.05212432e-01
-2.91770101e-01 4.69611436e-02 6.41333237e-02 -2.16336399e-01
2.12763667e-01 9.85965952e-02 -7.26096213e-01 2.43586108e-01
4.10190284e-01 6.34841681e-01 -3.04695487e-01 -4.26536500e-01
2.35715866e-01 2.52480268e-01 -3.08537111e-02 6.13189749e-02
-1.42744005e-01 -1.72075152e-01 -2.67726362e-01 9.99337673e-01
1.26677382e+00 -1.96654141e-01 1.70405999e-01 -5.70315048e-02
-1.92437872e-01 2.72275925e-01 -1.38200498e+00 1.19265544e+00
-8.65067840e-02 9.43076611e-01 -1.92700714e-01 -6.22037411e-01
1.29012048e+00 -1.00629903e-01 -3.50169897e-01 -5.11703908e-01
1.92175359e-01 1.30460009e-01 -5.45107052e-02 -5.88852167e-01
8.42115104e-01 6.05981588e-01 9.81607661e-02 2.40529910e-01
-4.78383660e-01 2.57756021e-02 2.91048408e-01 -6.01743273e-02
8.55246067e-01 2.94331431e-01 -8.89199898e-02 -1.71924308e-01
7.98458636e-01 -2.98170507e-01 2.41473556e-01 8.16537559e-01
-1.44218042e-01 8.10288668e-01 5.14365435e-01 -5.46018720e-01
-1.03345191e+00 -1.10628009e+00 -3.04563284e-01 1.05037832e+00
-1.21554926e-01 -5.38681030e-01 -7.24769950e-01 -9.18236971e-01
2.43341699e-01 3.51735264e-01 -5.34083247e-01 4.88295496e-01
-8.26830506e-01 -7.12514699e-01 7.10147142e-01 9.01908457e-01
5.83616376e-01 -8.60240936e-01 -8.60445499e-02 -1.93970963e-01
2.29988649e-01 -1.20839560e+00 -9.24465001e-01 4.75221276e-01
-8.63043547e-01 -8.57823730e-01 -1.02995312e+00 -1.08350670e+00
9.05337870e-01 5.46261430e-01 9.15028751e-01 -1.88464969e-02
-5.95533669e-01 -1.61646768e-01 -2.49318570e-01 -3.11405659e-01
-2.70928860e-01 1.45316124e-01 -2.74510145e-01 -2.15369090e-02
6.66912794e-01 -1.54900149e-01 -3.39492440e-01 6.43316388e-01
-7.83450484e-01 4.11069207e-02 6.24999285e-01 9.36060071e-01
3.14438581e-01 2.03951329e-01 -1.56122819e-02 -7.46840954e-01
8.83653522e-01 1.47323266e-01 -8.81710052e-01 6.75402701e-01
-6.12533748e-01 1.08010970e-01 1.12834764e+00 -8.63337696e-01
-7.62044191e-01 2.78191179e-01 2.21819282e-02 -1.35140970e-01
-2.47979015e-01 -2.25111004e-02 -5.58868289e-01 -1.07372306e-01
3.16055536e-01 5.31666219e-01 -1.12013243e-01 -1.04525590e+00
2.62495011e-01 1.18634009e+00 2.99473226e-01 -4.98689353e-01
1.08702409e+00 5.36303259e-02 -3.84419084e-01 -1.08858395e+00
-3.33754539e-01 -1.52814195e-01 -7.72946656e-01 -2.03356922e-01
6.97848499e-01 -7.87762821e-01 -8.75373065e-01 7.29885101e-01
-9.90633726e-01 9.68243927e-03 -2.77163535e-02 -6.87277243e-02
-5.14287082e-03 9.76086438e-01 -5.24783313e-01 -7.63256669e-01
-2.34088928e-01 -1.19046021e+00 8.10129941e-01 4.39395398e-01
-2.83028129e-02 -6.85841501e-01 -2.87648797e-01 2.54752725e-01
3.66450638e-01 -3.16300452e-01 1.26462209e+00 -4.89594012e-01
-8.01540673e-01 -4.21201497e-01 -7.02766538e-01 5.71135938e-01
3.20998907e-01 2.79602408e-01 -1.30015421e+00 -2.74288356e-01
-3.31654757e-01 -3.43040705e-01 9.68032837e-01 -4.36070710e-02
1.60122967e+00 -2.64983624e-04 -5.79838119e-02 8.61428320e-01
1.08280122e+00 2.05927968e-01 6.81624055e-01 3.87453467e-01
8.93427610e-01 4.01636511e-01 3.17981571e-01 3.18431824e-01
1.38037398e-01 5.72940469e-01 9.93631408e-02 -2.71429271e-01
-1.21577211e-01 -1.51340604e-01 5.20241678e-01 9.35471773e-01
9.21658706e-03 -3.69364947e-01 -5.41066825e-01 -3.53553407e-02
-1.62871540e+00 -7.65361309e-01 -1.86245844e-01 2.04137659e+00
8.40280652e-01 3.26006293e-01 6.61460534e-02 3.45036417e-01
8.90210629e-01 5.04688442e-01 -7.82658815e-01 -8.90299857e-01
-7.80732930e-01 7.26846652e-03 6.00658298e-01 1.35683790e-01
-1.25903738e+00 1.06114602e+00 7.01646423e+00 1.05048311e+00
-1.03600919e+00 -5.44872999e-01 7.13086963e-01 6.58797771e-02
4.07654047e-02 -2.78066341e-02 -1.41227186e+00 6.07830524e-01
4.40627754e-01 4.80703264e-01 7.37346709e-01 9.43652511e-01
-7.34626353e-02 4.40093651e-02 -1.11966860e+00 1.10662079e+00
4.27425683e-01 -8.62590909e-01 6.10148795e-02 -6.23755194e-02
1.72572687e-01 -4.50742006e-01 4.43017572e-01 3.09664577e-01
3.54215294e-01 -1.00843251e+00 7.34464884e-01 4.80043858e-01
9.78558004e-01 -4.82768863e-01 2.85795569e-01 2.41233781e-01
-1.20189571e+00 -2.11599067e-01 -8.78720105e-01 7.02074319e-02
-3.02291602e-01 5.18217504e-01 -8.52489650e-01 1.94322526e-01
7.68120587e-01 6.91031337e-01 -1.24576175e+00 9.64843571e-01
-3.37936848e-01 4.38809842e-01 -2.43174657e-01 -5.55187762e-01
9.05565768e-02 -4.14079815e-01 2.53195018e-01 1.52375507e+00
2.23679736e-01 -4.66506153e-01 5.42175882e-02 9.09658432e-01
-1.92791730e-01 2.81208247e-01 -2.36589834e-01 -4.26817745e-01
2.08621532e-01 1.41382658e+00 -7.03558743e-01 -4.30940241e-01
-5.63064158e-01 1.40833378e+00 5.51830888e-01 2.90262252e-01
-8.90370429e-01 -7.39802539e-01 4.46855634e-01 -2.46526197e-01
6.02501929e-01 -2.70440340e-01 -4.23106194e-01 -1.36628497e+00
7.12888181e-01 -1.28502095e+00 2.05546901e-01 -9.05426800e-01
-1.56315637e+00 6.69408619e-01 -5.61410904e-01 -1.48381770e+00
1.87073857e-01 -1.42681265e+00 -7.17828929e-01 1.38440764e+00
-1.52713490e+00 -9.83907402e-01 -5.01701117e-01 4.31499869e-01
9.20615733e-01 -4.70927805e-01 6.50855482e-01 1.86825395e-01
-1.04511380e+00 1.14820957e+00 6.06333315e-01 5.48959613e-01
1.19743884e+00 -1.66571712e+00 8.30204189e-01 8.14840138e-01
-7.47383088e-02 7.09655762e-01 5.03081203e-01 -6.98486626e-01
-1.58559632e+00 -6.70089006e-01 7.35022783e-01 -4.60577309e-01
7.17014909e-01 -1.12758029e+00 -8.91552448e-01 5.26699312e-02
3.34804237e-01 -1.52363583e-01 5.66419899e-01 4.89398353e-02
-1.20396960e+00 -1.80442899e-01 -7.90203273e-01 7.05505252e-01
9.11180973e-01 -6.56389177e-01 -4.58822548e-01 4.80634093e-01
3.10245216e-01 -2.81536311e-01 -6.57699287e-01 -2.32646137e-01
7.45292068e-01 -7.52863884e-01 8.99084806e-01 -6.26374841e-01
3.60821635e-01 -3.70696455e-01 1.86471939e-02 -1.00252509e+00
-3.60876203e-01 -6.99578702e-01 -1.24721311e-01 1.37758017e+00
5.74650705e-01 -6.92657828e-01 1.04666746e+00 7.47397661e-01
1.58634409e-01 -4.49881107e-01 -4.04309541e-01 -1.21985602e+00
2.04240128e-01 1.44408187e-02 9.01096523e-01 9.07294631e-01
-1.36457682e-01 9.63042527e-02 -7.13749528e-01 4.62629981e-02
2.73700833e-01 3.44692767e-01 7.62031853e-01 -1.27059650e+00
-7.28771925e-01 -5.73203743e-01 -1.32695839e-01 -1.58673334e+00
-1.46952793e-01 -7.65005708e-01 -1.97356030e-01 -1.15526128e+00
1.74906135e-01 -1.42369092e-01 -6.96110949e-02 4.49290037e-01
-3.17071766e-01 -9.78918076e-02 4.63527173e-01 1.68761656e-01
-4.26773071e-01 2.96383321e-01 1.23561239e+00 -6.73788309e-01
-1.45365417e-01 -1.15764365e-01 -5.41501224e-01 5.03859222e-01
8.04450929e-01 -1.74998537e-01 -1.39243498e-01 -6.51129782e-01
2.27542281e-01 -4.36865240e-01 6.11557951e-03 -9.03327167e-01
2.11117446e-01 2.11714298e-01 1.02385974e+00 -1.12360072e+00
3.03666621e-01 -9.24196959e-01 -5.37003458e-01 8.78406689e-02
-3.45716864e-01 4.49380487e-01 2.91512430e-01 3.56531799e-01
8.49702284e-02 -5.38610816e-01 4.88981336e-01 3.43130350e-01
-6.38078392e-01 2.65837591e-02 -5.71979642e-01 -1.67568251e-01
5.77078581e-01 -7.52121270e-01 -8.28498662e-01 -8.46665874e-02
-3.95115942e-01 -5.77816963e-02 4.31436628e-01 7.62454391e-01
5.69212258e-01 -9.39296007e-01 -5.27119637e-01 5.68655193e-01
2.58576632e-01 -5.11692822e-01 1.24705203e-01 4.16164547e-01
-5.45170963e-01 4.71792728e-01 -2.52046078e-01 -3.64216506e-01
-1.66305363e+00 7.00441539e-01 2.14507550e-01 -1.69782713e-01
-7.47599125e-01 7.33738244e-01 -1.04159445e-01 -3.75630468e-01
6.48108006e-01 -3.63934934e-01 -3.73294264e-01 4.42776456e-02
7.32833683e-01 4.58744794e-01 5.31076267e-02 -2.39048883e-01
-5.08125722e-02 8.14195156e-01 -8.99777472e-01 4.37643856e-01
1.07608938e+00 1.47990817e-02 5.80078848e-02 2.01312691e-01
1.04928231e+00 1.39262870e-01 -1.33949685e+00 -1.78688958e-01
3.73752937e-02 -6.66311979e-01 -3.09687823e-01 -1.10616410e+00
-1.06563663e+00 1.17550433e+00 1.03494883e+00 3.46696317e-01
1.15805781e+00 -7.58725643e-01 6.67669535e-01 9.47834492e-01
-1.08113734e-03 -1.62772501e+00 -2.45989069e-01 7.45485663e-01
6.69414401e-01 -1.36564267e+00 1.51511148e-01 -8.82889211e-01
-5.31179607e-01 1.56102812e+00 8.49168301e-01 -3.44173908e-02
3.71928364e-01 5.91602385e-01 1.94677725e-01 3.30386072e-01
-5.76641738e-01 1.33611277e-01 4.43056196e-01 5.83477437e-01
7.32762575e-01 2.15532631e-02 -1.74593821e-01 5.46878994e-01
-8.73913318e-02 -3.74099076e-01 4.73413467e-01 8.15033138e-01
-3.90717089e-01 -1.38418937e+00 -5.24706662e-01 6.32179916e-01
-1.27669141e-01 -6.51591659e-01 -9.07784283e-01 7.21594334e-01
-1.97927520e-01 7.04904974e-01 -8.19566380e-03 -4.10272956e-01
1.57918409e-01 3.00415605e-01 3.42753321e-01 -9.61827114e-02
-8.17002594e-01 1.30825043e-01 -7.51551823e-04 -3.30910593e-01
2.94430077e-01 -5.95128000e-01 -6.54192209e-01 -3.45946282e-01
-6.52094364e-01 -9.06338096e-02 8.22839260e-01 4.82923299e-01
2.35376060e-01 5.79092979e-01 7.62025356e-01 -3.51629615e-01
-1.11614990e+00 -8.20744157e-01 -7.27098286e-01 2.18729421e-01
8.04491267e-02 -2.61681676e-01 -2.78945267e-01 -3.53480279e-02]
|
[11.955574035644531, 2.094174385070801]
|
1f461b8a-a1bb-4e2e-b664-fd92f71ce438
|
alignscore-evaluating-factual-consistency
|
2305.16739
| null |
https://arxiv.org/abs/2305.16739v1
|
https://arxiv.org/pdf/2305.16739v1.pdf
|
AlignScore: Evaluating Factual Consistency with a Unified Alignment Function
|
Many text generation applications require the generated text to be factually consistent with input information. Automatic evaluation of factual consistency is challenging. Previous work has developed various metrics that often depend on specific functions, such as natural language inference (NLI) or question answering (QA), trained on limited data. Those metrics thus can hardly assess diverse factual inconsistencies (e.g., contradictions, hallucinations) that occur in varying inputs/outputs (e.g., sentences, documents) from different tasks. In this paper, we propose AlignScore, a new holistic metric that applies to a variety of factual inconsistency scenarios as above. AlignScore is based on a general function of information alignment between two arbitrary text pieces. Crucially, we develop a unified training framework of the alignment function by integrating a large diversity of data sources, resulting in 4.7M training examples from 7 well-established tasks (NLI, QA, paraphrasing, fact verification, information retrieval, semantic similarity, and summarization). We conduct extensive experiments on large-scale benchmarks including 22 evaluation datasets, where 19 of the datasets were never seen in the alignment training. AlignScore achieves substantial improvement over a wide range of previous metrics. Moreover, AlignScore (355M parameters) matches or even outperforms metrics based on ChatGPT and GPT-4 that are orders of magnitude larger.
|
['Zhiting Hu', 'Ruichen Li', 'Yichi Yang', 'Yuheng Zha']
|
2023-05-26
| null | null | null | null |
['fact-verification', 'semantic-textual-similarity', 'semantic-similarity']
|
['natural-language-processing', 'natural-language-processing', 'natural-language-processing']
|
[ 3.25457990e-01 2.93790221e-01 8.18019733e-02 -3.97043049e-01
-1.39927316e+00 -7.19808578e-01 1.07784283e+00 5.47759771e-01
2.43247226e-02 1.15178740e+00 7.16776669e-01 -1.55813619e-01
-2.74255633e-01 -5.01538396e-01 -6.89952970e-01 -1.72225460e-01
4.37064946e-01 6.25592649e-01 2.06553601e-02 -5.16598582e-01
7.78596818e-01 -3.19437176e-01 -1.44890273e+00 7.54598677e-01
1.34902871e+00 9.51420605e-01 -3.40295769e-03 5.18394232e-01
-4.96383458e-01 1.02260506e+00 -1.28377163e+00 -1.14048326e+00
-1.84190452e-01 -8.36535335e-01 -1.28524625e+00 -2.23980144e-01
9.17133272e-01 8.80211294e-02 1.66914985e-01 1.20352054e+00
6.14635527e-01 -1.24475770e-01 6.69027150e-01 -1.46709275e+00
-7.87350297e-01 1.02499926e+00 -1.64452404e-01 2.92047709e-01
8.67013574e-01 4.40652482e-02 1.27171791e+00 -1.11641729e+00
7.98521638e-01 1.36359680e+00 6.71691537e-01 3.20642114e-01
-8.07740808e-01 -5.08283854e-01 -2.22779661e-01 5.86214483e-01
-1.01774728e+00 -5.70603549e-01 7.26382494e-01 -8.68769065e-02
1.10644341e+00 5.98268449e-01 2.99200341e-02 1.42317164e+00
3.35306019e-01 6.96654797e-01 9.71776605e-01 -2.94152528e-01
9.86425281e-02 5.55816442e-02 1.88410431e-01 2.93399900e-01
5.77256560e-01 -6.24090731e-01 -8.09626997e-01 -3.00812423e-01
9.99837816e-02 -6.16028190e-01 -4.88179535e-01 3.65628928e-01
-1.56600392e+00 6.01999283e-01 1.30085796e-01 4.47895974e-01
-2.00550154e-01 -1.35010198e-01 6.36782110e-01 6.51123822e-01
2.12631822e-01 9.27465081e-01 -4.34785575e-01 -3.47896576e-01
-7.63820112e-01 7.21895278e-01 1.12269926e+00 1.25875497e+00
4.30132329e-01 -2.35614076e-01 -6.50113940e-01 1.01290023e+00
-1.16305523e-01 6.51571214e-01 8.67456436e-01 -1.13082874e+00
1.41003537e+00 7.20184684e-01 2.60512918e-01 -1.40974343e+00
-3.47843230e-01 -4.44672823e-01 -1.24780452e+00 -6.13833368e-01
2.78131574e-01 5.73642068e-02 -3.45560282e-01 1.68321121e+00
7.63351992e-02 -1.60806611e-01 4.78539795e-01 6.81726158e-01
1.32544303e+00 5.94644547e-01 -3.75028551e-01 -3.21726203e-01
1.17387521e+00 -1.06109512e+00 -1.06394482e+00 -3.23044002e-01
4.46053147e-01 -1.25369513e+00 1.22980130e+00 3.96008670e-01
-1.36012220e+00 -4.88765389e-01 -1.10869789e+00 -3.86198789e-01
-3.68022680e-01 -8.13980103e-02 2.25311786e-01 1.96650967e-01
-9.05241430e-01 6.68228149e-01 6.18112087e-02 -2.94529498e-01
8.59026760e-02 -3.03803444e-01 -3.09181064e-01 -1.46228030e-01
-1.51158059e+00 1.12051964e+00 6.20962501e-01 -6.80332258e-02
-4.15508270e-01 -5.36682487e-01 -8.30000103e-01 2.15700239e-01
6.69596493e-01 -1.22576416e+00 1.45158470e+00 -5.00747502e-01
-1.34410083e+00 7.67123044e-01 -1.90488726e-01 -5.20449281e-01
5.28170884e-01 -2.91313589e-01 -5.38996994e-01 1.97993033e-03
5.06267488e-01 2.75122970e-01 4.84787911e-01 -1.28301334e+00
-4.16235059e-01 -2.17268333e-01 2.09709723e-02 4.37325597e-01
1.72636495e-03 5.21116294e-02 -1.64759234e-01 -7.05642998e-01
1.93106189e-01 -5.21749973e-01 2.31934965e-01 -5.84408402e-01
-1.06764638e+00 -3.95874888e-01 3.74554664e-01 -7.98923075e-01
1.47366536e+00 -1.55614269e+00 5.24234995e-02 -1.67902961e-01
1.87939871e-02 1.38583243e-01 -2.57705957e-01 8.21100712e-01
1.39841229e-01 4.03030425e-01 -5.90281188e-01 -2.82723933e-01
2.08542764e-01 1.09013177e-01 -7.89004862e-01 -3.82139027e-01
2.70229220e-01 1.12079537e+00 -1.01322997e+00 -8.31638873e-01
-1.35488540e-01 -3.26488256e-01 -1.98446900e-01 2.53574282e-01
-3.40487301e-01 2.43283749e-01 -2.70450979e-01 5.26354909e-01
5.14508486e-01 -5.17051041e-01 1.30549967e-01 -2.24330038e-01
2.61956006e-01 8.96284997e-01 -9.79026675e-01 2.04624200e+00
-4.96203244e-01 5.59484303e-01 -5.85849464e-01 -8.99709582e-01
8.85072470e-01 4.48645145e-01 -6.16323063e-03 -8.99451375e-01
2.66514788e-03 3.70743632e-01 -3.48324291e-02 -7.66486943e-01
9.68098164e-01 7.59742483e-02 -4.51573938e-01 6.10874951e-01
2.69789189e-01 -6.15186214e-01 5.34748077e-01 6.80049896e-01
1.12403047e+00 -3.21277559e-01 6.27250552e-01 -1.12336822e-01
6.90109968e-01 2.36706644e-01 3.92238170e-01 1.10037947e+00
4.82093208e-02 8.28284144e-01 8.07498276e-01 1.78433824e-02
-1.04710472e+00 -1.04518414e+00 -1.27002299e-01 5.10092437e-01
3.48439366e-01 -9.30149317e-01 -5.95229626e-01 -5.83768427e-01
-1.67246401e-01 1.24685907e+00 -3.88128787e-01 -1.79796249e-01
-5.42228997e-01 -6.88227475e-01 8.17734122e-01 1.52063459e-01
8.08303535e-01 -1.06014454e+00 -3.43113869e-01 3.10099125e-01
-1.16563964e+00 -1.38603032e+00 -3.13926220e-01 -4.30612445e-01
-8.45240891e-01 -1.23351061e+00 -1.49349764e-01 -5.34302115e-01
2.71897286e-01 3.54875505e-01 1.83454514e+00 5.98903559e-02
2.00970486e-01 2.03998506e-01 -4.41551387e-01 -3.85657817e-01
-7.98782349e-01 -4.72956412e-02 -2.12305397e-01 -3.62236828e-01
1.94201421e-03 -5.44592619e-01 -4.60467279e-01 4.10688639e-01
-1.06682074e+00 3.39455336e-01 6.21128201e-01 1.11199450e+00
3.92746300e-01 5.83804911e-03 1.04582846e+00 -9.96739566e-01
1.24595332e+00 -7.96281576e-01 2.09311128e-01 8.05187464e-01
-5.21694601e-01 8.09515640e-02 9.51635122e-01 -1.18863367e-01
-1.27612638e+00 -9.77602065e-01 -7.19031841e-02 1.26970306e-01
8.71412903e-02 8.18623662e-01 -3.55545312e-01 6.70423985e-01
9.17557538e-01 2.33180553e-01 -2.81464994e-01 -1.73961878e-01
6.20569646e-01 8.10140848e-01 9.77910221e-01 -6.71495438e-01
6.93961501e-01 1.80518746e-01 -2.58578002e-01 -2.68117666e-01
-1.35114789e+00 -2.85317004e-01 -1.74382806e-01 -3.06996820e-03
1.64752200e-01 -6.97352886e-01 -4.21054751e-01 3.21824819e-01
-1.70961702e+00 1.75819859e-01 -8.50320607e-02 1.30885080e-01
-5.07975757e-01 6.62072539e-01 -3.40277791e-01 -4.43323225e-01
-9.28322732e-01 -5.05306184e-01 9.43192363e-01 5.83749637e-02
-6.97227478e-01 -7.94164240e-01 8.21753815e-02 6.99646056e-01
4.96165186e-01 3.59999955e-01 1.05934942e+00 -8.53755832e-01
-2.20542818e-01 -5.12508452e-02 -1.46081850e-01 3.18279773e-01
2.89599121e-01 9.22680125e-02 -6.87035263e-01 2.07793593e-01
4.96335059e-01 -7.33146429e-01 7.85447240e-01 -9.28552970e-02
1.23002028e+00 -9.74023402e-01 7.56788477e-02 -8.81021172e-02
1.18095481e+00 -1.13979258e-01 8.61374319e-01 2.55752206e-01
4.34255540e-01 7.12961912e-01 9.13203418e-01 3.74478698e-01
6.73922896e-01 6.86933517e-01 2.79990613e-01 3.94245505e-01
-1.21810898e-01 -4.15050328e-01 2.51018703e-01 1.15553379e+00
1.63647264e-01 -5.10526896e-01 -8.29324245e-01 4.71097350e-01
-2.13707805e+00 -1.47333539e+00 -3.19830835e-01 2.13391781e+00
1.48421013e+00 2.98592627e-01 -3.38962585e-01 3.14645678e-01
7.37003207e-01 8.22399780e-02 -4.00691479e-01 -4.64813828e-01
-5.72729468e-01 1.09933592e-01 -2.67807096e-01 4.11741793e-01
-7.30128109e-01 7.77891695e-01 5.92847061e+00 1.08495665e+00
-5.64584613e-01 4.10813205e-02 6.10605061e-01 -8.94710049e-02
-7.71498501e-01 -5.57100251e-02 -4.99212444e-01 6.67319417e-01
7.64478803e-01 -7.56137073e-01 1.83604136e-01 4.93208021e-01
2.38009430e-02 -3.55237722e-01 -1.30259740e+00 1.07574940e+00
6.32676125e-01 -1.37482560e+00 4.96758461e-01 -6.90174103e-01
1.07262611e+00 -3.03477585e-01 -7.94956908e-02 3.20255876e-01
1.82781249e-01 -1.07742858e+00 7.99040794e-01 4.51833069e-01
5.21204829e-01 -4.59624797e-01 1.07122755e+00 5.11346817e-01
-5.33196390e-01 1.86027080e-01 -4.24015224e-01 -3.83495055e-02
4.04409200e-01 9.99285817e-01 -7.69376814e-01 1.12967563e+00
5.05227566e-01 6.52756214e-01 -8.00128937e-01 8.00071478e-01
-5.40244162e-01 3.70468199e-01 -1.18433304e-01 -1.36288285e-01
6.76369742e-02 4.37184162e-02 6.93160772e-01 1.14885521e+00
3.83157641e-01 -1.00628905e-01 -4.26927388e-01 9.77901578e-01
-3.88114423e-01 1.75231248e-01 -6.82249904e-01 1.46223664e-01
7.86441684e-01 1.24111187e+00 -7.45191798e-02 -7.30724037e-01
-2.30418324e-01 9.50877070e-01 4.44883674e-01 7.53272250e-02
-9.63655293e-01 -4.71578717e-01 3.25689286e-01 -4.89801317e-01
-1.32030904e-01 1.02748394e-01 -5.66002011e-01 -1.40817881e+00
7.43252635e-01 -1.24293721e+00 5.40707529e-01 -1.06617212e+00
-1.71869028e+00 6.99527681e-01 6.53049573e-02 -1.28099144e+00
-5.98547876e-01 -8.06614757e-02 -8.90912890e-01 6.62252665e-01
-1.54904568e+00 -7.23915040e-01 -6.40628934e-01 4.64435220e-01
6.54590547e-01 1.15873292e-03 7.05727875e-01 9.00644213e-02
-4.74099010e-01 4.88431424e-01 -5.21094278e-02 -9.61911678e-02
1.05616522e+00 -1.38410378e+00 6.37932122e-01 8.50793600e-01
3.48742068e-01 8.41660500e-01 8.50704074e-01 -5.98376751e-01
-1.05611193e+00 -1.09309220e+00 1.53369796e+00 -8.80486131e-01
7.31686413e-01 8.49276967e-03 -1.08921885e+00 4.96176332e-01
6.30025744e-01 -7.40910470e-01 7.35173404e-01 4.59215678e-02
-5.69945216e-01 3.39182839e-02 -1.01657176e+00 7.31168091e-01
1.32435191e+00 -5.10412514e-01 -1.24245107e+00 7.13654578e-01
7.58533537e-01 -7.88727164e-01 -7.35983491e-01 4.03912187e-01
1.85335472e-01 -1.10221553e+00 7.73721099e-01 -9.46680844e-01
1.18539739e+00 -2.65052617e-01 -1.81335941e-01 -1.47457612e+00
1.45795709e-02 -6.33151710e-01 -1.26467615e-01 1.56817317e+00
6.44422770e-01 -5.67579687e-01 -5.28239608e-02 6.34522378e-01
-4.45527881e-01 -6.81318760e-01 -1.05751216e+00 -9.48562682e-01
1.74929768e-01 -3.40438783e-01 8.74866247e-01 1.14662874e+00
4.89336997e-01 8.65404427e-01 -2.20021963e-01 -1.76027939e-01
3.76551390e-01 5.06326079e-01 8.45801771e-01 -8.67527544e-01
-1.53587401e-01 -6.76986754e-01 -1.20044172e-01 -8.60020995e-01
7.23001594e-03 -1.12383628e+00 -1.25338370e-02 -1.96570349e+00
5.30509293e-01 -2.94432323e-02 -3.92207839e-02 4.14482892e-01
-6.15964532e-01 -7.71776661e-02 8.98827463e-02 4.50931937e-01
-8.65464807e-01 7.62860715e-01 1.23884046e+00 -4.01017100e-01
1.30297512e-01 -3.14671546e-01 -1.10777354e+00 6.19337201e-01
1.05130076e+00 -4.10363257e-01 -4.36085373e-01 -6.13000929e-01
6.15958333e-01 1.17286637e-01 5.58059990e-01 -9.63962257e-01
3.59629422e-01 -1.75538465e-01 2.22310908e-02 -6.25944376e-01
3.80777270e-02 -3.08898419e-01 2.00724229e-01 3.50527205e-02
-6.31101012e-01 3.53555173e-01 1.37455016e-02 3.38822216e-01
-7.38188267e-01 -5.13741851e-01 2.59518415e-01 -2.22136289e-01
-3.21929246e-01 -2.47097269e-01 1.60401091e-01 9.70792711e-01
5.05245745e-01 1.47998855e-01 -1.19185591e+00 -5.66046059e-01
-9.26492885e-02 4.07531232e-01 1.70797914e-01 5.47793627e-01
6.99364424e-01 -1.40660191e+00 -1.13643587e+00 -6.02082312e-01
4.05977547e-01 1.30907267e-01 3.16861719e-01 7.44365871e-01
-2.09167734e-01 7.05239356e-01 -1.83192909e-01 -4.35166329e-01
-9.44973648e-01 2.31359556e-01 1.31607309e-01 -6.97229207e-01
-3.92654896e-01 4.76510257e-01 -2.17682004e-01 -4.43885177e-01
-5.43894097e-02 -4.82785940e-01 -1.40005887e-01 5.36942296e-02
5.76242864e-01 4.72044647e-01 3.35127681e-01 -3.09971869e-01
-3.13728094e-01 4.39207077e-01 -1.51637524e-01 -2.18855247e-01
7.25108504e-01 -1.81396738e-01 -2.84801304e-01 4.51851904e-01
9.65254128e-01 -1.62998978e-02 -4.60054308e-01 -4.25737888e-01
2.74773151e-01 -4.70650285e-01 -5.88300347e-01 -1.29722524e+00
-4.34270620e-01 6.67019546e-01 -3.47313195e-01 4.58647847e-01
8.84956121e-01 8.94752666e-02 1.00304604e+00 7.97316849e-01
3.61040384e-01 -1.25953948e+00 4.77340043e-01 7.02614129e-01
1.74207509e+00 -1.42137158e+00 -2.04948038e-02 -4.70358670e-01
-1.04302430e+00 1.00294077e+00 6.69437647e-01 4.60857093e-01
-2.04422489e-01 -1.45476848e-01 4.28079441e-02 -3.29579234e-01
-1.07004440e+00 4.32397500e-02 3.89653355e-01 1.38598159e-01
4.43557560e-01 -8.17013234e-02 -6.84804082e-01 8.44274223e-01
-8.24602425e-01 -2.55332261e-01 6.93986416e-01 6.00927949e-01
-2.94587106e-01 -7.82997847e-01 -1.70200974e-01 6.76693141e-01
-3.08728904e-01 -3.43852878e-01 -7.89184272e-01 6.00922167e-01
-2.03278437e-01 1.40614104e+00 -1.13950774e-01 -2.05115363e-01
4.43395257e-01 9.87333730e-02 5.12124538e-01 -4.53654587e-01
-6.75243795e-01 -8.29794049e-01 5.28312504e-01 -5.48376381e-01
-4.92777944e-01 -5.25919735e-01 -1.12408841e+00 -4.12468314e-01
-1.59906656e-01 1.98632523e-01 5.46385884e-01 1.05234516e+00
5.62963009e-01 1.93569049e-01 5.70219517e-01 -8.64001065e-02
-7.84974873e-01 -1.32375145e+00 -1.60953283e-01 9.86287236e-01
-6.87191635e-02 -3.91593188e-01 -5.58892787e-01 -5.00141829e-02]
|
[12.00667953491211, 9.211189270019531]
|
0757e6df-fe0e-48ab-8b87-5591dc7b7f4e
|
curriculum-learning-meets-weakly-supervised
|
2212.07619
| null |
https://arxiv.org/abs/2212.07619v1
|
https://arxiv.org/pdf/2212.07619v1.pdf
|
Curriculum Learning Meets Weakly Supervised Modality Correlation Learning
|
In the field of multimodal sentiment analysis (MSA), a few studies have leveraged the inherent modality correlation information stored in samples for self-supervised learning. However, they feed the training pairs in a random order without consideration of difficulty. Without human annotation, the generated training pairs of self-supervised learning often contain noise. If noisy or hard pairs are used for training at the easy stage, the model might be stuck in bad local optimum. In this paper, we inject curriculum learning into weakly supervised modality correlation learning. The weakly supervised correlation learning leverages the label information to generate scores for negative pairs to learn a more discriminative embedding space, where negative pairs are defined as two unimodal embeddings from different samples. To assist the correlation learning, we feed the training pairs to the model according to difficulty by the proposed curriculum learning, which consists of elaborately designed scoring and feeding functions. The scoring function computes the difficulty of pairs using pre-trained and current correlation predictors, where the pairs with large losses are defined as hard pairs. Notably, the hardest pairs are discarded in our algorithm, which are assumed as noisy pairs. Moreover, the feeding function takes the difference of correlation losses as feedback to determine the feeding actions (`stay', `step back', or `step forward'). The proposed method reaches state-of-the-art performance on MSA.
|
['Haifeng Hu', 'Ya Sun', 'Sijie Mai']
|
2022-12-15
| null | null | null | null |
['multimodal-sentiment-analysis', 'multimodal-sentiment-analysis']
|
['computer-vision', 'natural-language-processing']
|
[ 2.31901497e-01 1.62305385e-01 -4.39097375e-01 -5.85732400e-01
-1.07671392e+00 -5.44319510e-01 4.00791734e-01 3.97776991e-01
-4.80629265e-01 4.67412651e-01 1.50445938e-01 -7.38349631e-02
2.02139709e-02 -6.95664644e-01 -5.45912087e-01 -9.81822968e-01
1.75091058e-01 4.45442379e-01 5.43064885e-02 -2.01029971e-01
1.17212243e-01 -3.41073215e-01 -1.44324553e+00 6.13443494e-01
1.07585204e+00 1.30529094e+00 5.32818250e-02 5.49857497e-01
-2.73069441e-01 6.70734167e-01 -2.85052270e-01 -7.68403769e-01
8.30497891e-02 -7.24734247e-01 -5.62022209e-01 -1.86460420e-01
1.98209941e-01 -4.23894934e-02 1.56346768e-01 1.15657890e+00
4.29385364e-01 1.69033796e-01 6.50991678e-01 -1.45509660e+00
-5.76262534e-01 8.06386590e-01 -5.84204137e-01 -2.18708545e-01
3.43620479e-01 1.12233274e-01 1.35380197e+00 -1.24819922e+00
3.98527175e-01 1.01507580e+00 4.92402226e-01 6.39445662e-01
-1.29007602e+00 -8.05281997e-01 2.15028599e-01 2.64908344e-01
-9.59392428e-01 -4.13335681e-01 1.21832359e+00 -3.40182632e-01
3.71152997e-01 1.03387609e-01 6.05965018e-01 1.14850092e+00
-2.36621365e-01 1.18388498e+00 1.14732635e+00 -4.11697596e-01
4.21547741e-01 6.76648915e-01 3.08477253e-01 6.70813382e-01
-2.02693060e-01 -7.26780156e-03 -7.88224876e-01 -1.38670355e-01
1.73105642e-01 -3.88190150e-02 -1.11722775e-01 -6.22147202e-01
-1.13216197e+00 9.75809455e-01 4.47812557e-01 7.94114023e-02
-9.31344777e-02 -3.68842602e-01 3.98767710e-01 5.25349319e-01
2.37990916e-01 3.97300124e-01 -6.65522039e-01 -2.90054709e-01
-5.50139248e-01 -5.90113550e-02 5.79913616e-01 6.82909131e-01
1.08045769e+00 -4.78583217e-01 -8.47522914e-02 1.14060640e+00
4.76583928e-01 3.37364227e-01 5.46367228e-01 -6.98730588e-01
9.56172764e-01 9.49870944e-01 -1.31460994e-01 -9.60556626e-01
-1.74562439e-01 -3.32420290e-01 -7.61463106e-01 -1.11345835e-01
4.70517576e-01 -3.09877217e-01 -6.42201066e-01 1.84214795e+00
3.22345734e-01 -6.46147951e-02 3.86770219e-01 1.04382920e+00
8.49523604e-01 4.90366548e-01 6.45922869e-02 -2.94340938e-01
1.10692275e+00 -1.18837976e+00 -6.35224879e-01 -2.78736055e-01
9.32806969e-01 -8.95680130e-01 1.36076701e+00 4.83871281e-01
-1.03231406e+00 -6.01733088e-01 -1.08674133e+00 1.53148413e-01
-4.50785398e-01 4.08089489e-01 5.17747402e-01 5.11536419e-01
-6.45809352e-01 6.32967055e-01 -6.55195415e-01 -1.04712479e-01
4.01620418e-01 3.50818634e-01 -3.79681438e-01 -1.39549255e-01
-1.25471652e+00 6.18734658e-01 1.40698105e-01 2.09536627e-01
-6.90817535e-01 -5.66845953e-01 -9.43196237e-01 -1.43476725e-01
1.67219207e-01 -4.54788178e-01 1.04586208e+00 -1.40690267e+00
-1.54403901e+00 7.64640152e-01 -1.67351604e-01 -3.24697606e-02
4.98674601e-01 -8.37607086e-02 -3.68251830e-01 4.95364256e-02
-3.14110587e-03 8.12024117e-01 9.02412057e-01 -1.64962113e+00
-6.27141118e-01 -2.95775712e-01 1.32472038e-01 5.69017112e-01
-8.27157617e-01 -2.76912242e-01 -5.23998439e-01 -4.53832120e-01
2.78691471e-01 -8.52002740e-01 -4.44504805e-02 -8.50386992e-02
-3.14904392e-01 -2.09754139e-01 7.40144849e-01 -3.65232557e-01
1.42462158e+00 -2.33060718e+00 4.31666970e-01 4.56172407e-01
1.21811859e-01 -1.59877110e-02 -5.09848297e-01 2.17156574e-01
-8.40114206e-02 2.01569218e-03 -2.64789879e-01 -6.52336121e-01
8.96665677e-02 1.84957445e-01 -1.24743178e-01 1.17612779e-01
5.25640428e-01 5.59278131e-01 -1.32072914e+00 -5.64590633e-01
9.01940614e-02 2.10355803e-01 -6.92890048e-01 6.51087284e-01
6.57443255e-02 2.96958864e-01 -4.04769301e-01 7.87706494e-01
7.05955505e-01 -5.82035258e-02 3.53215069e-01 -6.56372845e-01
1.66562483e-01 1.81552723e-01 -1.12479496e+00 1.54529452e+00
-4.90261018e-01 2.55707592e-01 -1.20392233e-01 -9.93306637e-01
9.70808983e-01 6.86194375e-02 2.89325953e-01 -5.67162335e-01
2.21735820e-01 2.07550526e-01 9.31089371e-02 -5.73942840e-01
5.12168586e-01 -1.87876508e-01 -2.04030186e-01 4.16318476e-01
2.40533426e-01 -6.39012828e-02 7.46870860e-02 2.06426620e-01
8.53068709e-01 2.10784018e-01 -1.81610793e-01 3.16419959e-01
6.90047026e-01 -1.96904495e-01 5.42382061e-01 3.51924628e-01
-3.49891305e-01 8.23961318e-01 8.82929683e-01 3.67702805e-02
-7.04522491e-01 -1.28484428e+00 4.98585030e-02 1.43045700e+00
1.87436253e-01 -4.40246642e-01 -5.85698485e-01 -1.14983594e+00
-1.84100866e-01 3.46619695e-01 -7.62554824e-01 -6.64949358e-01
-2.62295574e-01 -8.40436220e-01 1.43220097e-01 6.47527993e-01
2.36650005e-01 -9.19176280e-01 -1.61495693e-02 -3.05473320e-02
-2.33647510e-01 -6.34513080e-01 -4.52502310e-01 6.56764984e-01
-8.13227117e-01 -1.20979273e+00 -5.73261619e-01 -1.06711292e+00
1.12618649e+00 1.87131599e-01 9.08586025e-01 -3.14900205e-02
4.34229136e-01 6.30774796e-02 -5.75906038e-01 1.13979075e-02
-2.69385695e-01 1.79745153e-01 1.20695785e-01 3.87902230e-01
4.98940766e-01 -3.26876044e-01 -6.84480429e-01 5.07934570e-01
-6.86109543e-01 -6.42188489e-02 6.35510385e-01 1.32145584e+00
5.87569892e-01 -4.60463725e-02 5.80266476e-01 -8.55345786e-01
8.25450242e-01 -7.69296229e-01 -9.47523043e-02 1.48089767e-01
-5.64156473e-01 9.05436352e-02 9.11638737e-01 -7.39308357e-01
-8.67706716e-01 1.45412251e-01 3.03162392e-02 -5.78021765e-01
3.79922651e-02 7.02022374e-01 -3.37177008e-01 1.93494201e-01
4.05009836e-01 7.91081190e-02 9.45395306e-02 -2.52689928e-01
3.32930148e-01 6.82759225e-01 2.84216195e-01 -7.08359480e-01
7.34862149e-01 1.38256282e-01 -6.06605530e-01 -6.47972003e-02
-9.96261656e-01 -4.83619004e-01 -5.74251711e-01 -3.38837117e-01
6.72404468e-01 -7.88309872e-01 -4.65740979e-01 3.89515609e-01
-6.92175567e-01 -3.03908348e-01 -2.17877969e-01 4.41300422e-01
-2.65232801e-01 1.68595031e-01 -3.99368823e-01 -9.78639901e-01
-1.03998929e-01 -1.19270265e+00 1.04208577e+00 5.00619411e-01
-3.22607160e-01 -1.04767966e+00 1.28219828e-01 4.87398356e-01
8.70495066e-02 -2.28567213e-01 1.10071337e+00 -8.34004819e-01
-8.56252387e-02 -3.24629337e-01 -5.25256060e-02 6.94335163e-01
1.59421325e-01 5.60306571e-03 -1.22197866e+00 -1.91020891e-01
-2.56013423e-01 -8.60885799e-01 8.54141951e-01 -5.02568521e-02
9.44111347e-01 -1.09325103e-01 -1.06665738e-01 4.72249359e-01
1.15174127e+00 1.37081191e-01 4.01751757e-01 2.33917043e-01
6.87093139e-01 8.20983887e-01 8.39497149e-01 2.87430704e-01
5.98919153e-01 2.64050007e-01 3.36360872e-01 -5.28986529e-02
2.43575633e-01 -5.34511030e-01 7.51641750e-01 1.37436271e+00
2.49250710e-01 5.09270541e-02 -7.11167872e-01 4.71754581e-01
-1.96710002e+00 -7.89334953e-01 1.45168662e-01 2.38377714e+00
1.21336699e+00 3.56066376e-01 3.13469470e-02 3.66074741e-01
5.46510220e-01 1.49914891e-01 -4.83586431e-01 -4.34637725e-01
-2.34252021e-01 -5.62421158e-02 -3.17711346e-02 4.07063603e-01
-1.16916656e+00 7.61879086e-01 5.11341000e+00 7.07856774e-01
-1.11275697e+00 1.52798602e-02 7.66584098e-01 -2.01250520e-02
-6.63015902e-01 1.20256573e-01 -6.67312801e-01 6.19391978e-01
6.55693233e-01 5.17710268e-01 3.75188351e-01 7.26141155e-01
9.39886495e-02 -3.42301250e-01 -1.23361301e+00 9.06319261e-01
1.31800175e-01 -8.10311556e-01 -9.51071605e-02 -3.83368015e-01
7.85511494e-01 -2.94180423e-01 3.95266891e-01 7.08704054e-01
2.23954931e-01 -6.69260740e-01 6.58250809e-01 5.60412705e-01
3.23098391e-01 -9.34612393e-01 1.06463671e+00 2.76138216e-01
-1.03622007e+00 -2.16636956e-01 -3.75460953e-01 1.37609288e-01
-7.04103336e-02 5.99356115e-01 -6.61917865e-01 4.19696838e-01
4.83285308e-01 8.10122132e-01 -7.21219659e-01 7.02019930e-01
-2.32847705e-01 7.76692808e-01 -1.32313296e-01 -1.51851922e-01
1.31940767e-01 -5.11227667e-01 1.79929838e-01 1.00005996e+00
1.20759778e-01 -2.01133564e-01 2.32088491e-01 5.36069691e-01
-1.66110352e-01 2.41554350e-01 -3.14129680e-01 -2.15956233e-02
4.35581595e-01 1.41078520e+00 -3.86041939e-01 -2.69560456e-01
-4.26640451e-01 9.01619911e-01 5.49880087e-01 2.56080627e-01
-7.36966431e-01 -2.98701912e-01 4.13044572e-01 -2.74219036e-01
2.21324071e-01 1.73327059e-01 -6.01245701e-01 -1.11809266e+00
1.21759281e-01 -7.21198022e-01 5.38376033e-01 -6.43502176e-01
-1.82013118e+00 5.02614737e-01 -1.59649134e-01 -1.64276719e+00
-4.14564796e-02 -4.99747366e-01 -7.67784417e-01 6.61730945e-01
-1.46682000e+00 -8.48847747e-01 -2.42406115e-01 4.20944422e-01
2.67303735e-01 -3.62963855e-01 7.82562375e-01 4.09281075e-01
-6.80686116e-01 1.06445265e+00 8.44431370e-02 2.00009704e-01
1.06717348e+00 -1.40819025e+00 -5.47087193e-01 3.29981238e-01
-1.18934326e-01 5.99700809e-01 4.95320261e-01 -4.37202990e-01
-1.41400230e+00 -9.14308190e-01 8.82728159e-01 -3.38296354e-01
7.99608767e-01 -4.29022074e-01 -9.89087343e-01 1.54100567e-01
1.89756513e-01 9.58566591e-02 1.19106245e+00 2.90240347e-01
-5.60612440e-01 -4.12977457e-01 -1.13360691e+00 5.76336443e-01
6.21239126e-01 -6.79774284e-01 -5.38926661e-01 1.43782437e-01
4.48034495e-01 -1.65487841e-01 -8.31866503e-01 2.62571692e-01
6.43810749e-01 -7.91271865e-01 6.63925886e-01 -5.63078403e-01
9.52519774e-01 -3.91252428e-01 -2.79331684e-01 -1.55024874e+00
-1.11447990e-01 -1.43970340e-01 -1.93241626e-01 1.52425921e+00
8.10695171e-01 -2.99361229e-01 8.55641127e-01 4.44471091e-01
-2.06067804e-02 -1.13789952e+00 -7.08032310e-01 -3.22720259e-01
5.66061623e-02 -4.13255662e-01 3.50637108e-01 1.26921701e+00
4.13090110e-01 5.61284542e-01 -2.19017923e-01 4.06532437e-02
5.47339201e-01 6.91487268e-02 5.41692197e-01 -9.99716818e-01
-3.23638588e-01 -4.08525258e-01 -6.12857603e-02 -9.70760822e-01
1.27511099e-01 -9.55057263e-01 1.49225920e-01 -1.04875350e+00
3.78433734e-01 -6.36113822e-01 -7.22659707e-01 5.64912319e-01
-5.20309031e-01 3.32579076e-01 2.22580925e-01 1.47654802e-01
-7.96978951e-01 8.03386629e-01 1.17436373e+00 -3.59105736e-01
-3.98356676e-01 -5.36955819e-02 -7.70838082e-01 6.67080641e-01
7.83964157e-01 -3.65878403e-01 -5.34039319e-01 -3.04988176e-01
4.60411668e-01 -6.88841492e-02 1.65792089e-02 -7.43146777e-01
2.74256140e-01 -1.55968934e-01 4.57145125e-01 -4.36577618e-01
3.26944530e-01 -9.49372470e-01 -4.05358434e-01 5.27578257e-02
-7.10068166e-01 -3.32463942e-02 -1.31596267e-01 4.31306928e-01
-4.57227379e-01 -4.13734198e-01 7.15062380e-01 1.74965009e-01
-3.22393149e-01 1.15531638e-01 4.91146604e-03 9.63265821e-03
8.43010604e-01 -1.67929038e-01 -1.08535528e-01 -3.17980856e-01
-9.00618315e-01 6.28518879e-01 4.00848120e-01 5.73257148e-01
7.02855766e-01 -1.67042422e+00 -3.54414940e-01 3.76954883e-01
4.74189132e-01 1.24319278e-01 3.04418266e-01 1.01648986e+00
2.55950868e-01 -2.43373945e-01 9.46894065e-02 -7.36686289e-01
-1.07042193e+00 2.95027912e-01 1.84085086e-01 -3.00346464e-01
1.46620810e-01 9.45752859e-01 -2.51106787e-02 -8.07519138e-01
3.78654689e-01 -2.36412793e-01 -3.43150407e-01 6.67734027e-01
4.17599916e-01 6.28376529e-02 2.95348782e-02 -4.80201751e-01
-2.83991098e-01 5.02055526e-01 -2.59100288e-01 -1.55452803e-01
1.20996535e+00 -5.80926239e-02 1.74045581e-02 7.88714349e-01
1.58392596e+00 3.90472673e-02 -1.27197087e+00 -2.26157010e-01
-1.80284917e-01 -2.89802879e-01 -1.08387835e-01 -8.95020545e-01
-1.19842482e+00 9.22631025e-01 7.44366348e-01 8.37131683e-03
1.10763741e+00 8.11689794e-02 7.30296791e-01 2.14001924e-01
7.58374929e-02 -1.43589866e+00 4.80256796e-01 6.15104795e-01
5.26098311e-01 -1.62276983e+00 -3.37035418e-01 -2.23104492e-01
-1.00916767e+00 1.05160415e+00 9.60849583e-01 4.45181690e-02
5.44505596e-01 1.81210712e-01 2.34612256e-01 1.63125411e-01
-8.87974262e-01 -4.01838608e-02 3.04513276e-01 7.32028246e-01
4.77168292e-01 4.02226262e-02 -3.42887074e-01 1.14186895e+00
-1.43250033e-01 -2.68543363e-01 2.27336586e-01 7.88652658e-01
-2.69430667e-01 -1.17658353e+00 -2.61130273e-01 6.02311015e-01
1.27320334e-01 -8.94648954e-02 -5.00893354e-01 2.36811861e-01
4.36430007e-01 9.08400059e-01 7.70566761e-02 -9.04827774e-01
4.89339262e-01 2.09502876e-01 2.08460763e-01 -4.93031502e-01
-8.38750899e-01 -4.51620445e-02 2.26676669e-02 -3.55682045e-01
-4.18912172e-01 -6.43560231e-01 -1.39788926e+00 9.96510461e-02
-4.49909657e-01 3.06290954e-01 3.36214364e-01 9.36342895e-01
6.06938861e-02 3.05320352e-01 1.24004352e+00 -7.21982002e-01
-6.92162991e-01 -9.43661571e-01 -1.87713861e-01 7.43996263e-01
3.61841053e-01 -5.90576768e-01 -6.27455890e-01 -2.14499775e-02]
|
[12.961502075195312, 4.985927104949951]
|
bd07bdea-f050-4b5c-a561-5d35ffaf621e
|
dense-procedure-captioning-in-narrated
| null | null |
https://aclanthology.org/P19-1641
|
https://aclanthology.org/P19-1641.pdf
|
Dense Procedure Captioning in Narrated Instructional Videos
|
Understanding narrated instructional videos is important for both research and real-world web applications. Motivated by video dense captioning, we propose a model to generate procedure captions from narrated instructional videos which are a sequence of step-wise clips with description. Previous works on video dense captioning learn video segments and generate captions without considering transcripts. We argue that transcripts in narrated instructional videos can enhance video representation by providing fine-grained complimentary and semantic textual information. In this paper, we introduce a framework to (1) extract procedures by a cross-modality module, which fuses video content with the entire transcript; and (2) generate captions by encoding video frames as well as a snippet of transcripts within each extracted procedure. Experiments show that our model can achieve state-of-the-art performance in procedure extraction and captioning, and the ablation studies demonstrate that both the video frames and the transcripts are important for the task.
|
['Zhendong Niu', 'Botian Shi', 'Yaobo Liang', 'Nan Duan', 'Ming Zhou', 'Peng Chen', 'Lei Ji']
|
2019-07-01
| null | null | null |
acl-2019-7
|
['dense-captioning']
|
['computer-vision']
|
[ 8.80032182e-01 2.88761854e-01 -4.11720902e-01 -3.69886100e-01
-1.42401493e+00 -9.20534313e-01 5.33719063e-01 -5.85329197e-02
-5.52201904e-02 7.82042205e-01 1.03883779e+00 -4.97713983e-02
3.15154821e-01 -3.40156078e-01 -1.29027915e+00 -3.61197412e-01
8.57748464e-02 5.92374206e-02 -1.06775336e-01 1.25976190e-01
7.39160627e-02 -1.01390682e-01 -1.85467064e+00 1.38374019e+00
7.06517816e-01 6.44826889e-01 2.33119264e-01 9.23247993e-01
-4.59150583e-01 1.54122019e+00 -7.26843476e-01 -4.50924009e-01
2.00651642e-02 -9.56018567e-01 -9.80965078e-01 5.78850925e-01
8.97227407e-01 -6.57523334e-01 -6.86151862e-01 8.30309033e-01
1.34746850e-01 3.23629051e-01 4.24003303e-01 -1.33163488e+00
-7.23545015e-01 1.30029964e+00 -4.86674160e-01 3.34311455e-01
1.34516644e+00 1.62390023e-01 6.79282069e-01 -8.25507939e-01
1.12805462e+00 9.33602452e-01 2.58222967e-01 1.00064790e+00
-6.44426405e-01 -5.49115717e-01 4.23608750e-01 3.83529007e-01
-1.08131647e+00 -4.97977436e-01 5.78111231e-01 -4.74869490e-01
6.12248659e-01 3.92867506e-01 9.56901848e-01 1.82679379e+00
-2.17720404e-01 1.49746895e+00 5.29835165e-01 -3.29991579e-01
-7.47384205e-02 -1.88621014e-01 1.03421742e-02 6.89762652e-01
1.00915015e-01 -4.27179843e-01 -8.93131554e-01 3.42570871e-01
6.80459619e-01 6.88336268e-02 -7.19792902e-01 -1.45600811e-01
-1.56768954e+00 5.81894934e-01 -1.21848576e-01 2.15999275e-01
-4.58168328e-01 1.67521209e-01 5.79080403e-01 -1.03229612e-01
1.71379715e-01 6.10874891e-01 7.62290061e-02 -8.88428807e-01
-1.39947164e+00 3.50095093e-01 6.92218721e-01 1.74762809e+00
4.04743582e-01 1.34902345e-02 -9.34568465e-01 3.35857779e-01
-2.48133808e-01 4.31143612e-01 4.47400659e-01 -1.06495106e+00
9.97173250e-01 2.51519352e-01 1.53253272e-01 -6.01210594e-01
1.19704425e-01 1.83275998e-01 -2.01946884e-01 -8.06623161e-01
2.05909416e-01 -3.24335247e-01 -1.25629461e+00 1.56686354e+00
1.94261707e-02 8.51334691e-01 6.51104674e-02 9.24307227e-01
1.59626913e+00 1.01157165e+00 2.97736347e-01 -4.06375796e-01
1.53079414e+00 -1.59723902e+00 -1.25413847e+00 -4.41385150e-01
5.40744722e-01 -7.45363533e-01 9.49686766e-01 -5.71551621e-02
-1.61798751e+00 -4.91638899e-01 -8.28447282e-01 -2.73933589e-01
-5.26917353e-02 1.09959617e-02 3.27788144e-01 2.31444255e-01
-7.50939488e-01 1.22191332e-01 -7.97978163e-01 -2.02575162e-01
5.78792870e-01 -4.58348766e-02 -6.30318284e-01 -4.37627912e-01
-8.93234730e-01 4.77445185e-01 5.02134442e-01 -9.66983885e-02
-1.18087983e+00 -9.73698854e-01 -1.60913527e+00 3.65099698e-01
6.27025783e-01 -6.81621134e-01 1.75045824e+00 -1.27332473e+00
-1.27369189e+00 6.94597065e-01 -4.68377471e-01 -3.89763415e-01
1.45424336e-01 -5.34076452e-01 -2.55895078e-01 1.00051057e+00
1.50861040e-01 1.07540774e+00 8.27158391e-01 -1.23026180e+00
-6.93047523e-01 1.52061000e-01 4.05531555e-01 4.71496493e-01
-5.27503788e-01 7.74034187e-02 -8.70379627e-01 -7.20638037e-01
-3.77990425e-01 -9.55284715e-01 2.31230602e-01 -3.82813215e-01
-3.77170652e-01 3.55832517e-01 8.67008924e-01 -1.00680816e+00
1.52182090e+00 -2.07809567e+00 4.15414840e-01 -3.45781237e-01
1.95134491e-01 1.61660433e-01 -5.04629254e-01 5.87855995e-01
-2.58619517e-01 1.48931175e-01 -1.23039052e-01 -3.54258955e-01
-1.22661456e-01 -7.76040321e-03 -4.87838894e-01 1.53518111e-01
3.53079706e-01 1.01759553e+00 -1.39567375e+00 -9.33848321e-01
3.11078608e-01 7.13825047e-01 -4.99338150e-01 6.03577137e-01
-3.05531919e-01 4.10209924e-01 -5.71032822e-01 7.53207743e-01
3.79524231e-01 -2.92913467e-01 1.05599016e-01 -3.60206842e-01
5.81637695e-02 1.74140632e-01 -7.76872158e-01 2.16397834e+00
-4.12291169e-01 1.05558515e+00 -5.04909530e-02 -6.86436296e-01
2.16429114e-01 8.76118124e-01 6.85777009e-01 -4.52776551e-01
1.72732279e-01 -3.97482336e-01 -7.45908320e-01 -1.18413329e+00
7.56624520e-01 1.46937385e-01 -5.11489093e-01 3.97960663e-01
5.56367993e-01 -9.25915018e-02 8.53645504e-01 6.46854758e-01
1.23833883e+00 7.79214799e-01 3.01030606e-01 3.80246460e-01
3.53730381e-01 2.50386596e-01 1.85907423e-01 7.10645258e-01
-1.90517560e-01 1.10956430e+00 5.69847584e-01 -1.90945476e-01
-1.01074946e+00 -9.25277650e-01 3.16187710e-01 1.02787077e+00
6.18819110e-02 -8.19135308e-01 -1.17218161e+00 -8.37391913e-01
-5.78815877e-01 7.87482619e-01 -8.07552397e-01 -1.48729593e-01
-8.32308114e-01 4.44386192e-02 3.31304401e-01 7.91668177e-01
1.79898337e-01 -1.16594613e+00 -4.26821738e-01 1.38932362e-03
-1.00234032e+00 -1.76611280e+00 -9.30883527e-01 -2.98020661e-01
-4.40866441e-01 -1.06010246e+00 -8.56248558e-01 -1.19974661e+00
9.78518605e-01 6.61953688e-01 1.33663905e+00 -8.89756531e-02
1.64665598e-02 8.43637764e-01 -8.48460674e-01 -2.68458366e-01
-4.54621226e-01 -4.01056446e-02 -4.47680056e-01 -7.43316785e-02
4.79463309e-01 -1.60979897e-01 -1.94313899e-01 -1.20485649e-01
-1.30834031e+00 8.28517914e-01 4.41954702e-01 4.48700368e-01
6.31326377e-01 -6.17354393e-01 1.88468814e-01 -1.20982158e+00
4.25879121e-01 -6.88521564e-01 -1.37061745e-01 3.41824114e-01
3.43555242e-01 -6.10249043e-02 6.21325076e-01 -4.31997508e-01
-1.38038635e+00 4.72490579e-01 7.01879933e-02 -8.99166346e-01
-4.19236511e-01 5.88352382e-01 -1.50732487e-03 3.22935700e-01
3.26123327e-01 3.71734321e-01 -2.56002069e-01 -2.48681251e-02
6.19814157e-01 6.30940855e-01 9.17549670e-01 -7.52999306e-01
7.88942277e-01 2.80238807e-01 -4.14302021e-01 -7.04966307e-01
-1.35262537e+00 -7.42689610e-01 -5.22153437e-01 -7.65496254e-01
1.19329941e+00 -1.41180766e+00 -4.85372961e-01 -2.55115598e-01
-1.20385253e+00 5.08719273e-02 -3.81057024e-01 6.28058434e-01
-9.49705243e-01 3.16509128e-01 -9.35506999e-01 -4.10262436e-01
-1.17037572e-01 -1.28516746e+00 1.47378886e+00 5.93143165e-01
-3.45529139e-01 -7.00628877e-01 -1.99461475e-01 8.60387743e-01
-2.31871217e-01 6.84077024e-01 3.12205672e-01 -5.54720819e-01
-7.05918610e-01 -2.90439129e-01 -6.54156357e-02 -7.04869479e-02
5.23632690e-02 2.20735162e-01 -9.75549042e-01 -2.53110025e-02
-2.61933118e-01 -4.46119815e-01 5.73447585e-01 3.26971561e-01
1.32253170e+00 -5.26151538e-01 -2.59798944e-01 5.59261918e-01
1.21755815e+00 2.84703612e-01 8.42305064e-01 2.76624441e-01
1.07427037e+00 6.23435497e-01 7.01548934e-01 3.49951267e-01
2.34474570e-01 3.47432822e-01 2.47630268e-01 1.88820511e-01
-4.09321457e-01 -7.32818902e-01 7.36301601e-01 1.23117673e+00
4.50531067e-03 -4.36299384e-01 -7.57993221e-01 9.41415906e-01
-2.04481220e+00 -1.66582835e+00 -8.51948000e-03 1.66016781e+00
9.50009227e-01 -4.88720238e-02 5.74426278e-02 -2.47782066e-01
1.00690877e+00 2.57681996e-01 -1.08889103e-01 -3.28045398e-01
1.22848742e-01 5.83455376e-02 1.11752458e-01 1.40999109e-01
-1.16196346e+00 8.25127900e-01 6.49612808e+00 5.94148517e-01
-7.74336457e-01 -1.14895217e-01 3.13047260e-01 -5.75588167e-01
-4.11294669e-01 -2.59602875e-01 -7.23315179e-01 6.07471168e-01
1.36349082e+00 -5.52021801e-01 1.71373516e-01 8.36845577e-01
2.55301654e-01 2.37625569e-01 -1.43497145e+00 9.65894818e-01
8.68420720e-01 -1.78885591e+00 5.65492570e-01 -4.24984902e-01
1.19976759e+00 -6.13479912e-01 -1.62233561e-01 6.48004711e-01
6.54191390e-05 -8.96415114e-01 9.82466102e-01 2.46906132e-01
8.74267042e-01 -5.38853526e-01 7.69218147e-01 -9.22148004e-02
-1.44434321e+00 4.41788249e-02 2.06178516e-01 8.70545432e-02
5.82153916e-01 7.72543177e-02 -9.84973192e-01 6.18956566e-01
5.10783911e-01 1.12324071e+00 -4.69540387e-01 1.26261938e+00
-5.44399858e-01 7.03124702e-01 1.97738543e-01 3.33627500e-02
6.12000167e-01 9.91555452e-02 3.49993050e-01 1.66475630e+00
4.59443629e-01 2.53731698e-01 2.56381243e-01 4.91669595e-01
-3.84862512e-01 -4.64667976e-02 -7.92899251e-01 -5.66968858e-01
4.12965804e-01 1.28548372e+00 -7.85415232e-01 -6.43560708e-01
-8.56899440e-01 9.13988709e-01 6.99044811e-03 5.99347293e-01
-1.33792484e+00 -3.03207248e-01 5.15522420e-01 3.83914337e-02
6.17733419e-01 1.19234346e-01 1.26318946e-01 -1.54656672e+00
1.77484259e-01 -9.79730129e-01 4.57448840e-01 -1.38837540e+00
-8.51076722e-01 6.70151055e-01 3.61283213e-01 -1.80086315e+00
-6.29049242e-01 -2.53066391e-01 -3.72414172e-01 7.15226904e-02
-1.45295596e+00 -1.21056068e+00 -7.34766483e-01 6.02078199e-01
1.29063714e+00 3.48220825e-01 7.55013108e-01 5.21894276e-01
-6.74613953e-01 3.34382892e-01 -2.08956033e-01 4.93745834e-01
7.78467059e-01 -1.03092992e+00 9.81361941e-02 1.14898598e+00
3.16366732e-01 5.09090185e-01 8.77400935e-01 -6.42128408e-01
-1.59053230e+00 -1.30776322e+00 6.30562901e-01 -6.52842700e-01
4.58886057e-01 -4.21393633e-01 -6.22471929e-01 1.15661705e+00
6.92260265e-01 -1.16054416e-01 1.13893151e+00 -3.01028818e-01
-2.27788493e-01 4.70672369e-01 -7.96090782e-01 7.51998067e-01
1.09879100e+00 -8.01021159e-01 -1.03819621e+00 6.26850724e-01
1.09768534e+00 -9.26761210e-01 -7.67505586e-01 3.83452922e-02
5.91110408e-01 -5.14206648e-01 9.01291132e-01 -8.03194940e-01
1.47372186e+00 -1.63118958e-01 2.37988845e-01 -1.12625015e+00
-3.50320041e-02 -8.32458496e-01 -6.23613298e-01 1.43255901e+00
3.87995034e-01 8.74719977e-01 8.84136677e-01 6.73569322e-01
-7.31436610e-01 -3.58740151e-01 -3.51679325e-01 -4.57182437e-01
-5.04555702e-01 -3.02388221e-01 3.87286365e-01 9.56770778e-01
5.92134297e-01 4.75565940e-01 -5.91739237e-01 6.11997582e-03
2.37077057e-01 2.97974676e-01 5.29454589e-01 -5.34709454e-01
2.85961684e-02 -3.19223218e-02 -2.23795742e-01 -1.12801826e+00
5.83812475e-01 -6.36353016e-01 5.25002003e-01 -2.01670647e+00
6.02954328e-01 7.42313802e-01 6.90596700e-02 4.62559462e-01
-5.16049862e-01 2.49545500e-01 4.62792724e-01 -7.73258954e-02
-1.34448767e+00 2.87117720e-01 1.50227964e+00 -3.48538429e-01
6.31762296e-02 -4.51471984e-01 -6.56895757e-01 5.00676274e-01
1.16807610e-01 -3.64336044e-01 -7.04733074e-01 -7.03619421e-01
-1.64473392e-02 6.20931327e-01 -1.66842699e-01 -1.16501987e+00
1.67143121e-01 -3.96707654e-01 2.91988522e-01 -3.50309402e-01
4.81797367e-01 -8.11421454e-01 -1.16408281e-02 4.27192572e-04
-7.45491803e-01 2.02524811e-02 5.19543588e-01 5.79453409e-01
-7.48329163e-01 -3.99538159e-01 1.50701046e-01 -3.82399112e-01
-1.05557144e+00 2.97509342e-01 -8.78155470e-01 1.92818567e-01
1.49705660e+00 -4.50219512e-01 -6.69041514e-01 -8.60562384e-01
-7.13855326e-01 3.08847278e-01 3.99208695e-01 7.46331096e-01
7.76119471e-01 -1.44130230e+00 -7.08566368e-01 -1.63287178e-01
5.06478965e-01 -3.84924375e-02 5.56398332e-01 3.18625242e-01
-6.91030681e-01 4.83123869e-01 -4.98293817e-01 -4.15520251e-01
-1.53185034e+00 8.66222680e-01 -4.70330060e-01 4.83752117e-02
-6.40823543e-01 1.05863154e+00 2.91392028e-01 3.25933367e-01
3.94426644e-01 -5.28053343e-01 -5.16158700e-01 3.41195315e-01
1.16802526e+00 -2.21584216e-01 -2.23314106e-01 -6.66641057e-01
-3.94420028e-02 3.50855470e-01 -2.13931024e-01 4.40756679e-02
1.40598440e+00 -1.87671736e-01 3.68113726e-01 2.91224182e-01
1.28576934e+00 -7.42769102e-03 -1.45752108e+00 1.90854833e-01
-3.46589774e-01 -5.10209560e-01 -8.98743048e-02 -5.39671361e-01
-8.89037132e-01 7.19533026e-01 -1.38976485e-01 6.00069985e-02
1.20134294e+00 4.17349581e-03 1.25100958e+00 1.76398143e-01
1.40431464e-01 -9.01943803e-01 2.65514374e-01 2.69069016e-01
6.49521172e-01 -1.01455402e+00 -5.01886383e-02 -9.07286644e-01
-9.97572362e-01 1.29415822e+00 9.72584486e-01 2.41265595e-01
-1.62343174e-01 4.82215613e-01 -1.65452689e-01 -2.82853786e-02
-8.84680092e-01 -2.15967253e-01 2.90887415e-01 5.77379942e-01
6.04423106e-01 -2.19146922e-01 -7.30842873e-02 7.60688066e-01
-9.62730348e-02 3.29430759e-01 1.23772252e+00 1.47052562e+00
-2.39564434e-01 -7.00232387e-01 -4.23519790e-01 3.95803839e-01
-4.69495565e-01 -3.18681806e-01 -3.61070782e-01 5.76104999e-01
-1.32581219e-01 7.70863175e-01 3.60878229e-01 -2.54377782e-01
2.91922510e-01 3.07388306e-01 6.23650432e-01 -1.03009903e+00
-6.52663708e-01 -6.47454858e-02 4.10880566e-01 -7.73155689e-01
-8.28871191e-01 -5.86069167e-01 -1.18587887e+00 -9.36985537e-02
2.67722122e-02 6.14043236e-01 4.74895269e-01 9.50300276e-01
4.00034875e-01 9.62176681e-01 1.65709361e-01 -9.67007101e-01
9.59190652e-02 -8.15440536e-01 -1.35036800e-02 9.20259416e-01
3.73943120e-01 -4.31159884e-01 -2.12889567e-01 1.09458089e+00]
|
[10.362436294555664, 0.7279420495033264]
|
07c23fc1-6bb2-4e69-b792-9470f8c075f9
|
robust-statistics-and-no-reference-image
|
1902.03842
| null |
http://arxiv.org/abs/1902.03842v1
|
http://arxiv.org/pdf/1902.03842v1.pdf
|
Robust statistics and no-reference image quality assessment in Curvelet domain
|
This paper uses robust statistics and curvelet transform to learn a
general-purpose no-reference (NR) image quality assessment (IQA) model. The new
approach, here called M1, competes with the Curvelet Quality Assessment
proposed in 2014 (Curvelet2014). The central idea is to use descriptors based
on robust statistics to extract features and predict the human opinion about
degraded images. To show the consistency of the method the model is tested with
3 different datasets, LIVE IQA, TID2013 and CSIQ. To test evaluation, it is
used the Wilcoxon test to verify the statistical significance of results and
promote an accurate comparison between new model M1 and Curvelet2014. The
results show a gain when robust statistics are used as descriptor.
|
['Ramon Giostri Campos', 'Evandro Ottoni Teatini Salles']
|
2019-02-11
| null | null | null | null |
['no-reference-image-quality-assessment']
|
['computer-vision']
|
[-8.94180462e-02 -4.02832121e-01 2.04297706e-01 -1.11676492e-01
-1.06065476e+00 -5.25913894e-01 6.46954834e-01 2.98785210e-01
-4.82892543e-01 5.84845185e-01 3.08594882e-01 1.51654318e-01
-4.98518020e-01 -4.40203846e-01 -3.53407890e-01 -7.42779374e-01
-3.54738444e-01 -1.43562227e-01 3.03677917e-01 -2.15315953e-01
4.65146661e-01 7.82070518e-01 -1.78079069e+00 4.19530064e-01
7.79191315e-01 1.22697592e+00 -1.58050001e-01 7.58992076e-01
4.16560024e-01 5.17974496e-01 -9.13986981e-01 -4.87629563e-01
6.83947325e-01 -3.04055989e-01 -6.11054778e-01 -2.02254713e-01
4.90909994e-01 -1.75134420e-01 -1.58492878e-01 1.00214040e+00
8.94139230e-01 5.55977747e-02 7.45711803e-01 -1.22301662e+00
-3.65054786e-01 -2.02098325e-01 -2.48757243e-01 4.31141108e-01
7.96990633e-01 3.41715962e-01 7.87278473e-01 -8.46688151e-01
8.72797310e-01 1.02274954e+00 8.80952418e-01 -9.29470137e-02
-1.04655206e+00 -4.33824986e-01 -6.64485514e-01 9.44717288e-01
-1.42137659e+00 -1.22119531e-01 5.30385971e-01 -4.69461739e-01
5.25506496e-01 3.23291838e-01 6.11735702e-01 9.73691106e-01
5.61999321e-01 4.38406169e-01 2.01283240e+00 -4.63921636e-01
4.23356593e-01 1.59468085e-01 1.52600124e-01 4.91719097e-01
2.56877750e-01 6.97662175e-01 -3.64688933e-01 -9.76666734e-02
4.57975417e-01 -4.88820612e-01 -3.79870296e-01 -2.92467862e-01
-1.17196321e+00 6.07470095e-01 3.06806594e-01 7.01747596e-01
-6.08493865e-01 -2.52664357e-01 6.33353889e-01 8.31172526e-01
2.36614525e-01 2.71262765e-01 -6.67782798e-02 -2.47129828e-01
-1.19276750e+00 3.82999271e-01 7.20252872e-01 3.12488437e-01
4.54863816e-01 -3.64107266e-02 -6.47445798e-01 5.87147355e-01
6.74207658e-02 9.02264476e-01 6.15599036e-01 -1.25758076e+00
-1.54875234e-01 2.11410969e-01 1.40513808e-01 -1.35759354e+00
-1.49803922e-01 -7.11065114e-01 -7.00567126e-01 9.06204402e-01
5.21435201e-01 3.57655138e-01 -7.15490341e-01 9.51682210e-01
-1.67875722e-01 -3.95968407e-02 1.38144985e-01 8.94674957e-01
8.64906549e-01 6.54270709e-01 -1.06377535e-01 -1.97973132e-01
9.63826895e-01 -5.03455400e-01 -6.98459864e-01 7.08379924e-01
3.60373110e-01 -1.01767898e+00 9.84666407e-01 1.09026825e+00
-9.11658883e-01 -1.12961090e+00 -1.26444864e+00 3.19358677e-01
-4.36338842e-01 8.24931860e-02 6.89847618e-02 8.31090987e-01
-1.36738837e+00 8.96634519e-01 -2.34210715e-01 -4.87125427e-01
2.66161203e-01 3.94860096e-02 -6.34975910e-01 -4.53802198e-01
-1.01103854e+00 1.14616477e+00 3.30906540e-01 -3.98961037e-01
-9.68748629e-01 -5.70085287e-01 -4.71193463e-01 -2.56410129e-02
-2.38883063e-01 -3.95909429e-01 5.67378521e-01 -8.34558606e-01
-1.50510216e+00 1.14624166e+00 1.01506114e-01 -5.30864656e-01
7.49043047e-01 -2.00010855e-02 -5.35267711e-01 6.70604050e-01
-3.45121771e-02 3.78145814e-01 9.22028959e-01 -1.44199288e+00
-4.51385587e-01 -2.88937241e-01 -1.67756677e-01 -2.49494538e-01
1.71065733e-01 2.26108029e-01 -1.56963244e-01 -7.79479086e-01
-3.35540101e-02 -6.19936109e-01 3.59001786e-01 4.98180874e-02
3.33694145e-02 -1.74383730e-01 6.76609039e-01 -1.03108382e+00
9.48953331e-01 -2.09902358e+00 -3.62662345e-01 5.79381526e-01
-1.06987879e-02 3.93897802e-01 -3.75873774e-01 3.59774470e-01
-1.49602130e-01 -9.69367772e-02 -2.18449429e-01 5.95212579e-02
-2.05435678e-02 1.29858643e-01 1.35310851e-02 7.82234728e-01
8.44054222e-02 7.43273556e-01 -6.14561975e-01 -5.76605201e-01
4.45594549e-01 5.06098807e-01 -2.55970985e-01 2.18674451e-01
4.32152539e-01 5.06200492e-01 6.73721582e-02 7.47283161e-01
1.07232821e+00 2.81612903e-01 -5.33588469e-01 -8.17884505e-01
-2.97106445e-01 -4.92434114e-01 -1.25736821e+00 1.58112478e+00
-3.06413859e-01 7.85806894e-01 -1.95048794e-01 -8.86619151e-01
1.18962741e+00 4.07643080e-01 6.14549696e-01 -1.42007875e+00
2.64501005e-01 3.68359804e-01 -1.91516355e-01 -7.33504236e-01
1.34492889e-01 -4.79646139e-02 3.17215085e-01 8.48644041e-03
4.12389964e-01 -2.08269551e-01 4.49494779e-01 -3.51965204e-02
1.15175343e+00 2.03783914e-01 3.85998398e-01 -5.97772121e-01
9.85376775e-01 -1.03467338e-01 2.71235615e-01 6.64953113e-01
-7.57826984e-01 8.02793443e-01 2.69831061e-01 -2.41734758e-01
-1.13973367e+00 -1.43075097e+00 -2.43339956e-01 4.46854621e-01
7.16130361e-02 -2.02210337e-01 -5.60487032e-01 -6.88580930e-01
-1.31990999e-01 6.03614450e-01 -6.88065588e-01 1.25340337e-03
-3.21495056e-01 -5.47168434e-01 3.51021051e-01 -4.54912446e-02
8.09556484e-01 -9.77762520e-01 -7.48918653e-01 -1.78811789e-01
-2.58199960e-01 -1.07345045e+00 4.98038381e-02 -1.09073736e-01
-7.82218277e-01 -1.19036818e+00 -1.08998644e+00 -3.24168503e-01
8.13867003e-02 5.75449541e-02 1.24314904e+00 -1.55420318e-01
-4.43206787e-01 7.74455786e-01 -7.64753580e-01 -1.73400015e-01
-6.31597877e-01 -4.82552499e-01 -1.75465375e-01 5.91318794e-02
2.67193727e-02 -4.56455648e-01 -6.24562204e-01 4.64369357e-01
-1.06134176e+00 -7.62116015e-01 9.62280095e-01 5.66076994e-01
5.84364355e-01 1.36349440e-01 3.17966670e-01 -2.32411027e-01
6.43959582e-01 -1.18371680e-01 -5.18300951e-01 2.47472629e-01
-1.04786766e+00 -2.03815624e-02 4.10327345e-01 -8.97539109e-02
-6.24808133e-01 -4.67976928e-01 -1.51163533e-01 -5.14291048e-01
-2.73534864e-01 3.66750956e-01 -9.92358476e-02 -6.13024116e-01
8.63382041e-01 1.85591742e-01 -6.77357242e-02 -3.95062834e-01
1.53798521e-01 6.33016050e-01 1.02875328e+00 -2.54277229e-01
1.07510448e+00 5.48224151e-01 2.92628050e-01 -8.05035055e-01
-4.29818571e-01 -8.66674721e-01 -6.42386913e-01 -6.41026616e-01
7.68786252e-01 -6.63686275e-01 -8.55509877e-01 6.50870740e-01
-1.16662741e+00 -3.40740345e-02 -6.12065613e-01 6.89179361e-01
-7.72537112e-01 5.71904123e-01 -2.47230276e-01 -7.51897335e-01
-2.88203567e-01 -7.87256420e-01 7.06497908e-01 2.29956135e-01
2.04378635e-01 -8.52833807e-01 4.79336143e-01 2.10819095e-01
7.40969598e-01 7.51756549e-01 5.25593817e-01 -4.37516302e-01
-2.40612984e-01 -4.46226835e-01 -5.39948046e-01 9.60540950e-01
-2.57057160e-01 -6.82607759e-04 -1.14283943e+00 -4.29771215e-01
2.39522725e-01 -3.02917629e-01 5.38622439e-01 3.50223213e-01
8.46073747e-01 -6.14526682e-02 4.01971042e-01 5.18978894e-01
1.79307473e+00 2.02584907e-01 1.21771288e+00 6.39268219e-01
-7.40091726e-02 5.69640040e-01 9.24416840e-01 3.80042255e-01
8.69982168e-02 9.98918474e-01 4.69875276e-01 -1.84832856e-01
-5.51925659e-01 1.52865902e-01 6.26059830e-01 8.41261148e-01
-3.33450019e-01 -9.98771265e-02 -8.02174926e-01 3.44494939e-01
-1.13877189e+00 -1.16757023e+00 -1.84708461e-01 2.45757675e+00
3.29786807e-01 2.97706962e-01 1.75594866e-01 8.20993364e-01
4.35114026e-01 -5.25994822e-02 1.17606120e-02 -5.61268270e-01
-6.81072593e-01 5.45741439e-01 4.92890596e-01 3.02419245e-01
-1.01382971e+00 1.97594419e-01 6.52657700e+00 1.09454978e+00
-1.21309769e+00 3.19528341e-01 5.26772559e-01 4.48026091e-01
1.48273125e-01 -1.41258240e-01 -1.18627502e-02 3.88307333e-01
1.32868922e+00 -3.23581189e-01 1.79916203e-01 3.74678075e-01
5.86067319e-01 -3.52283925e-01 -6.14208519e-01 1.44940150e+00
5.68420827e-01 -8.29981625e-01 1.81477845e-01 -6.14280393e-03
6.71459436e-01 2.68836878e-02 2.09451988e-01 3.54327619e-01
-2.22207934e-01 -1.05038786e+00 7.25740671e-01 1.15225577e+00
8.64847004e-01 -5.99080324e-01 1.31693900e+00 -1.74990088e-01
-7.96269000e-01 -1.18068807e-01 -3.50452125e-01 3.01920801e-01
-7.07180426e-02 6.67889893e-01 -3.67510945e-01 1.08505857e+00
9.64642942e-01 6.74820840e-01 -1.14276659e+00 1.75981879e+00
-9.33089107e-02 7.24019706e-01 6.19399399e-02 4.57065910e-01
1.00106969e-01 -1.16737723e-01 9.10770535e-01 1.20049369e+00
5.83283126e-01 -1.30842879e-01 -1.65782481e-01 5.40181339e-01
3.63351017e-01 4.83362615e-01 -4.07788098e-01 2.70498216e-01
-5.38272299e-02 1.16733336e+00 -6.09079719e-01 -2.62318760e-01
-2.57766277e-01 1.14146650e+00 -7.14516938e-01 2.04272553e-01
-6.57352984e-01 -5.46919703e-01 -1.01188600e-01 2.25797713e-01
1.80944577e-01 -2.41193116e-01 -1.18473187e-01 -8.29026937e-01
5.12202494e-02 -1.01076782e+00 2.86805391e-01 -1.22477543e+00
-1.33210254e+00 7.65470564e-01 2.48373240e-01 -1.76178408e+00
-1.01139247e-01 -7.48886168e-01 -6.21123075e-01 9.15022552e-01
-1.62308955e+00 -1.03969336e+00 -7.27351308e-01 6.70146048e-01
5.77517524e-02 -2.36900672e-01 6.73771501e-01 3.77678812e-01
1.49998322e-01 6.07026815e-01 3.41205001e-01 -1.14366025e-01
9.07447278e-01 -1.25618839e+00 -2.07854211e-01 8.12322080e-01
1.31138740e-02 -5.04656285e-02 1.16165984e+00 -3.44085068e-01
-7.90111005e-01 -7.10686445e-01 7.61631131e-01 -4.02389526e-01
2.54695952e-01 1.98017016e-01 -8.86107802e-01 -3.74937296e-01
3.14850062e-01 4.11634445e-01 5.14491200e-01 -5.87381899e-01
-5.66516638e-01 -5.66466570e-01 -1.62991595e+00 -2.58240074e-01
4.64150697e-01 -3.90763670e-01 -5.26954174e-01 -9.30537060e-02
2.00922757e-01 2.48829111e-01 -1.33917594e+00 5.96692920e-01
6.98606431e-01 -1.67572320e+00 9.80697513e-01 8.75966847e-02
2.70034403e-01 -3.52802217e-01 -4.18680757e-01 -1.26561403e+00
7.96791539e-03 -3.96156222e-01 1.26945645e-01 1.02319050e+00
-2.14977320e-02 -5.92073977e-01 4.60017651e-01 -1.14352390e-01
2.99726762e-02 -4.01429564e-01 -1.16019642e+00 -1.39611089e+00
7.56115541e-02 -3.48383933e-01 3.38508487e-01 6.84902549e-01
-4.50295627e-01 -2.91457891e-01 -4.56449091e-02 -1.41715817e-02
9.54038143e-01 -1.68449789e-01 7.80094564e-01 -1.49295068e+00
-2.52577603e-01 -4.44228441e-01 -1.24374819e+00 -1.76146522e-01
-3.89202714e-01 -7.25153506e-01 -6.72535226e-02 -1.36855054e+00
9.92249772e-02 -1.00449957e-01 -7.62846529e-01 -6.11073524e-02
1.52135774e-01 7.34175086e-01 5.08202791e-01 3.60368878e-01
-6.70000017e-01 6.25772297e-01 1.22135389e+00 -2.11509109e-01
3.26148160e-02 4.67420965e-02 -1.39885202e-01 2.67931938e-01
5.87361515e-01 -4.41074699e-01 -2.25717165e-02 1.78990617e-01
2.38680243e-02 6.58519492e-02 8.46711397e-01 -1.89118421e+00
-9.43668485e-02 3.94594401e-01 5.51149786e-01 -6.36056542e-01
6.58032149e-02 -9.02897775e-01 2.45749373e-02 8.07993829e-01
-2.59320498e-01 3.10029328e-01 1.36088103e-01 4.25464272e-01
-6.44186795e-01 -3.24147761e-01 1.14533293e+00 1.00025617e-01
-7.22157240e-01 1.50234938e-01 -1.06832542e-01 -6.22825138e-02
8.94661188e-01 -5.70840776e-01 -1.18289039e-01 -5.89260757e-01
-9.41332698e-01 -4.34591562e-01 5.56718707e-01 1.23001419e-01
6.08603716e-01 -1.32233262e+00 -1.20853293e+00 -7.87893310e-02
5.16612649e-01 -1.08283687e+00 3.86542052e-01 1.02627110e+00
-6.71293080e-01 3.31694365e-01 -7.65430689e-01 -6.51377141e-01
-1.32561660e+00 5.95169365e-01 3.44303340e-01 -5.35350442e-01
-3.63512784e-01 2.07276702e-01 -4.04880077e-01 -4.66508679e-02
-4.54853568e-03 -1.50648057e-01 -6.13739252e-01 2.23573716e-03
5.06490529e-01 9.37446535e-01 4.75659281e-01 -9.15727854e-01
-2.14730099e-01 8.05336416e-01 4.36745286e-01 -4.53070760e-01
1.23706424e+00 -1.14204064e-01 -2.61825491e-02 4.45035934e-01
1.45936286e+00 2.30044156e-01 -1.05763948e+00 -1.49437748e-02
2.84595221e-01 -8.50937784e-01 3.02196026e-01 -1.35673201e+00
-9.15283203e-01 9.30333674e-01 1.84629452e+00 1.37127593e-01
1.46949959e+00 -4.21588928e-01 5.21397710e-01 5.53407997e-04
3.76155168e-01 -8.59162152e-01 2.32897386e-01 2.36799374e-01
1.19681597e+00 -1.21760786e+00 3.69279608e-02 -1.20878994e-01
-4.37964946e-01 1.25260806e+00 4.59819380e-03 -3.69361252e-01
7.11818397e-01 -8.47109184e-02 3.23207587e-01 -1.06510483e-01
-2.47994378e-01 -3.67818534e-01 5.51015973e-01 1.09365880e+00
1.66918471e-01 -1.95593804e-01 -7.77404010e-01 3.20443511e-01
-1.36074245e-01 3.48119974e-01 5.01516104e-01 5.68518937e-01
-2.47268111e-01 -1.09505808e+00 -6.02742314e-01 2.43470371e-01
-4.73968208e-01 3.04744482e-01 -6.91758096e-02 8.46483767e-01
5.45307696e-01 1.13770247e+00 -5.09849072e-01 -6.00575209e-01
5.90900719e-01 -2.48553067e-01 5.54584026e-01 5.98738641e-02
-8.55148613e-01 -2.16010138e-01 -3.75835091e-01 -1.00932801e+00
-7.75330603e-01 -7.57041216e-01 -5.74933767e-01 -1.80964068e-01
2.67035142e-02 2.59561867e-01 1.06554258e+00 6.56041861e-01
5.80538958e-02 2.33281612e-01 9.20441091e-01 -6.72137439e-01
-6.16172373e-01 -1.02132726e+00 -6.38556659e-01 9.30835843e-01
2.34805942e-01 -6.86455309e-01 -5.00203073e-01 2.37440392e-01]
|
[11.777389526367188, -1.931288480758667]
|
c58da3d9-dffc-4a05-a827-f8e3291a03ca
|
a-robust-attentional-framework-for-license
|
2006.03919
| null |
https://arxiv.org/abs/2006.03919v2
|
https://arxiv.org/pdf/2006.03919v2.pdf
|
A Robust Attentional Framework for License Plate Recognition in the Wild
|
Recognizing car license plates in natural scene images is an important yet still challenging task in realistic applications. Many existing approaches perform well for license plates collected under constrained conditions, eg, shooting in frontal and horizontal view-angles and under good lighting conditions. However, their performance drops significantly in an unconstrained environment that features rotation, distortion, occlusion, blurring, shading or extreme dark or bright conditions. In this work, we propose a robust framework for license plate recognition in the wild. It is composed of a tailored CycleGAN model for license plate image generation and an elaborate designed image-to-sequence network for plate recognition. On one hand, the CycleGAN based plate generation engine alleviates the exhausting human annotation work. Massive amount of training data can be obtained with a more balanced character distribution and various shooting conditions, which helps to boost the recognition accuracy to a large extent. On the other hand, the 2D attentional based license plate recognizer with an Xception-based CNN encoder is capable of recognizing license plates with different patterns under various scenarios accurately and robustly. Without using any heuristics rule or post-processing, our method achieves the state-of-the-art performance on four public datasets, which demonstrates the generality and robustness of our framework. Moreover, we released a new license plate dataset, named "CLPD", with 1200 images from all 31 provinces in mainland China. The dataset can be available from: https://github.com/wangpengnorman/CLPD_dataset.
|
['Peng Wang', 'Yanning Zhang', 'Chunhua Shen', 'Linjiang Zhang', 'Hui Li', 'Zhen Li']
|
2020-06-06
| null | null | null | null |
['license-plate-recognition']
|
['computer-vision']
|
[ 1.66725311e-02 -8.83032620e-01 4.14172746e-02 -2.63546765e-01
-7.75959611e-01 -9.64582682e-01 4.96443123e-01 -1.02343726e+00
-2.02919886e-01 5.28278828e-01 -2.25374043e-01 -1.59934461e-01
4.31984395e-01 -7.76919186e-01 -7.00526297e-01 -9.41316068e-01
7.18558133e-01 1.86906710e-01 4.40814704e-01 -3.23145688e-01
5.50788701e-01 5.68868399e-01 -1.50359094e+00 6.11851104e-02
7.59299994e-01 8.28679442e-01 4.45251316e-01 5.37032247e-01
7.76610225e-02 6.30699039e-01 -4.92328614e-01 -7.94630051e-01
7.34308362e-01 -1.15803234e-01 -4.63294908e-02 6.36246741e-01
3.53552431e-01 -4.85143065e-01 -6.64872408e-01 1.29404902e+00
5.86255193e-01 -5.91386855e-02 5.23405075e-01 -1.03643000e+00
-8.90030801e-01 -1.54240265e-01 -4.70782161e-01 4.74893814e-03
2.98780072e-02 7.30215251e-01 5.43571055e-01 -1.19565189e+00
4.57251370e-01 5.54089427e-01 6.54649973e-01 5.61326087e-01
-5.04526079e-01 -8.34972143e-01 -2.47307196e-01 1.50312871e-01
-1.60662663e+00 -7.09709466e-01 7.17522979e-01 -4.99277771e-01
5.86636424e-01 3.28957886e-01 4.62007582e-01 8.56207848e-01
-1.75333787e-02 8.10001075e-01 9.34936523e-01 -2.14515865e-01
-2.97955088e-02 1.74440056e-01 -2.10164621e-01 6.15089536e-01
3.96960258e-01 -1.40858069e-01 -6.49606576e-03 4.49322402e-01
9.64043021e-01 4.94178176e-01 -4.21605319e-01 1.51329741e-01
-9.96309817e-01 4.56728399e-01 4.39621843e-02 1.29065409e-01
-2.35245936e-02 -1.98039785e-01 1.27671480e-01 -3.95452268e-02
2.52817571e-01 1.63564965e-01 -1.50899976e-01 -1.56299192e-02
-9.92532253e-01 2.02384040e-01 5.00452042e-01 1.40017271e+00
8.04021299e-01 4.11593854e-01 -1.69865540e-04 1.08087647e+00
3.22923660e-01 8.65842581e-01 4.90123689e-01 -5.07988811e-01
8.95245969e-01 4.54081327e-01 1.12140782e-01 -1.05293024e+00
-6.23020604e-02 -3.28967541e-01 -8.59888375e-01 3.41577381e-01
2.82186657e-01 -2.51699477e-01 -1.04140985e+00 9.91754949e-01
-1.91954747e-01 1.54077858e-01 6.90851435e-02 1.13558960e+00
6.48607492e-01 9.64529693e-01 -4.88241196e-01 5.18991463e-02
1.57679307e+00 -1.18474066e+00 -6.03228271e-01 -5.52893400e-01
1.92015126e-01 -1.05668390e+00 1.02223098e+00 2.68893450e-01
-9.00591373e-01 -4.71956700e-01 -1.28021026e+00 5.51885962e-02
-2.96872914e-01 7.29691267e-01 2.32126832e-01 7.99359202e-01
-9.40969765e-01 -1.56348154e-01 -4.57454056e-01 -7.77747259e-02
4.53000724e-01 3.89386952e-01 -5.00686467e-01 -6.58896089e-01
-9.02790904e-01 7.10212409e-01 9.78414118e-02 4.46781665e-01
-7.01291502e-01 -2.97968209e-01 -7.08881676e-01 -7.35824481e-02
3.95853996e-01 -1.04825959e-01 9.13389027e-01 -9.64737117e-01
-1.68038571e+00 9.71902192e-01 3.99538055e-02 1.62834629e-01
6.70090914e-01 8.77554640e-02 -7.91454971e-01 -2.67407019e-02
-4.29729261e-02 3.20471257e-01 8.09191823e-01 -1.12890422e+00
-5.53223372e-01 -3.30073059e-01 -1.25629395e-01 3.92575800e-01
-2.25350305e-01 2.25919247e-01 -1.24699163e+00 -6.11860812e-01
-1.10273376e-01 -1.18255222e+00 1.78990569e-02 -2.46007636e-01
-5.85164249e-01 1.67124152e-01 7.48858154e-01 -7.96804726e-01
7.17401505e-01 -2.39873672e+00 -7.03810751e-01 4.38542031e-02
-4.88097332e-02 7.97300875e-01 -4.90122847e-03 4.63313550e-01
4.14680988e-02 2.40425631e-01 -2.99222142e-01 -2.75546491e-01
7.85046518e-02 -2.45819092e-01 -2.15314046e-01 6.71289682e-01
3.59588474e-01 7.19694972e-01 -2.78800458e-01 -3.02068293e-01
3.94856900e-01 2.64821142e-01 -2.58993298e-01 2.10680187e-01
1.76514700e-01 1.25789121e-01 -3.52438390e-01 1.11711621e+00
1.34259582e+00 -1.18659239e-03 -1.44619182e-01 1.44813061e-01
-3.83035898e-01 -3.78722459e-01 -1.05369759e+00 1.20759559e+00
-1.48300275e-01 9.30047512e-01 2.73128580e-02 -5.30770481e-01
1.33263564e+00 2.45055199e-01 -8.34546909e-02 -7.90576637e-01
4.92477380e-02 4.97796327e-01 -2.99506366e-01 -7.36666262e-01
6.09693944e-01 -9.89240259e-02 -1.18825890e-01 -5.58771119e-02
-3.38925928e-01 -1.18012160e-01 2.64409214e-01 -2.24182084e-01
7.89430261e-01 -1.53525323e-01 -8.98690969e-02 5.93160503e-02
9.02910113e-01 4.01584767e-02 7.22685754e-01 3.99470568e-01
-2.50235438e-01 1.03405857e+00 1.80883065e-01 -4.19535190e-01
-1.41302276e+00 -7.12263405e-01 -1.37435328e-02 6.36800826e-01
4.08960760e-01 3.13330531e-01 -6.47890747e-01 -2.04652652e-01
-3.61974567e-01 3.69618863e-01 -9.74251926e-02 1.25497073e-01
-7.27840960e-01 -8.52388620e-01 1.00913596e+00 4.24127251e-01
1.20242918e+00 -8.43639076e-01 1.63979411e-01 -1.66513801e-01
-8.36323649e-02 -1.36864662e+00 -9.04807687e-01 -3.99038255e-01
-3.77095580e-01 -9.67716932e-01 -1.03239465e+00 -1.20022917e+00
7.04984784e-01 6.56852007e-01 7.57211924e-01 9.85553712e-02
1.02252280e-02 -3.88993055e-01 -1.84504300e-01 -2.96579957e-01
-1.99322984e-01 -1.16027288e-01 -3.25165801e-02 3.34141552e-01
4.06712502e-01 -1.34351268e-01 -6.44405007e-01 7.06153274e-01
-1.12044990e+00 1.65011808e-01 8.53988945e-01 5.17710924e-01
2.25151017e-01 2.55687982e-01 3.02511811e-01 -4.82044339e-01
4.82550353e-01 -3.63960803e-01 -9.69982922e-01 2.37625659e-01
-2.49289721e-01 -5.48531830e-01 8.72557819e-01 -8.15310329e-02
-1.18020356e+00 2.61790723e-01 -4.56087261e-01 -4.66870517e-01
-4.13280606e-01 7.54132271e-02 -6.62185550e-01 -1.76541954e-01
1.68679193e-01 9.62629557e-01 4.35946435e-02 -2.02479333e-01
-1.84801474e-01 1.12324476e+00 6.87119961e-01 -1.14094026e-01
1.14494467e+00 4.17634577e-01 -3.24029088e-01 -9.66095388e-01
-1.62195623e-01 -4.78657931e-01 -4.92481947e-01 -4.98629123e-01
8.21951032e-01 -1.06902742e+00 -8.63655031e-01 1.25339293e+00
-9.52808380e-01 -7.67821968e-02 4.96515095e-01 4.29540128e-01
-9.90374610e-02 5.23556292e-01 -7.47556210e-01 -8.13461900e-01
-1.74014628e-01 -1.34841859e+00 9.72552121e-01 5.08848667e-01
7.49140143e-01 -6.16587102e-01 -5.70348538e-02 7.00582862e-01
4.88819480e-01 1.33207981e-02 4.57711339e-01 -6.72187448e-01
-1.03781164e+00 -5.90940237e-01 -4.64148551e-01 5.93290687e-01
-1.02767507e-02 2.60957152e-01 -1.06467581e+00 2.10398957e-02
-1.23086914e-01 -1.98475301e-01 7.28503585e-01 1.92707077e-01
8.69190037e-01 -3.64262253e-01 6.42396882e-02 7.93891490e-01
1.81293869e+00 5.67062557e-01 1.19044781e+00 5.56481481e-01
7.25143194e-01 3.74754965e-01 4.19407219e-01 4.13075149e-01
2.48930901e-01 6.85919225e-01 3.94944549e-01 -1.56729028e-01
8.52737799e-02 -1.72508240e-01 5.26109993e-01 8.32559824e-01
-2.98601180e-01 -7.97481775e-01 -1.08246863e+00 2.17372477e-01
-1.38514841e+00 -1.24281859e+00 -3.10764581e-01 2.06335878e+00
5.29588878e-01 -5.89472428e-02 -2.40786523e-02 5.65871932e-02
1.28515112e+00 1.32184163e-01 -4.99634564e-01 1.02196947e-01
-4.72574443e-01 -4.41408902e-01 8.90803576e-01 1.59755901e-01
-1.10010672e+00 1.03336966e+00 5.36780262e+00 1.05387318e+00
-1.43821883e+00 -2.62463212e-01 7.67329395e-01 1.61585853e-01
-3.48981842e-02 -2.25998640e-01 -9.49282944e-01 8.82306516e-01
4.55025792e-01 8.94223303e-02 3.99135083e-01 7.32564449e-01
1.72206119e-01 2.32318833e-01 -3.77518326e-01 1.29278672e+00
5.01074255e-01 -1.44658601e+00 -2.40360454e-01 3.54574054e-01
6.83422208e-01 1.67094156e-01 3.40629011e-01 1.68670580e-01
-2.77214535e-02 -9.79769826e-01 7.06552744e-01 5.77914059e-01
1.23249793e+00 -7.73751557e-01 1.03154194e+00 5.15909612e-01
-9.68713880e-01 3.33791366e-03 -6.07258677e-01 1.48997620e-01
8.78734794e-03 3.85854602e-01 -8.80150259e-01 4.37542915e-01
2.69578367e-01 7.49020636e-01 -5.58312714e-01 1.20973551e+00
-1.42001092e-01 6.80740237e-01 -5.86880036e-02 -1.71154559e-01
4.63842660e-01 -4.90439147e-01 2.91658163e-01 1.35602748e+00
6.98445857e-01 1.96719661e-01 -9.51504484e-02 6.03852510e-01
-1.65948212e-01 -7.37852752e-02 -6.57880306e-01 1.15665436e-01
5.34075558e-01 1.33545589e+00 -6.42242253e-01 -1.05229974e-01
-6.21541917e-01 9.56454575e-01 -2.31214926e-01 4.57287937e-01
-1.03252506e+00 -5.62484622e-01 4.49821711e-01 1.15500838e-01
4.46005344e-01 -3.07245255e-01 -1.14673994e-01 -1.56559944e+00
2.45227098e-01 -8.84638071e-01 3.44934575e-02 -9.57696140e-01
-1.13089824e+00 6.04498565e-01 -5.02853572e-01 -1.81693590e+00
8.39704946e-02 -1.11296391e+00 -7.93921888e-01 7.90233314e-01
-1.62019002e+00 -1.11763585e+00 -5.50755382e-01 8.31519127e-01
9.49299395e-01 -6.67805910e-01 3.40965390e-01 5.34277320e-01
-1.04915202e+00 5.57058811e-01 6.38786376e-01 7.71482587e-01
7.87372053e-01 -7.53199577e-01 4.26966220e-01 1.29975665e+00
-3.62438560e-01 2.94220239e-01 2.90422142e-01 -3.93671334e-01
-1.69947314e+00 -1.26669014e+00 5.87703586e-01 -1.09289184e-01
3.33221674e-01 -5.53352952e-01 -8.09689522e-01 5.78189194e-01
2.53729582e-01 4.24006395e-02 6.61403656e-01 -7.36777306e-01
-1.02131642e-01 -2.72822261e-01 -1.01168132e+00 6.26601696e-01
7.33966410e-01 -2.00978145e-01 -1.88268900e-01 4.45664525e-01
2.38558695e-01 -4.41182435e-01 -4.10827339e-01 -8.18526223e-02
5.97049832e-01 -1.03420699e+00 8.53704631e-01 6.13041408e-02
7.19490469e-01 -6.79647863e-01 -3.14103991e-01 -8.92734647e-01
-3.57776850e-01 -4.11997497e-01 6.41196549e-01 1.51682615e+00
3.47984821e-01 -8.47139299e-01 7.74164498e-01 7.82215536e-01
-5.67486227e-01 -4.87704664e-01 -6.67091250e-01 -8.01880598e-01
-1.04997978e-01 -3.28679323e-01 8.36595178e-01 7.80510008e-01
-4.34459060e-01 -2.11579651e-02 -8.21211278e-01 4.00223643e-01
4.34160203e-01 1.92896258e-02 9.35254216e-01 -7.07710981e-01
4.19277698e-02 -4.29031730e-01 -7.91310489e-01 -1.07316327e+00
5.64749204e-02 -6.38628066e-01 2.98970044e-01 -1.23166716e+00
1.24948315e-01 -4.02278692e-01 9.85920653e-02 2.01508597e-01
-1.09560974e-01 7.83814728e-01 4.61190403e-01 8.40611637e-01
-5.14636457e-01 3.30892056e-01 1.23801839e+00 -3.17043513e-01
1.87955230e-01 1.48855984e-01 -6.69244170e-01 7.06799150e-01
9.45174694e-01 -1.82104692e-01 -1.47574022e-01 -7.18073666e-01
1.18170314e-01 6.92725852e-02 2.05556199e-01 -1.02068257e+00
6.33295596e-01 -1.75185770e-01 6.51462078e-01 -5.96342444e-01
3.99000376e-01 -6.52984619e-01 2.40823567e-01 2.59150237e-01
1.81047872e-01 1.85968935e-01 1.18107013e-01 3.88850659e-01
-4.64996845e-01 -4.77138400e-01 6.61276579e-01 -2.70829946e-01
-1.17353678e+00 5.18291712e-01 -5.58413923e-01 -7.38003179e-02
1.16089404e+00 -6.97195351e-01 -6.14120007e-01 -2.74552971e-01
8.68348256e-02 3.62462527e-03 9.01008725e-01 3.55595827e-01
7.55650401e-01 -1.37125504e+00 -1.10842752e+00 5.14561236e-01
2.63854086e-01 -8.20282623e-02 5.77410817e-01 5.29886842e-01
-1.28377700e+00 5.87817550e-01 -3.93488348e-01 -4.74926800e-01
-1.04371929e+00 2.98740119e-01 4.02884752e-01 2.58441865e-02
-3.50567788e-01 6.30566478e-01 2.14913920e-01 -4.59626287e-01
-3.92172009e-01 2.44869992e-01 -8.53604823e-02 -2.52142340e-01
6.41189694e-01 3.21241230e-01 2.73555934e-01 -8.77764523e-01
-4.16954130e-01 7.58437634e-01 -4.17810269e-02 -1.11356884e-01
1.34062374e+00 -2.92916019e-02 9.11239907e-02 -6.17764257e-02
1.15914035e+00 3.05068135e-01 -1.45607722e+00 6.40577376e-02
-4.44250822e-01 -9.55463052e-01 -3.13836515e-01 -5.48020959e-01
-1.28912830e+00 6.76254988e-01 2.85140097e-01 -6.74206540e-02
1.16794562e+00 -4.77287054e-01 7.94965684e-01 4.02732998e-01
2.31256649e-01 -1.06271112e+00 -1.49152860e-01 5.54993451e-01
7.25833833e-01 -1.25233924e+00 -2.05241933e-01 -2.82259583e-01
-1.07041144e+00 1.32093251e+00 6.80242836e-01 -2.35733911e-01
1.44387409e-01 4.20377910e-01 2.78698534e-01 1.12044878e-01
-4.51772481e-01 -3.27656157e-02 3.90825905e-02 5.53053439e-01
2.04464436e-01 6.60644332e-03 1.32889062e-01 4.57740158e-01
-1.06906682e-01 -1.91002429e-01 8.30688655e-01 6.50921643e-01
-4.08937812e-01 -6.89991355e-01 -7.63901055e-01 3.13392252e-01
-5.09681344e-01 -1.94222242e-01 -1.98983997e-01 1.01689970e+00
2.25152567e-01 9.27464843e-01 2.06313714e-01 -5.11387467e-01
2.29629517e-01 -2.78889816e-02 -8.61483514e-02 -3.38307172e-01
-1.97175801e-01 1.44353837e-01 -8.34249258e-02 -4.59370986e-02
-2.74664164e-01 -8.42243612e-01 -8.68090630e-01 -5.67608356e-01
-5.00356197e-01 -2.83418506e-01 7.34203458e-01 8.52310479e-01
2.67696768e-01 1.11922540e-01 9.56752658e-01 -7.91111708e-01
-4.12784219e-01 -8.54243934e-01 -8.60825479e-01 1.18876576e-01
1.92867965e-01 -2.07262591e-01 -2.34023228e-01 3.17760050e-01]
|
[9.85318374633789, -4.921056747436523]
|
5fb4edad-6eec-4c4e-b5b8-090e08ec4da6
|
implicit-behavioral-cloning
|
2109.00137
| null |
https://arxiv.org/abs/2109.00137v1
|
https://arxiv.org/pdf/2109.00137v1.pdf
|
Implicit Behavioral Cloning
|
We find that across a wide range of robot policy learning scenarios, treating supervised policy learning with an implicit model generally performs better, on average, than commonly used explicit models. We present extensive experiments on this finding, and we provide both intuitive insight and theoretical arguments distinguishing the properties of implicit models compared to their explicit counterparts, particularly with respect to approximating complex, potentially discontinuous and multi-valued (set-valued) functions. On robotic policy learning tasks we show that implicit behavioral cloning policies with energy-based models (EBM) often outperform common explicit (Mean Square Error, or Mixture Density) behavioral cloning policies, including on tasks with high-dimensional action spaces and visual image inputs. We find these policies provide competitive results or outperform state-of-the-art offline reinforcement learning methods on the challenging human-expert tasks from the D4RL benchmark suite, despite using no reward information. In the real world, robots with implicit policies can learn complex and remarkably subtle behaviors on contact-rich tasks from human demonstrations, including tasks with high combinatorial complexity and tasks requiring 1mm precision.
|
['Jonathan Tompson', 'Igor Mordatch', 'Johnny Lee', 'Adrian Wong', 'Laura Downs', 'Ayzaan Wahid', 'Oscar Ramirez', 'Andy Zeng', 'Corey Lynch', 'Pete Florence']
|
2021-09-01
| null | null | null | null |
['d4rl']
|
['robots']
|
[ 3.05261780e-02 3.75114113e-01 -5.80855668e-01 2.88384198e-03
-5.58777928e-01 -5.65241933e-01 8.09310436e-01 -1.89377889e-01
-7.78862715e-01 1.21525300e+00 -1.49198964e-01 -3.27934384e-01
-4.71966743e-01 -2.49442514e-02 -1.14637566e+00 -1.05720246e+00
-5.50456941e-01 9.15739775e-01 2.75476128e-02 -3.23189706e-01
3.62825185e-01 3.81453544e-01 -1.49223137e+00 -1.79306030e-01
7.49869943e-01 1.02172196e+00 5.40244460e-01 7.19414353e-01
3.98820162e-01 1.14053118e+00 -3.47985327e-01 1.24258056e-01
4.57412720e-01 1.26963615e-01 -8.14375103e-01 -1.94803253e-01
1.18016161e-01 -5.89292586e-01 -4.84265029e-01 9.46370065e-01
5.06456196e-01 3.94500941e-01 1.06155562e+00 -1.57550967e+00
-8.25256348e-01 4.11042780e-01 -4.33311701e-01 -1.97087422e-01
-6.30694702e-02 1.02216172e+00 7.29793429e-01 -5.17850816e-01
5.84703624e-01 1.60354042e+00 9.20622110e-01 9.08382356e-01
-1.58302712e+00 -3.79223049e-01 5.76713622e-01 3.03452034e-02
-5.90482950e-01 -4.63753402e-01 1.06851570e-01 -4.63256568e-01
1.50831962e+00 -4.34476733e-01 5.09209275e-01 1.64974451e+00
3.83012831e-01 1.01617527e+00 1.34703267e+00 -3.21877748e-02
4.72598732e-01 -1.21811435e-01 -1.73918381e-01 9.69637334e-01
2.83379048e-01 8.64589870e-01 -2.09254175e-01 -4.47332978e-01
9.60887432e-01 -2.63322294e-01 -1.55623704e-01 -1.18508852e+00
-1.30177188e+00 7.80256271e-01 4.82318372e-01 -1.93672732e-01
-6.51028216e-01 1.15528274e+00 5.67843497e-01 5.80422819e-01
-1.12230577e-01 9.61290896e-01 -9.33825910e-01 -4.99623090e-01
-1.91221163e-02 6.94004893e-01 9.95821893e-01 1.33299291e+00
7.16145337e-01 2.88101614e-01 -2.17067897e-01 7.04669654e-01
1.86476260e-01 6.65952861e-01 4.56115067e-01 -1.78371727e+00
2.17513174e-01 -1.14058286e-01 8.72618556e-01 -1.78096086e-01
-6.07764184e-01 -1.21403493e-01 -6.39072955e-01 9.69558656e-01
4.19294238e-01 -3.82211566e-01 -1.16820908e+00 2.04280782e+00
7.28909522e-02 -1.12340651e-01 1.46505967e-01 9.26616907e-01
9.76812541e-02 5.65201283e-01 4.43550140e-01 -1.61663711e-01
7.19523489e-01 -1.14832926e+00 -5.49102545e-01 -3.17011058e-01
8.28481793e-01 -3.07611916e-02 1.38176870e+00 5.08946478e-01
-9.46689725e-01 -3.42325270e-01 -8.41209590e-01 -7.59279877e-02
-2.71594971e-01 1.80358723e-01 8.19505274e-01 -8.27410147e-02
-1.16580212e+00 1.02061510e+00 -1.04627144e+00 -4.97958899e-01
2.20526576e-01 6.45363510e-01 -1.92605466e-01 2.59164840e-01
-6.80114627e-01 1.46803784e+00 3.45171660e-01 -2.21117005e-01
-1.76059902e+00 -7.38774180e-01 -7.57596195e-01 -1.70369446e-01
7.97512352e-01 -6.35159373e-01 2.12804842e+00 -1.04007554e+00
-2.11942744e+00 4.69290376e-01 4.27890629e-01 -8.59741926e-01
5.89469850e-01 -5.82968771e-01 5.05232275e-01 -1.17451161e-01
1.75375715e-01 1.11351979e+00 1.03024602e+00 -1.32297671e+00
-5.12275934e-01 -5.07790968e-02 1.96707308e-01 3.48359883e-01
1.85601681e-01 -7.36828387e-01 3.45132738e-01 -3.63774955e-01
-7.06764340e-01 -1.30531216e+00 -6.63794577e-01 4.06620741e-01
-9.03352536e-03 -4.58656847e-01 8.68432403e-01 -2.88297087e-01
1.06927671e-01 -1.86625350e+00 6.82167768e-01 -3.34493041e-01
-1.04111105e-01 1.59957305e-01 -3.36633861e-01 2.66093433e-01
4.16928232e-01 -2.71525741e-01 -3.84957105e-01 -3.04800004e-01
7.12706506e-01 8.21064830e-01 -4.83979106e-01 6.13334835e-01
1.95640102e-01 1.09357691e+00 -1.31913483e+00 -3.06002647e-01
1.68473482e-01 -8.30810741e-02 -5.24984539e-01 3.11049491e-01
-9.19081986e-01 6.87216818e-01 -6.92854106e-01 5.23517668e-01
1.27744883e-01 -1.73509583e-01 2.68718988e-01 4.73150522e-01
-5.45156449e-02 2.49368057e-01 -5.45692742e-01 1.95397961e+00
-4.33370143e-01 4.72630948e-01 6.44431770e-01 -1.34979892e+00
5.23973405e-01 2.33786300e-01 7.10892856e-01 -7.83610940e-01
7.48535544e-02 3.95860463e-01 1.58040792e-01 -4.04507756e-01
3.45031351e-01 -7.06129596e-02 -1.24264836e-01 2.81026542e-01
5.23244560e-01 -5.24664164e-01 -3.48189138e-02 -1.61616847e-01
1.47411275e+00 9.16979730e-01 3.07693332e-01 -5.79695106e-01
1.04428306e-02 5.05304456e-01 3.31622422e-01 1.16869628e+00
-6.05104804e-01 -2.41665542e-01 7.02795208e-01 -2.36460581e-01
-1.28446269e+00 -1.10856688e+00 1.68982893e-01 1.39957595e+00
2.34130576e-01 8.15728307e-02 -3.82767975e-01 -7.17825353e-01
7.79628813e-01 8.64744961e-01 -8.84702325e-01 -4.40049917e-01
-8.01618576e-01 -3.02432299e-01 5.04601419e-01 6.64407849e-01
1.79023996e-01 -1.48983133e+00 -1.10834622e+00 4.35082287e-01
3.80275458e-01 -1.05588257e+00 -2.36546353e-01 9.39545870e-01
-1.02628195e+00 -1.16182292e+00 -9.36820865e-01 -7.29171515e-01
4.03821886e-01 4.09511179e-02 1.13083303e+00 -1.76913247e-01
-3.95342767e-01 1.08803439e+00 -3.47021908e-01 -4.63794142e-01
-4.39571321e-01 -5.11137992e-02 4.77627486e-01 -8.24332654e-01
-1.02404088e-01 -5.29480636e-01 -4.56151962e-01 2.51150399e-01
-2.34873131e-01 -1.13887332e-01 1.02461779e+00 1.31772459e+00
5.92933893e-01 -6.19683146e-01 7.29375482e-01 -2.75622398e-01
8.21851790e-01 -5.80988646e-01 -9.67554629e-01 1.03819616e-01
-7.48522758e-01 6.50037050e-01 7.85849929e-01 -1.06278300e+00
-1.09702313e+00 1.23598069e-01 3.05955231e-01 -6.82264209e-01
-1.12152487e-01 -1.63316786e-01 5.39702654e-01 -1.59322917e-01
8.27820241e-01 3.20075870e-01 4.89339948e-01 -4.30513799e-01
8.03040564e-01 1.15211777e-01 6.48805380e-01 -1.41154826e+00
4.95063335e-01 2.47794524e-01 2.16095939e-01 -5.38682759e-01
-3.88249397e-01 -1.90094486e-01 -4.18171167e-01 2.60749962e-02
8.90709519e-01 -7.13133633e-01 -1.08072257e+00 3.77289176e-01
-8.68505776e-01 -1.62377989e+00 -6.87276959e-01 5.12104869e-01
-1.78506005e+00 -4.71685827e-02 -7.36260593e-01 -9.94763434e-01
-8.35207552e-02 -1.35716343e+00 1.20369875e+00 -2.75549777e-02
-7.04118833e-02 -7.95676053e-01 3.10220629e-01 -6.00162983e-01
4.31540042e-01 9.07187313e-02 1.20677388e+00 -2.60508448e-01
-5.46623468e-01 6.39422596e-01 9.41670612e-02 2.58089751e-01
1.88493337e-02 -4.03306425e-01 -5.70297241e-01 -5.01694024e-01
-2.70228356e-01 -1.19708395e+00 8.22261214e-01 6.48583531e-01
1.24175334e+00 -2.80985415e-01 -6.25577867e-01 5.18661022e-01
1.06312549e+00 2.39024639e-01 2.57874221e-01 5.41363358e-01
3.02838922e-01 4.90008086e-01 1.01422381e+00 5.07318854e-01
9.57086831e-02 5.18882573e-01 8.81309390e-01 2.86424816e-01
1.15223952e-01 -3.67336005e-01 8.19037437e-01 -2.71588508e-02
-2.06503049e-02 8.03379342e-02 -4.93400335e-01 5.51490605e-01
-2.43906450e+00 -8.82318497e-01 3.25327694e-01 1.92424691e+00
8.96993041e-01 -5.01067340e-02 3.11041415e-01 -6.12549841e-01
2.99146742e-01 -1.57274798e-01 -1.49038088e+00 -6.28713846e-01
1.97798476e-01 1.09210044e-01 9.32039797e-01 1.96326658e-01
-1.10444903e+00 1.03551960e+00 7.09479952e+00 6.10756099e-01
-8.59022379e-01 1.35434017e-01 4.07599300e-01 -3.61189753e-01
2.91179687e-01 -2.47653961e-01 -4.77139592e-01 2.54888564e-01
7.08323240e-01 -7.67318439e-03 9.66591775e-01 1.47650206e+00
-6.72666654e-02 -2.12668389e-01 -1.69227219e+00 9.68683064e-01
-7.54354596e-01 -1.12460697e+00 -4.80060250e-01 3.15017492e-01
9.02328134e-01 5.65704048e-01 4.05572563e-01 1.12977862e+00
1.33221972e+00 -1.27704251e+00 1.02645564e+00 4.23129916e-01
7.92654872e-01 -3.41361493e-01 2.59989530e-01 7.59336531e-01
-8.26424599e-01 -6.81502223e-01 -7.09012091e-01 -1.88959222e-02
-8.55251849e-02 -4.16069478e-01 -7.65178621e-01 6.39852211e-02
1.04619563e+00 8.39033127e-01 1.50852680e-01 8.45135450e-01
-9.69703943e-02 1.57045439e-01 -3.23152721e-01 -4.57974046e-01
8.74838650e-01 -2.05760553e-01 6.38299763e-01 9.99902666e-01
2.25703180e-01 9.27027687e-02 2.02951565e-01 9.92249191e-01
2.38474607e-02 -4.55587298e-01 -1.09790659e+00 -2.58763045e-01
4.68899608e-02 1.00640023e+00 -1.97085515e-01 -2.77874410e-01
-7.09665939e-02 8.40174854e-01 9.75408733e-01 5.79865336e-01
-9.47780490e-01 9.52503085e-02 1.23635602e+00 -3.60503048e-01
5.30537188e-01 -6.28946781e-01 1.62027299e-01 -7.48078108e-01
-2.59561449e-01 -8.62516999e-01 7.37832338e-02 -8.12135756e-01
-1.44549274e+00 -1.15162000e-01 2.82129109e-01 -1.12681651e+00
-6.91877782e-01 -1.33836699e+00 -2.82229543e-01 3.96476299e-01
-1.32733941e+00 -5.27614057e-01 7.23871216e-02 5.11631787e-01
6.46773219e-01 -2.67407179e-01 7.63272464e-01 -2.91625321e-01
2.64660623e-02 3.01251262e-01 5.15479088e-01 -2.10888028e-01
7.40136325e-01 -1.56291711e+00 2.77905911e-01 2.40798723e-02
-2.81687379e-01 4.01097536e-01 8.25645745e-01 -4.89020735e-01
-2.07483959e+00 -9.54686999e-01 -7.01223612e-01 -4.60955054e-01
7.26020694e-01 -1.75391585e-01 -6.05642021e-01 1.04200327e+00
3.72857898e-01 5.25743850e-02 -2.55258411e-01 -3.32907960e-02
-6.75541982e-02 3.68337661e-01 -1.26760101e+00 7.07708001e-01
1.53013647e+00 -1.68314233e-01 -7.23368168e-01 3.91665280e-01
9.20454681e-01 -3.53682697e-01 -6.51818037e-01 7.08334982e-01
6.37051463e-01 -5.49493790e-01 9.07686174e-01 -1.11116159e+00
1.61368057e-01 7.65393600e-02 -1.70042396e-01 -1.88345826e+00
-4.96691942e-01 -8.32162440e-01 -6.90345287e-01 3.04603606e-01
2.19804123e-01 -9.04652953e-01 6.03931904e-01 1.23583801e-01
-4.49192375e-01 -1.04259753e+00 -8.11029553e-01 -1.38375473e+00
6.09990716e-01 -1.61763933e-02 1.89982370e-01 5.23814142e-01
1.09583579e-01 6.47027642e-02 -3.81857812e-01 -1.60143420e-01
7.61212349e-01 -1.33650124e-01 9.34402704e-01 -9.36745942e-01
-5.40553212e-01 -8.14708114e-01 7.07531273e-02 -1.66562271e+00
6.81060493e-01 -6.35143697e-01 6.73559368e-01 -1.45821309e+00
-7.45228529e-02 -9.13775027e-01 -2.12009355e-01 5.01630723e-01
3.25959384e-01 -6.23337448e-01 1.04860254e-01 2.25451186e-01
-7.45753109e-01 1.29160094e+00 1.47559857e+00 -3.52426976e-01
-5.45885116e-02 -2.38141611e-01 -3.84182692e-01 8.15891683e-01
8.35934997e-01 -4.37079549e-01 -7.40730524e-01 -5.73790133e-01
-6.57967627e-02 1.56074703e-01 6.76193178e-01 -9.23127651e-01
-4.31185439e-02 -6.99034095e-01 2.87266970e-01 -1.78525716e-01
7.39021957e-01 -6.98684216e-01 -3.98577422e-01 1.00444067e+00
-6.30731404e-01 8.96085575e-02 3.83375674e-01 1.04092169e+00
4.52793598e-01 -1.95331141e-01 8.37926745e-01 -4.26930428e-01
-1.09900427e+00 6.04964852e-01 -6.68802559e-01 2.42225826e-01
1.22275066e+00 6.50697425e-02 -6.32700980e-01 -3.50193650e-01
-7.55505741e-01 6.89392984e-01 6.90970838e-01 4.98014599e-01
3.78733158e-01 -1.17602932e+00 -3.97435129e-01 -4.63486940e-01
6.03399947e-02 -7.15304539e-02 -3.26696992e-01 9.98313725e-01
-2.10742772e-01 5.12687027e-01 -5.33694863e-01 -6.85589850e-01
-6.19442582e-01 7.95382142e-01 6.67543411e-01 -3.06442320e-01
-8.23207855e-01 7.51012862e-01 5.10197878e-01 -9.24518824e-01
5.70903361e-01 -8.28440309e-01 2.73522019e-01 -6.39997482e-01
-1.23597719e-01 4.10856605e-01 -6.68922782e-01 -1.31942496e-01
-1.11344613e-01 3.55573386e-01 2.41098866e-01 -2.58735657e-01
1.33865738e+00 3.37884247e-01 3.06937724e-01 4.82105166e-01
8.49738896e-01 -8.98290277e-01 -2.22391510e+00 -1.31585911e-01
6.04729317e-02 1.58536492e-03 -4.28178042e-01 -1.00418079e+00
-4.20678288e-01 8.97492886e-01 5.90968311e-01 -1.99269056e-02
4.87422317e-01 8.91283453e-02 3.23446453e-01 1.29416585e+00
1.08444929e+00 -1.52394974e+00 5.62494218e-01 9.41580832e-01
1.13600731e+00 -1.40673876e+00 -1.16021158e-02 3.26615959e-01
-9.47095573e-01 9.30331469e-01 1.04218459e+00 -6.33820057e-01
4.30800557e-01 3.27945054e-01 -3.64126980e-01 -3.83575656e-03
-1.34979153e+00 -2.89521068e-01 -1.40699312e-01 1.13415289e+00
-2.02637807e-01 1.60643473e-01 1.17329538e-01 1.38177991e-01
-8.57583713e-03 -1.79959968e-01 9.94039327e-02 1.21629679e+00
-8.50565255e-01 -8.67873073e-01 -3.77445482e-02 4.42002356e-01
-4.51889783e-02 3.63038510e-01 -2.72620201e-01 1.24003613e+00
-2.82232195e-01 6.57215059e-01 -7.09909424e-02 9.80704185e-03
9.26111266e-02 -4.34310213e-02 1.16264760e+00 -4.53619808e-01
-3.17047507e-01 -4.35333282e-01 1.34272665e-01 -1.12293041e+00
-2.55984157e-01 -6.50480330e-01 -1.65420711e+00 -1.15030594e-01
1.92427352e-01 -3.35834511e-02 7.55498111e-01 9.64624107e-01
3.86180699e-01 3.79995316e-01 7.50148818e-02 -1.55506837e+00
-1.67768538e+00 -1.17381978e+00 -3.30363899e-01 2.84768373e-01
7.32974589e-01 -1.45500910e+00 -3.06490213e-01 -2.92968690e-01]
|
[4.3894243240356445, 1.03813898563385]
|
081be8ef-5c1d-4d65-babf-fac5432383c1
|
deep-brain-state-classification-of-meg-data
|
2007.00897
| null |
https://arxiv.org/abs/2007.00897v2
|
https://arxiv.org/pdf/2007.00897v2.pdf
|
Deep brain state classification of MEG data
|
Neuroimaging techniques have shown to be useful when studying the brain's activity. This paper uses Magnetoencephalography (MEG) data, provided by the Human Connectome Project (HCP), in combination with various deep artificial neural network models to perform brain decoding. More specifically, here we investigate to which extent can we infer the task performed by a subject based on its MEG data. Three models based on compact convolution, combined convolutional and long short-term architecture as well as a model based on multi-view learning that aims at fusing the outputs of the two stream networks are proposed and examined. These models exploit the spatio-temporal MEG data for learning new representations that are used to decode the relevant tasks across subjects. In order to realize the most relevant features of the input signals, two attention mechanisms, i.e. self and global attention, are incorporated in all the models. The experimental results of cross subject multi-class classification on the studied MEG dataset show that the inclusion of attention improves the generalization of the models across subjects.
|
['Jesus Garcia Fernandez', 'Siamak Mehrkanoon', 'Ismail Alaoui Abdellaoui', 'Caner Sahinli']
|
2020-07-02
| null | null | null | null |
['brain-decoding', 'brain-decoding']
|
['medical', 'miscellaneous']
|
[ 5.36005311e-02 -9.79653969e-02 5.65736890e-01 -6.97340667e-01
-3.52682732e-02 -1.31959781e-01 7.07629383e-01 1.26780391e-01
-5.03766894e-01 7.76678622e-01 4.78659004e-01 1.17747031e-01
-4.57957029e-01 -5.77830255e-01 -5.48899651e-01 -7.24043846e-01
-2.24874243e-01 2.34316304e-01 1.22148305e-01 -1.53955922e-01
4.95029986e-01 5.77988386e-01 -1.80499208e+00 7.43230820e-01
7.56777823e-01 1.02393603e+00 4.63999629e-01 2.74650186e-01
3.70455645e-02 3.88847888e-01 -4.29882824e-01 -1.95594937e-01
-6.66305870e-02 -5.03691554e-01 -6.11575067e-01 -3.47328037e-01
5.58497421e-02 -7.40538388e-02 -1.96205944e-01 8.84537101e-01
6.28068924e-01 1.19615547e-01 8.34470928e-01 -7.04404235e-01
-5.01297474e-01 5.89747727e-01 -4.07540083e-01 1.03304088e+00
2.57484555e-01 9.41851512e-02 4.54218149e-01 -8.57933939e-01
4.52371746e-01 9.93595243e-01 4.21799064e-01 3.20799977e-01
-9.91231382e-01 -6.43549621e-01 2.43164286e-01 6.41457021e-01
-1.08684111e+00 -5.64453185e-01 7.33973205e-01 -6.61862075e-01
9.66951132e-01 3.38238329e-02 8.83355319e-01 1.35766208e+00
8.84803355e-01 2.88048118e-01 1.64939272e+00 -1.78022847e-01
2.44163290e-01 2.25369319e-01 5.38837254e-01 2.34868944e-01
1.84862018e-01 9.30439010e-02 -6.64648473e-01 7.28588377e-04
5.20109117e-01 4.91047278e-02 -3.50722969e-01 2.05803383e-03
-1.33537793e+00 5.34260035e-01 6.61730826e-01 9.42573607e-01
-1.05567062e+00 -1.73085123e-01 5.09406984e-01 7.93880224e-02
5.10762274e-01 3.84880215e-01 -5.18373311e-01 2.08394095e-01
-1.00706637e+00 -1.72794938e-01 3.80465865e-01 3.30258012e-01
5.40070832e-01 1.17003515e-01 -4.56866115e-01 5.96651971e-01
4.99629468e-01 2.11170390e-01 1.06064034e+00 -3.55763763e-01
5.11039615e-01 4.52067256e-01 -4.44907069e-01 -9.79679525e-01
-7.63769567e-01 -6.73667014e-01 -7.49983311e-01 -3.32470499e-02
2.21292064e-01 -2.18685165e-01 -8.09505105e-01 1.69684744e+00
-8.73732120e-02 2.02852547e-01 3.42428125e-02 7.43712366e-01
6.82646096e-01 1.86134011e-01 4.22186404e-01 -1.74536958e-01
1.65390277e+00 -6.42724156e-01 -7.56572008e-01 -4.57954139e-01
5.81303656e-01 -9.35622975e-02 3.25683445e-01 3.28981280e-01
-1.28474915e+00 -9.10002232e-01 -1.14735425e+00 1.44651830e-01
-8.14264834e-01 -4.75432761e-02 6.48487508e-01 4.88082767e-01
-1.17417693e+00 6.59861267e-01 -7.71812499e-01 -3.99843723e-01
7.04079390e-01 3.06336313e-01 -5.51548243e-01 1.22593604e-02
-1.40996230e+00 1.30495322e+00 5.32337904e-01 4.18608487e-01
-8.77439141e-01 -5.94400048e-01 -5.55867732e-01 4.79359597e-01
-3.22012216e-01 -9.16320860e-01 4.51542467e-01 -1.17074227e+00
-1.33836257e+00 7.90731430e-01 -2.55110025e-01 -3.25655043e-01
1.11350417e-01 1.83926046e-01 -6.03917897e-01 3.89272362e-01
9.40202028e-02 5.84106207e-01 7.48753667e-01 -7.63064444e-01
-3.83544713e-01 -1.20940554e+00 -3.63124251e-01 7.05344677e-02
-2.72600472e-01 1.45449549e-01 5.84184006e-02 -5.45104742e-01
3.63363713e-01 -6.30163610e-01 1.56284347e-01 -5.29634774e-01
-9.27235708e-02 -2.84978241e-01 4.04171586e-01 -1.17780960e+00
8.91085327e-01 -2.18681860e+00 4.34066981e-01 2.93533862e-01
2.42396161e-01 1.18269302e-01 -5.00813276e-02 3.29625070e-01
-5.30082107e-01 -1.00994237e-01 -2.29160324e-01 1.65957715e-02
-2.28362679e-01 -6.27159625e-02 6.83451369e-02 6.42234683e-01
1.14370555e-01 9.94354248e-01 -6.67209327e-01 3.79945263e-02
-5.73199801e-03 3.42199445e-01 -3.55839282e-01 8.03765282e-02
3.60125422e-01 9.12613153e-01 -2.91493654e-01 9.96465385e-02
7.67519116e-01 7.47915581e-02 7.74232969e-02 -3.08175206e-01
-7.57080503e-03 2.86469638e-01 -5.71529329e-01 1.96242654e+00
-3.18889499e-01 6.20322764e-01 -7.04786321e-03 -1.57626927e+00
9.16210949e-01 6.40308738e-01 2.53589123e-01 -1.15576053e+00
4.84416455e-01 3.73240747e-02 7.07715094e-01 -1.01541746e+00
-2.73092628e-01 -9.13113132e-02 2.38068447e-01 5.27254760e-01
7.03002572e-01 5.10102630e-01 2.15056568e-01 -1.65531069e-01
7.68366337e-01 1.24627531e-01 2.16615692e-01 -6.54464066e-01
6.94180012e-01 -6.04898274e-01 2.43005618e-01 4.80493098e-01
-2.12852255e-01 3.25757176e-01 5.65267682e-01 -4.07134712e-01
-6.58572793e-01 -6.90395653e-01 -1.87732339e-01 1.05756307e+00
-3.22964787e-01 2.01942056e-01 -8.18074048e-01 -3.83517981e-01
-2.75833309e-01 6.11318290e-01 -9.96304214e-01 -5.23577631e-01
-4.94720131e-01 -1.08929777e+00 3.85564327e-01 7.68211544e-01
6.99105442e-01 -1.20522928e+00 -1.09471524e+00 2.73741543e-01
1.10838622e-01 -1.05797064e+00 1.37881264e-01 2.97392488e-01
-1.13011086e+00 -1.03533757e+00 -7.55835593e-01 -6.58705890e-01
4.74384725e-01 -7.59506300e-02 7.69633412e-01 -1.26007646e-01
-2.31603265e-01 3.55748326e-01 -2.85316050e-01 -3.55025738e-01
4.71468754e-02 1.55680716e-01 -4.53569889e-02 5.69664419e-01
5.68745255e-01 -9.92151558e-01 -7.32432663e-01 3.12265866e-02
-7.95167625e-01 -6.55627716e-03 9.30134773e-01 5.64040124e-01
2.95731220e-02 -4.13819015e-01 1.12572241e+00 -8.43616366e-01
7.53999233e-01 -1.04260325e+00 -1.05532341e-01 2.97288835e-01
-3.41692358e-01 1.70202404e-01 2.92492449e-01 -2.56425560e-01
-1.22152555e+00 -2.98913449e-01 -2.27755263e-01 -1.54982537e-01
-6.62522793e-01 7.57020056e-01 -2.01346844e-01 3.25146616e-02
5.39078534e-01 5.05311608e-01 3.39341313e-02 -5.26960492e-01
4.27052788e-02 5.83078146e-01 1.99836910e-01 -1.72682181e-01
-9.61641371e-02 3.90901327e-01 -5.13995327e-02 -6.80894971e-01
-4.31212515e-01 -2.36583248e-01 -1.09935045e+00 -3.88878077e-01
1.17730200e+00 -8.59489322e-01 -6.60635173e-01 5.49237490e-01
-1.34087729e+00 4.02104855e-02 1.71045586e-01 9.72949147e-01
-4.30473298e-01 -6.01406693e-02 -4.22641188e-01 -5.25696874e-01
-3.25960815e-01 -1.10602450e+00 7.17804074e-01 3.23071212e-01
-4.21149395e-02 -1.11401153e+00 -1.06981598e-01 1.78884596e-01
6.77721798e-01 7.99461156e-02 1.11677241e+00 -1.04456663e+00
-1.08304292e-01 1.52130723e-02 -2.23815590e-01 4.25289683e-02
-1.00603826e-01 -6.52206838e-01 -1.42167902e+00 -1.19311847e-01
7.31963515e-01 1.90190628e-01 1.17395270e+00 6.42705798e-01
1.23663056e+00 7.56339058e-02 -4.66090590e-01 5.04430115e-01
1.25069857e+00 3.28388095e-01 1.00139296e+00 3.65502052e-02
2.82108366e-01 1.03672075e+00 -1.77445039e-01 2.06939623e-01
3.37653339e-01 3.77574712e-01 4.49742138e-01 3.64898235e-01
-5.72729371e-02 2.34523460e-01 2.10431412e-01 9.41583514e-01
-3.74478787e-01 4.58232239e-02 -8.70255530e-01 4.94180709e-01
-1.63673973e+00 -1.03711259e+00 -2.76059121e-01 2.06768107e+00
1.38235927e-01 7.86466151e-02 -2.61318743e-01 8.76879320e-02
6.27093017e-01 6.53065890e-02 -5.28690696e-01 -6.59733534e-01
-1.12978920e-01 4.82812971e-01 6.66577220e-02 -3.14105451e-02
-6.87082827e-01 2.75362492e-01 6.12389517e+00 3.35327029e-01
-1.28434861e+00 5.83750486e-01 5.40029526e-01 -7.05939233e-02
-7.76859000e-02 -3.86039674e-01 -3.76510173e-01 6.46858156e-01
1.51839089e+00 -4.76670219e-03 5.03345609e-01 1.66998178e-01
2.84651220e-01 -3.90447706e-01 -1.01832950e+00 8.56642067e-01
4.29126769e-01 -1.04918706e+00 -8.36774185e-02 7.12066442e-02
3.24638546e-01 1.60678700e-01 2.20971778e-02 3.22318912e-01
-6.51337862e-01 -1.02649331e+00 5.26106477e-01 1.30329394e+00
2.90188342e-01 -3.40084672e-01 8.66046965e-01 6.37372077e-01
-8.46584618e-01 -3.13796192e-01 -1.79489687e-01 -2.89751291e-01
2.46812142e-02 3.74723911e-01 -3.20908040e-01 5.88509619e-01
6.59628749e-01 8.78796756e-01 -9.75396395e-01 1.14166963e+00
-1.96729198e-01 2.83293545e-01 1.57785058e-01 -6.73016012e-02
6.37522042e-02 -5.61605357e-02 3.43872666e-01 9.87596452e-01
4.92336661e-01 2.39028528e-01 -4.88444895e-01 1.27628922e+00
1.17874674e-01 7.89018720e-02 -7.56905735e-01 1.81491613e-01
7.54912570e-02 1.36022568e+00 -8.00754845e-01 -2.99816668e-01
-2.42355376e-01 9.11450684e-01 3.56972545e-01 5.31257212e-01
-6.86517298e-01 -3.72027516e-01 1.85794234e-01 6.96122572e-02
3.55338424e-01 -2.39978865e-01 -4.17226374e-01 -1.06195867e+00
-3.27351727e-02 -3.13082308e-01 1.80555657e-01 -1.11344993e+00
-1.21923590e+00 8.77346218e-01 2.37782791e-01 -5.79369426e-01
-3.63314182e-01 -7.21411169e-01 -7.58274794e-01 1.51893198e+00
-1.43468726e+00 -1.05120707e+00 -2.66951799e-01 7.66080737e-01
4.86616284e-01 -1.92171797e-01 9.26387668e-01 5.80416083e-01
-5.08383930e-01 1.84364125e-01 -1.44926861e-01 -1.00958250e-01
5.05953193e-01 -6.77644253e-01 -1.41605571e-01 6.71715140e-01
9.95065179e-03 6.87963545e-01 1.94804668e-01 -5.81337869e-01
-1.06319809e+00 -7.44854629e-01 9.52049017e-01 -3.97263885e-01
2.71463156e-01 -4.04033065e-01 -1.26284337e+00 7.43563712e-01
6.16129458e-01 1.09570555e-01 6.89824581e-01 -6.84772059e-02
-1.23624183e-01 -1.02285491e-02 -1.22316146e+00 -2.40096413e-02
9.34558392e-01 -5.18373132e-01 -1.09511137e+00 1.56035513e-01
2.17977569e-01 8.92338753e-02 -9.47844386e-01 2.50091106e-01
5.86435318e-01 -1.21685636e+00 7.79382586e-01 -9.72390354e-01
2.48392314e-01 2.03842923e-01 3.31998318e-02 -1.74218869e+00
-6.60856605e-01 2.70443410e-01 -7.29272366e-02 1.03067303e+00
3.11340630e-01 -1.11343312e+00 5.20293787e-02 4.19112295e-01
-2.90745705e-01 -6.80234194e-01 -1.03499675e+00 -2.84369081e-01
4.83689979e-02 -2.47026384e-01 5.57183385e-01 7.98125803e-01
1.57391459e-01 3.52872312e-01 6.88095391e-02 1.93669885e-01
2.01610640e-01 -1.92952961e-01 -8.83191824e-02 -1.56793475e+00
9.94094759e-02 -4.66907769e-01 -6.82492793e-01 -2.75012016e-01
4.11916405e-01 -1.21258748e+00 -3.14379752e-01 -1.52882886e+00
2.78681338e-01 3.94296087e-02 -6.54939711e-01 2.63366759e-01
-2.56959498e-02 2.15709686e-01 9.05282050e-03 9.82402731e-03
-1.33897111e-01 4.74859625e-01 1.06824231e+00 1.56725928e-01
4.12313864e-02 -3.11012752e-02 -7.07148075e-01 6.65150821e-01
9.95678604e-01 -3.60931247e-01 -2.02446893e-01 -6.19562685e-01
5.02000153e-02 1.29888296e-01 5.03883123e-01 -1.45961249e+00
3.70359004e-01 4.86265182e-01 1.00121367e+00 -2.61079192e-01
4.05827135e-01 -7.35586822e-01 5.29850274e-02 5.97790241e-01
-2.72433132e-01 2.86203325e-01 4.45468009e-01 5.46431124e-01
-2.81012177e-01 -2.66987830e-01 6.81648374e-01 -3.62024248e-01
-5.60553551e-01 2.18358561e-01 -5.73828936e-01 -2.38216549e-01
1.00603211e+00 -2.87565172e-01 -3.57808173e-01 2.81861853e-02
-1.28572810e+00 -8.03341642e-02 -5.42255402e-01 4.56896335e-01
5.54914355e-01 -1.29222226e+00 -7.12118924e-01 6.94514811e-01
-6.71864673e-02 -8.94476593e-01 7.23739088e-01 1.46627498e+00
-2.02951021e-02 6.75221920e-01 -9.62666035e-01 -3.16571534e-01
-6.44730151e-01 6.59082711e-01 5.89564264e-01 -2.48906352e-02
-4.80888695e-01 6.06465459e-01 1.90285847e-01 -3.09708584e-02
-2.02555567e-01 -3.58318508e-01 -1.03062749e+00 7.15024233e-01
6.66849017e-01 2.82010972e-01 4.17953044e-01 -8.63341749e-01
-3.13688934e-01 2.95538008e-01 7.08544552e-02 -2.22167015e-01
1.70316792e+00 -1.47531763e-01 -3.89435977e-01 7.06150711e-01
1.14918411e+00 -4.83919233e-01 -7.75938034e-01 2.85190810e-02
1.22554429e-01 -1.58305332e-01 2.10200951e-01 -9.22948658e-01
-1.21500194e+00 1.54063237e+00 9.41352785e-01 1.72874093e-01
1.24843335e+00 -1.17747784e-01 3.92597020e-01 -9.04186592e-02
4.35694426e-01 -9.81894433e-01 -2.24372238e-01 4.65255260e-01
1.00035346e+00 -9.19995904e-01 -3.18430424e-01 1.25572801e-01
-4.05865610e-01 1.38474655e+00 6.08718574e-01 -4.14272010e-01
8.62370670e-01 -3.20397049e-01 -5.59104502e-01 -5.21418571e-01
-5.51740825e-01 -1.80400535e-01 3.96869123e-01 4.79651004e-01
4.13672864e-01 -8.72678831e-02 -5.88409901e-01 1.14459312e+00
1.83250114e-01 1.20109811e-01 3.74095410e-01 5.90297699e-01
-4.65591788e-01 -5.85734129e-01 -3.62181962e-01 8.67124915e-01
-4.01320159e-01 -8.98535326e-02 8.22417215e-02 5.55516660e-01
3.83659184e-01 6.75152659e-01 1.93604603e-01 -1.02170013e-01
3.66935343e-01 7.80479670e-01 5.83343744e-01 -6.84549451e-01
-9.72067297e-01 -3.35013956e-01 -1.81843862e-01 -4.96123612e-01
-6.50442779e-01 -7.08047688e-01 -9.43978429e-01 1.32432923e-01
-8.44797939e-02 1.12155583e-02 9.43058431e-01 1.43012249e+00
6.91572368e-01 1.13067877e+00 3.19910318e-01 -1.20374405e+00
-1.70578882e-01 -1.50930631e+00 -7.91823566e-01 3.99991810e-01
1.07974164e-01 -8.94322574e-01 -1.61562681e-01 -5.28800823e-02]
|
[12.68569278717041, 3.3846426010131836]
|
a49f2c10-d4d2-49d9-bf91-a1f7ecee14cd
|
dynamic-observation-policies-in-observation
|
2307.02620
| null |
https://arxiv.org/abs/2307.02620v1
|
https://arxiv.org/pdf/2307.02620v1.pdf
|
Dynamic Observation Policies in Observation Cost-Sensitive Reinforcement Learning
|
Reinforcement learning (RL) has been shown to learn sophisticated control policies for complex tasks including games, robotics, heating and cooling systems and text generation. The action-perception cycle in RL, however, generally assumes that a measurement of the state of the environment is available at each time step without a cost. In applications such as deep-sea and planetary robot exploration, materials design and medicine, however, there can be a high cost associated with measuring, or even approximating, the state of the environment. In this paper, we survey the recently growing literature that adopts the perspective that an RL agent might not need, or even want, a costly measurement at each time step. Within this context, we propose the Deep Dynamic Multi-Step Observationless Agent (DMSOA), contrast it with the literature and empirically evaluate it on OpenAI gym and Atari Pong environments. Our results, show that DMSOA learns a better policy with fewer decision steps and measurements than the considered alternative from the literature.
|
['Isaac Tamblyn', 'Mark Crowley', 'Colin Bellinger']
|
2023-07-05
| null | null | null | null |
['reinforcement-learning-1', 'text-generation', 'openai-gym']
|
['methodology', 'natural-language-processing', 'playing-games']
|
[ 1.11229308e-01 3.86895180e-01 3.13063823e-02 1.60112038e-01
-1.60444185e-01 -5.87527454e-01 6.69804335e-01 3.41835976e-01
-1.00866461e+00 1.32385635e+00 -3.65324587e-01 -1.59560725e-01
-4.34236884e-01 -8.91283095e-01 -8.05310130e-01 -9.71973896e-01
-7.94470087e-02 7.42556632e-01 1.20170325e-01 -3.07494819e-01
1.23033591e-01 4.94046569e-01 -1.83510911e+00 -3.62710357e-01
6.74580693e-01 8.83155525e-01 6.46745145e-01 7.57343769e-01
4.11118090e-01 8.37668240e-01 -4.24915165e-01 3.73010635e-01
2.90247619e-01 -4.78965312e-01 -8.09551835e-01 1.01289138e-01
-3.87005001e-01 -4.10994977e-01 3.53805944e-02 8.70178699e-01
6.12220705e-01 6.11067414e-01 4.37295228e-01 -1.04613340e+00
2.17885479e-01 4.21192408e-01 -1.06035858e-01 -1.88176721e-01
3.66392434e-01 5.30812085e-01 5.75401127e-01 -5.57586923e-02
8.26472342e-01 1.01188207e+00 1.38051406e-01 6.64328396e-01
-1.14705658e+00 -3.39603543e-01 2.20387667e-01 3.29042435e-01
-1.11704814e+00 -8.30718800e-02 5.04560709e-01 -2.34812170e-01
8.19507658e-01 2.14767575e-01 1.05269372e+00 1.25031102e+00
2.02026173e-01 6.05772197e-01 1.77115178e+00 -3.25696111e-01
1.14030838e+00 -1.22056743e-02 -6.13846898e-01 5.22527516e-01
2.60452390e-01 4.52232182e-01 -2.70912677e-01 1.37264073e-01
6.01865709e-01 -3.94020468e-01 -2.24322543e-01 -5.88003099e-01
-9.85741258e-01 6.85151339e-01 2.84287333e-01 2.99214661e-01
-9.09378052e-01 2.99837679e-01 2.32176393e-01 4.83802885e-01
-7.51831681e-02 7.65614331e-01 -3.41381907e-01 -6.06531203e-01
-5.29100716e-01 5.44825494e-01 1.08042181e+00 4.71185684e-01
7.13514507e-01 2.24087909e-01 1.54835388e-01 1.68067813e-01
5.94634488e-02 4.33433890e-01 5.32509089e-01 -1.09988964e+00
1.13293514e-01 3.17809582e-01 7.00152278e-01 -5.22621214e-01
-7.11670399e-01 -1.65119469e-01 -5.06681621e-01 9.65359032e-01
4.21418846e-01 -5.98120689e-01 -6.32112563e-01 1.79613209e+00
8.07714403e-01 3.07272911e-01 4.28789556e-01 1.18176949e+00
9.28386524e-02 4.85495120e-01 -8.61825868e-02 -4.11839038e-01
1.07965386e+00 -4.77327794e-01 -6.97478950e-01 -3.45708400e-01
3.83818448e-01 -3.01100761e-01 8.69556785e-01 9.80395675e-01
-1.03190088e+00 -3.49935621e-01 -1.17103040e+00 2.99645334e-01
-5.28580070e-01 -9.54397470e-02 4.59324598e-01 4.01841044e-01
-8.90957713e-01 1.00261950e+00 -9.75683272e-01 -5.89414597e-01
-2.17926830e-01 4.34939563e-01 -3.07415992e-01 9.67348069e-02
-1.06207657e+00 1.34986234e+00 6.69498265e-01 2.45145574e-01
-1.36291003e+00 -2.38750532e-01 -5.21310508e-01 -1.95514113e-02
9.74919438e-01 -3.66749257e-01 1.55214655e+00 -8.74320686e-01
-2.48877668e+00 3.96919906e-01 3.97688448e-01 -6.95532203e-01
9.14274275e-01 -4.10167664e-01 2.90418178e-01 7.11489394e-02
-3.94862473e-01 4.70474809e-01 6.76519394e-01 -1.08880854e+00
-4.64143068e-01 -3.32017660e-01 7.43197381e-01 6.76814258e-01
3.56781244e-01 -6.37906015e-01 3.90788198e-01 8.10900424e-03
-1.10800549e-01 -1.19701409e+00 -6.19374156e-01 -4.52172942e-02
-1.33471489e-01 -1.63793996e-01 5.69334626e-01 -2.35570982e-01
5.12359321e-01 -1.75775838e+00 6.65237784e-01 1.44704998e-01
-2.58916497e-01 1.06593683e-01 1.62045941e-01 7.04499662e-01
2.74933994e-01 -3.26638460e-01 -2.03846276e-01 -2.07162634e-01
2.17601582e-01 8.16243887e-01 -2.78800484e-02 5.69832683e-01
-3.23552936e-01 5.36378920e-01 -1.11837626e+00 -2.56250024e-01
5.59981704e-01 2.65648752e-01 -3.72943193e-01 2.45235562e-01
-7.19829082e-01 7.26226211e-01 -7.06549406e-01 1.90614760e-01
2.27233931e-01 2.38414109e-01 5.50465107e-01 3.55233699e-01
-6.75166547e-01 1.99470744e-01 -1.48796356e+00 1.93651485e+00
-7.20083058e-01 4.08012390e-01 3.03713381e-01 -1.36297059e+00
7.59379447e-01 4.78791416e-01 5.72640717e-01 -1.10799122e+00
4.39944416e-01 3.78442645e-01 2.23373681e-01 -5.33169985e-01
5.15657902e-01 -2.31002465e-01 9.43097845e-02 3.50165188e-01
-2.10372508e-01 -6.14431024e-01 1.83296934e-01 -4.85839427e-01
1.08469915e+00 7.07082033e-01 4.73958373e-01 -3.26625913e-01
5.99581540e-01 7.26510435e-02 3.72662574e-01 8.19019496e-01
-5.51090240e-02 -7.01467972e-03 5.51539242e-01 -2.67632753e-01
-1.00909483e+00 -8.25009584e-01 2.46368587e-01 7.71341681e-01
2.08229691e-01 1.85546800e-01 -8.04018199e-01 -3.63920212e-01
-7.80381709e-02 8.19903135e-01 -6.72671974e-01 -1.75432593e-01
-6.93942130e-01 -5.37095666e-01 4.64830287e-02 8.77285302e-02
3.76981735e-01 -1.40399671e+00 -1.73902583e+00 4.92650449e-01
7.07142279e-02 -9.30195630e-01 3.55710179e-01 5.01022279e-01
-7.72761405e-01 -1.07390893e+00 -4.10380721e-01 -1.77015498e-01
2.89590091e-01 -2.77960241e-01 8.67935002e-01 -9.60425958e-02
-3.49781036e-01 5.32184422e-01 -5.47464550e-01 -5.41213930e-01
-3.93092543e-01 -2.63115540e-02 3.21216285e-01 -1.96956024e-01
-1.55316293e-01 -6.18668020e-01 -5.71169615e-01 7.93320388e-02
-8.37101281e-01 2.54710075e-02 6.29354239e-01 6.52447343e-01
6.07847929e-01 9.94722098e-02 2.82639325e-01 -5.37805557e-01
5.12186944e-01 -4.05257523e-01 -1.07620454e+00 1.67856459e-03
-5.73700845e-01 3.51612866e-01 8.11281562e-01 -5.43656290e-01
-8.32421243e-01 2.57883891e-02 -6.67445511e-02 -4.98141386e-02
-3.41932118e-01 2.81922996e-01 -1.67280972e-01 -1.45318434e-01
4.52294737e-01 3.39123368e-01 8.82648081e-02 -4.17392880e-01
2.59858221e-01 3.41664076e-01 2.10322797e-01 -7.96726525e-01
6.24299943e-01 4.05454785e-01 5.66814184e-01 -9.34684277e-01
-3.55519652e-01 -1.01155587e-01 -5.14290750e-01 -4.81232733e-01
8.44816387e-01 -4.53599095e-01 -1.43842530e+00 3.98799717e-01
-8.34800661e-01 -9.93890524e-01 -7.23936319e-01 7.04555988e-01
-1.23466861e+00 5.37096374e-02 -1.82699546e-01 -1.28040159e+00
-9.57851112e-02 -1.10210562e+00 7.45257795e-01 5.42226195e-01
2.48749778e-02 -7.52786815e-01 2.50703454e-01 -1.11505106e-01
4.47916031e-01 6.10315382e-01 5.35066783e-01 -2.57010698e-01
-5.01975358e-01 1.64027944e-01 5.82844675e-01 1.02164783e-01
3.09670083e-02 -4.50690180e-01 -9.28083122e-01 -4.40977544e-01
3.92805971e-02 -6.29717648e-01 2.36975446e-01 2.59009093e-01
9.97623205e-01 -2.82648474e-01 7.60620758e-02 1.27572715e-01
1.55511677e+00 6.30682588e-01 4.56553429e-01 6.86953604e-01
2.31713559e-02 7.44142771e-01 1.04969609e+00 8.07858407e-01
2.11571574e-01 7.86715686e-01 9.83864129e-01 1.31257579e-01
4.21603352e-01 -7.84169659e-02 3.73370588e-01 3.38087946e-01
-4.18187499e-01 -3.42240155e-01 -4.40768361e-01 2.77185082e-01
-1.90769994e+00 -8.14816117e-01 3.85652304e-01 2.47096825e+00
4.78905022e-01 -4.88112308e-03 1.57948844e-02 3.33174288e-01
1.52460948e-01 -4.54722978e-02 -9.34433579e-01 -6.07382059e-01
1.21449888e-01 3.05860251e-01 6.26934111e-01 7.00849235e-01
-6.54769719e-01 7.08250403e-01 5.27284288e+00 3.98547560e-01
-1.20022726e+00 3.52105498e-02 2.16052622e-01 -1.89569041e-01
1.26707956e-01 8.95842444e-03 -2.23789677e-01 2.79622525e-01
1.08798838e+00 -1.36147037e-01 9.12243962e-01 8.75480950e-01
5.31207323e-01 -7.91956007e-01 -1.10231030e+00 6.68612957e-01
-4.19350058e-01 -7.43672192e-01 -5.24819493e-01 2.61513770e-01
3.90395284e-01 -4.16540802e-02 -8.62588435e-02 4.81218725e-01
5.33745706e-01 -8.87980044e-01 7.84446657e-01 6.42539144e-01
1.91021506e-02 -9.54389751e-01 8.22902143e-01 8.55739236e-01
-8.97191286e-01 -3.42749178e-01 -2.66825527e-01 -5.24630308e-01
1.09682843e-01 4.66747843e-02 -8.33051324e-01 7.05125868e-01
4.54043359e-01 1.83521956e-01 -2.48019286e-02 1.03495967e+00
-3.43450129e-01 3.19046527e-01 -5.55036306e-01 -7.84159243e-01
6.13544881e-01 -7.58039117e-01 5.84498525e-01 5.06957471e-01
4.77724999e-01 2.93135196e-01 3.02039504e-01 7.79252708e-01
4.95135427e-01 6.62445575e-02 -7.04762936e-01 9.81493220e-02
1.41837105e-01 1.02160382e+00 -6.96467757e-01 -2.76995629e-01
1.24493159e-01 9.02773440e-01 2.24483639e-01 2.88562626e-01
-6.67824566e-01 -4.59601693e-02 7.44796574e-01 -1.29258931e-01
2.29581505e-01 -3.97810340e-01 2.67546386e-01 -8.68034065e-01
-1.01604812e-01 -6.67680740e-01 4.13976423e-02 -7.04376221e-01
-4.56869662e-01 2.69085407e-01 2.02869132e-01 -1.15765584e+00
-7.16403604e-01 -7.52132654e-01 -1.54974714e-01 5.32441854e-01
-1.56812716e+00 -5.18205047e-01 -1.95754245e-01 3.91277701e-01
5.23426235e-01 2.04861209e-01 8.24522257e-01 -1.01473764e-01
-5.19456029e-01 -2.59654105e-01 3.12214345e-01 -5.15921354e-01
1.57569051e-01 -1.60570526e+00 -2.08187178e-01 4.50828999e-01
-3.02054852e-01 2.66576484e-02 1.34626651e+00 -3.55575800e-01
-1.80363262e+00 -4.66864079e-01 1.03955708e-01 2.05557704e-01
5.95104992e-01 -3.60919416e-01 -7.49438345e-01 4.05593097e-01
3.44542593e-01 -1.00170851e-01 2.95835920e-02 -2.42445827e-01
5.71557999e-01 -1.39292464e-01 -1.35888672e+00 5.73048532e-01
6.38739645e-01 -2.52009660e-01 -3.27227592e-01 1.99442521e-01
2.66689986e-01 -6.59372270e-01 -8.24225187e-01 1.93421856e-01
6.91895008e-01 -1.05818677e+00 6.90881968e-01 -5.56523025e-01
-1.27847925e-01 -3.22026908e-01 3.98911759e-02 -1.63850999e+00
1.78407609e-01 -8.06909382e-01 -1.37701809e-01 5.93480945e-01
5.73419221e-02 -6.99682772e-01 9.48194325e-01 3.96949172e-01
3.67134288e-02 -7.47469425e-01 -1.46943140e+00 -8.20610821e-01
2.03157693e-01 -3.28043967e-01 4.56164122e-01 6.31499887e-01
5.86189367e-02 -3.75351273e-02 -4.74474788e-01 4.12974358e-01
5.75595021e-01 1.36811167e-01 8.61879706e-01 -1.18001521e+00
-7.46661663e-01 -1.79020822e-01 -2.72440732e-01 -6.94953918e-01
1.29493907e-01 -7.55771473e-02 5.42183161e-01 -1.58567226e+00
-3.97836447e-01 -3.75776649e-01 -1.43517002e-01 2.32744291e-01
1.96489379e-01 -3.61465156e-01 2.82049865e-01 -3.25105667e-01
-5.01239002e-01 7.02744722e-01 1.39753568e+00 1.79644644e-01
-1.49329454e-01 1.24156430e-01 1.60994276e-01 7.23228276e-01
9.92883623e-01 -3.73329848e-01 -6.59494877e-01 -1.06642872e-01
4.86398399e-01 3.25050622e-01 2.02144772e-01 -1.35828340e+00
4.12248075e-02 -4.83779222e-01 -5.99955469e-02 -7.32214227e-02
7.06846595e-01 -1.05487585e+00 2.08329618e-01 1.21182668e+00
-1.83723032e-01 1.58460662e-01 4.15461771e-02 6.06146216e-01
1.19567893e-01 -5.79319358e-01 8.21109176e-01 -6.14955425e-01
-7.56227970e-01 -6.56506643e-02 -8.05180609e-01 -2.07211271e-01
1.20528555e+00 -8.95409882e-02 1.83972542e-03 -3.70401531e-01
-9.38144684e-01 4.01574016e-01 4.60755527e-01 3.25670779e-01
4.07026321e-01 -7.82347202e-01 -4.06557381e-01 -2.36126691e-01
-2.84019113e-01 1.71054617e-01 3.17450017e-01 8.16777885e-01
-6.58983767e-01 1.60189480e-01 -4.22206432e-01 -2.37056181e-01
-1.06266332e+00 6.50796652e-01 6.46222532e-01 -4.70140845e-01
-5.06325185e-01 1.86477840e-01 -1.13331385e-01 -4.36417073e-01
7.10367560e-02 -6.52871549e-01 -3.69436532e-01 -6.84340075e-02
2.80316919e-01 5.24546981e-01 -1.15382098e-01 -7.00660646e-02
2.38837507e-02 5.31585217e-01 4.50909913e-01 -5.17845333e-01
1.34165967e+00 -1.24934100e-01 2.57795364e-01 8.98753941e-01
4.32792902e-01 -4.91799176e-01 -1.55926359e+00 1.05356552e-01
1.47007421e-01 -1.32241189e-01 1.03898242e-01 -7.10499406e-01
-5.38940907e-01 7.01718688e-01 9.52737033e-01 2.87902683e-01
1.15920687e+00 -4.11218882e-01 2.37315863e-01 8.94906819e-01
9.68658149e-01 -1.57173765e+00 5.58963045e-02 5.30121863e-01
8.61816764e-01 -1.06416821e+00 4.91187535e-02 1.81880817e-01
-6.65165424e-01 1.10035431e+00 6.42411947e-01 -2.94039309e-01
2.76055098e-01 2.72122413e-01 -1.96884409e-01 8.83435234e-02
-8.73537302e-01 -5.65724611e-01 -4.06172872e-01 6.33366823e-01
-5.46357930e-02 2.74860352e-01 -4.65132535e-01 -1.61113694e-01
-2.38201335e-01 1.87090918e-01 8.34698498e-01 1.23390925e+00
-6.83389127e-01 -1.14726663e+00 -4.83932793e-01 -1.19682044e-01
-1.93195999e-01 3.89535755e-01 -1.54544443e-01 1.25924361e+00
2.24716753e-01 7.72600949e-01 -8.23183432e-02 1.13155678e-01
4.79342759e-01 -1.14841633e-01 6.92576885e-01 -3.97135556e-01
-5.06901443e-01 -1.01089247e-01 2.12844923e-01 -9.01491702e-01
-5.64116836e-01 -8.31234336e-01 -1.45462561e+00 6.74508214e-02
4.18559089e-02 3.94787371e-01 1.06440067e+00 1.21126652e+00
-4.98019941e-02 6.24022961e-01 6.10191524e-01 -1.04515028e+00
-6.42806172e-01 -1.14209509e+00 -7.87465513e-01 1.21056773e-02
4.79173541e-01 -9.42293525e-01 -2.20795214e-01 -5.47656715e-01]
|
[4.381244659423828, 1.9527369737625122]
|
45f2037c-9f80-4f32-815a-9e5da75c3d73
|
groupformer-group-activity-recognition-with
|
2108.12630
| null |
https://arxiv.org/abs/2108.12630v1
|
https://arxiv.org/pdf/2108.12630v1.pdf
|
GroupFormer: Group Activity Recognition with Clustered Spatial-Temporal Transformer
|
Group activity recognition is a crucial yet challenging problem, whose core lies in fully exploring spatial-temporal interactions among individuals and generating reasonable group representations. However, previous methods either model spatial and temporal information separately, or directly aggregate individual features to form group features. To address these issues, we propose a novel group activity recognition network termed GroupFormer. It captures spatial-temporal contextual information jointly to augment the individual and group representations effectively with a clustered spatial-temporal transformer. Specifically, our GroupFormer has three appealing advantages: (1) A tailor-modified Transformer, Clustered Spatial-Temporal Transformer, is proposed to enhance the individual representation and group representation. (2) It models the spatial and temporal dependencies integrally and utilizes decoders to build the bridge between the spatial and temporal information. (3) A clustered attention mechanism is utilized to dynamically divide individuals into multiple clusters for better learning activity-aware semantic representations. Moreover, experimental results show that the proposed framework outperforms state-of-the-art methods on the Volleyball dataset and Collective Activity dataset. Code is available at https://github.com/xueyee/GroupFormer.
|
['Shuai Yi', 'Jun Hou', 'Shinan Liu', 'Kunlin Yang', 'Lingbo Liu', 'Qianggang Cao', 'Shuaicheng Li']
|
2021-08-28
| null |
http://openaccess.thecvf.com//content/ICCV2021/html/Li_GroupFormer_Group_Activity_Recognition_With_Clustered_Spatial-Temporal_Transformer_ICCV_2021_paper.html
|
http://openaccess.thecvf.com//content/ICCV2021/papers/Li_GroupFormer_Group_Activity_Recognition_With_Clustered_Spatial-Temporal_Transformer_ICCV_2021_paper.pdf
|
iccv-2021-1
|
['group-activity-recognition']
|
['computer-vision']
|
[ 0.04636412 -0.30907422 -0.4435891 -0.53892165 -0.5072464 -0.2787737
0.58423996 0.16616358 -0.26108158 0.4272255 0.56206346 0.18082123
-0.4800772 -0.7686857 -0.5796931 -0.7321687 -0.1316728 0.07193415
0.21695668 0.11148048 0.31111914 0.04087991 -1.692633 0.4168192
1.3448132 1.0981296 0.3397388 0.24947894 -0.08852537 1.0627483
-0.41229331 0.14683644 -0.25633937 -0.86237377 -0.53952986 0.3977091
0.22664359 -0.0819503 -0.3928772 0.682393 0.46492392 0.6331319
0.38580072 -1.254658 -1.0907654 0.6010217 -0.5387361 0.39088368
0.4330641 0.20627573 0.8859384 -0.8390057 0.07357834 1.0526286
0.2684276 0.44592163 -0.8869592 -1.0100855 0.95038646 0.6316952
-1.5997831 -0.27152848 0.7484557 -0.35386536 0.95631385 0.14252295
1.1052281 1.0130334 -0.19920774 1.2012368 0.8224653 -0.06529454
0.19898286 -0.33756492 0.2887652 0.8240599 -0.02200472 -0.3027221
-0.9844858 0.15593407 0.87891924 0.5261956 -0.3250005 -0.18585576
-1.4119964 0.6070622 0.770503 0.61804384 -0.4993624 0.45377466
0.11439677 -0.07862011 0.41010118 0.08063173 0.0444541 -0.24977773
-0.84346366 0.21710199 0.11108825 0.8085651 0.8172691 -0.19252448
-0.57703835 0.98887277 0.5420308 0.21788114 0.8023858 -0.86253446
0.8067622 1.0119277 -0.14206657 -0.94172335 -0.249548 -0.64000386
-0.78655005 -0.47300464 0.25747693 0.13048345 -0.7483042 1.8841221
0.225951 0.75363797 -0.26574227 0.90568537 0.8875823 0.6794978
0.46047574 -0.1826436 1.3367088 -1.3281524 -0.87325454 -0.58805996
0.6275276 -0.16610642 0.966374 0.1340919 -1.2293292 -0.83415353
-0.9940422 -0.17224291 -0.3227393 0.27158666 0.88888097 0.36787343
-0.86857843 0.41277936 -1.0206782 -0.24944134 0.81593096 0.51763934
-0.27064258 -0.14474836 -1.104036 0.2357532 0.28384674 0.25096437
-0.6761068 -0.64713466 -0.87342715 0.24955188 0.45465595 -0.74584204
1.0121976 -1.0303975 -1.3543324 0.5490458 -0.5821016 -0.14670481
0.18668155 -0.20850888 -0.3153582 0.09710182 0.18856789 0.5205717
0.36727902 -0.8073237 -0.7323775 -0.67345035 -0.20883301 0.70950174
-0.6726277 -0.08939276 -0.8162963 -0.8881253 0.3770871 -0.6917377
-0.14100493 -0.09020538 -0.1038466 -0.64101714 0.5444838 -0.5829641
1.6164205 -2.078928 0.41148403 0.14623274 0.21730568 0.16027223
-0.05191366 0.49163687 -0.03803942 0.03170038 -0.07930572 -0.59746313
-0.04274039 0.1469744 0.02326019 0.45448786 -0.03188624 1.1741015
-0.9638604 -0.42831174 0.2648777 0.57475173 -0.6143412 0.17764999
0.06898869 0.6171805 -0.8641328 0.66432935 0.48918292 -0.44718117
0.16736355 -0.09276612 -0.21352953 0.42809814 -1.0907024 1.9463292
-0.27573603 0.4147466 -0.29610878 -1.232141 0.8902798 0.14021467
0.7004394 -1.0854304 0.04871819 -0.01352261 -0.16496685 -0.35842875
0.16816524 0.29594874 -0.210519 0.6908339 0.1814733 0.828879
0.17038366 0.24350527 1.0004998 0.19506702 0.25014994 -0.16327006
0.5735061 -0.5883789 0.93452203 0.49459103 -0.3730945 0.48781896
0.27844456 -0.31321803 -0.24875358 -1.1037345 0.40849882 1.4767321
0.53827727 -0.5820804 -0.60233516 -0.6779466 -0.19917075 0.38765532
-0.93048364 -0.3721471 -0.6967178 -0.7081984 0.25065404 1.0938673
1.0734854 -1.0214198 -0.2539639 0.16539563 -0.5703326 -0.6122487
-0.9622371 -0.16867335 -0.8204199 -1.0307562 -0.611873 -0.9630479
0.77817404 0.67740476 0.7847324 0.33588114 -0.0452842 0.32917562
-0.47537896 -0.20279044 0.56787425 0.06236046 -0.15252814 0.47343978
0.59111905 -0.9080294 -0.976492 0.62694055 -0.6409564 0.272885
0.67307836 0.62339884 0.67808205 -0.02307358 0.7207482 -0.5059875
0.5676496 -0.5660071 -0.06248235 0.3750295 -0.30336812 -0.17931198
0.3350094 -0.62166387 -1.252463 -0.13123819 0.08886056 -0.29805985
-0.03014009 0.5090451 -0.4997358 0.23470007 0.32627922 0.65115565
-0.24626675 -0.58671707 0.08299661 0.5415251 0.3992535 -0.6167413
0.39269873 0.441686 -0.40579396 -0.58995754 -0.9401656 -0.61443144
-0.7174757 -0.39247903 0.9974547 -1.2237812 -0.8997898 0.69169736
-0.8063503 -0.6342208 0.0120675 0.5590969 -0.41246337 0.2914972
-0.55573666 -0.73407054 -0.0687962 -0.81626666 0.975737 0.70553994
-0.11848665 -1.0823523 -0.14755715 0.86456376 0.16248539 0.08621874
0.41729197 -0.47939318 -0.8590863 0.08954701 -0.19995841 0.04675748
0.32457155 -0.3378756 -0.63407075 -0.1721134 -0.37845317 -0.17628735
1.0230106 0.34745258 1.6852946 -0.28833574 -0.5034217 0.5365319
0.89794856 0.3575274 0.6141089 -0.01182732 0.95435625 0.55346656
0.6234816 0.48444292 0.87959796 0.8758754 0.03097652 0.06304445
-0.13422856 -0.63771755 0.42831987 0.8788935 -0.35659373 -0.2803396
-0.7637262 0.7363461 -2.4437976 -1.2557331 -0.02654493 2.007163
0.5256128 -0.25974548 0.30010024 0.1730865 0.712858 0.3571065
-0.560129 0.2349156 0.0518375 -0.0773105 0.13371539 0.22985534
-1.0553516 0.811307 4.97941 0.94107634 -0.7105531 0.23286757
0.4995884 -0.46998242 -0.20240201 -0.26826453 -0.71816576 0.84504426
0.5374077 -0.34864935 0.45429426 0.5549153 0.3491511 -0.00652533
-1.0095897 1.0922314 0.28014717 -1.1239284 -0.00800384 0.29658252
0.70007914 -0.34574765 0.13184866 0.4983407 0.28970152 -1.0595185
0.6973844 0.7407076 0.20672311 -0.8149629 0.4326192 0.42079207
-1.7962495 -0.34513485 -0.00621223 -0.19802386 0.06628628 0.2065829
-0.11593286 0.4252134 0.72174346 1.3393122 -0.8125725 0.9883975
-0.3637915 0.5724256 -0.18249305 -0.04015843 0.2186179 -0.28213552
0.07617265 0.8751857 0.5543193 0.39483032 0.34308934 0.7850823
-0.05229284 0.05742291 -0.13094819 -0.19636299 0.79671353 0.9679768
-0.65091187 -0.3056891 -0.418197 0.94449335 0.57293355 0.47199
-0.99932456 -0.22120413 0.62982756 0.19108655 0.39263818 -0.28347424
-0.16431177 -1.2482836 0.18504502 -0.49795863 0.84320396 -0.7921096
-1.0700805 0.19425708 -0.02776515 -1.2109709 0.01285255 -0.17045613
-0.9448462 0.68770033 -1.1282188 -1.4293967 -0.73759645 0.81547046
0.7277634 -0.11064524 0.53374434 0.4622505 -0.9669298 0.648773
-0.32393828 0.21141542 0.49132603 -1.0670152 0.07471318 0.88493466
0.1507176 0.8568424 0.13788047 -0.67384696 -1.1303822 -1.1908953
1.0082816 -0.31643733 0.4064553 -0.45636576 -1.0859506 0.92036694
0.04842828 0.07549636 1.1887362 0.28181255 -0.2045711 -0.2043631
-0.65953124 0.5805687 1.6731911 -0.50289434 -0.5322774 0.16584514
0.46606755 -0.03686072 -0.8236593 0.13229454 0.4606365 -1.0453916
0.91257685 -0.26550892 0.19326396 -0.5315158 0.01124448 -1.356952
-0.67958766 -0.38514355 -0.41829413 1.3203205 -0.01323715 -0.60543466
1.0157175 0.46160942 -0.31570008 -0.8910235 -0.81857973 -0.5985639
-0.26166654 -0.3734452 0.75940794 0.920246 0.20728324 0.22899391
-0.37709194 0.04998976 0.57196254 0.23886384 0.45488143 -1.0707753
-0.24228603 -0.56738913 -0.3086596 -1.4506943 0.05691726 -0.88218796
-0.16802253 -1.8495454 0.3078073 -0.38681605 -0.7112057 0.75239927
-0.43732148 0.16221927 0.11700069 0.24003945 -1.275766 0.9333033
1.2735113 0.01215832 -0.39610112 0.01454052 -1.0566545 0.4970155
0.882716 -0.12827146 -0.80961907 -0.4403503 -0.09884058 -0.12872864
0.4341519 -1.2774774 0.594485 -0.4344172 0.60178626 -0.58758044
0.44809797 -0.41045722 0.0484176 0.3179697 -0.51318395 -0.16988459
-0.2171204 0.74884456 -0.413002 0.38667974 0.51468927 -0.05622889
-0.9333879 0.70683116 -0.39665386 -0.10106709 1.2793067 -0.32643154
-0.38179293 -0.35090008 -0.79908127 0.7105877 0.124842 0.77921677
0.63161457 -1.8309264 -0.4033874 0.37688124 0.33900207 0.03055679
1.011714 0.9996471 -0.12724976 0.3651044 -0.2106151 -0.56759566
-1.3392354 0.39861324 0.45327035 -0.18689188 -0.27431026 1.0268059
0.50700706 -0.16399665 0.32605308 -0.20398587 -0.45498255 0.15410203
0.9215601 0.45521072 -0.3046625 -0.73609227 -0.55195826 0.6162781
0.03775616 0.17141342 1.3531815 -0.20983537 0.09345316 0.3742749
1.0017791 -0.36573166 -1.5129278 -0.44431305 -0.19287406 -0.696402
0.12866807 -0.6246931 -1.2505351 1.0471509 0.4283614 -0.02858241
1.3506798 0.10944639 0.6891686 0.07032686 0.27060246 -1.1655097
0.66805995 0.16382223 0.76571524 -0.88256687 -0.04848188 -0.5474371
-0.7650896 0.54995906 1.027644 -0.11401358 0.5236175 -0.2405559
-0.330698 -0.19636348 -0.9970122 -0.29444543 0.57947594 0.6038004
0.58494055 0.03886082 -0.32441708 1.2608726 0.19595662 0.15823053
-0.06165867 0.99037063 -0.5232291 -1.1266546 -0.16045219 0.37396786
-0.15816139 0.27915666 -0.35535657 0.4921692 0.63354516 1.0322541
0.32635286 -0.52940536 0.35962716 -0.14277185 0.39530328 -0.62510484
-0.6341492 0.15473089 -0.08319321 -0.89950156 -0.5145999 -0.9312832
-1.2860781 -0.22668822 0.26963326 0.22172149 0.22901915 0.9452946
0.49938816 0.81853914 0.45439786 -0.8672207 0.15351102 -0.96628976
-0.61063427 0.485664 0.03587423 -0.92251605 -0.12170894 0.02519023]
|
[8.206324577331543, 0.7047435641288757]
|
1f3bc620-31c9-4dbc-a1d3-0caf1452e0cf
|
pay-better-attention-to-attention-head
|
2106.10840
| null |
https://arxiv.org/abs/2106.10840v1
|
https://arxiv.org/pdf/2106.10840v1.pdf
|
Pay Better Attention to Attention: Head Selection in Multilingual and Multi-Domain Sequence Modeling
|
Multi-head attention has each of the attention heads collect salient information from different parts of an input sequence, making it a powerful mechanism for sequence modeling. Multilingual and multi-domain learning are common scenarios for sequence modeling, where the key challenge is to maximize positive transfer and mitigate negative transfer across languages and domains. In this paper, we find that non-selective attention sharing is sub-optimal for achieving good generalization across all languages and domains. We further propose attention sharing strategies to facilitate parameter sharing and specialization in multilingual and multi-domain sequence modeling. Our approach automatically learns shared and specialized attention heads for different languages and domains to mitigate their interference. Evaluated in various tasks including speech recognition, text-to-text and speech-to-text translation, the proposed attention sharing strategies consistently bring gains to sequence models built upon multi-head attention. For speech-to-text translation, our approach yields an average of $+2.0$ BLEU over $13$ language directions in multilingual setting and $+2.0$ BLEU over $3$ domains in multi-domain setting.
|
['Xian Li', 'Juan Pino', 'Yun Tang', 'Hongyu Gong']
|
2021-06-21
| null |
http://proceedings.neurips.cc/paper/2021/hash/15c00b5250ddedaabc203b67f8b034fd-Abstract.html
|
http://proceedings.neurips.cc/paper/2021/file/15c00b5250ddedaabc203b67f8b034fd-Paper.pdf
|
neurips-2021-12
|
['speech-to-text-translation']
|
['natural-language-processing']
|
[ 6.42451197e-02 -7.76285455e-02 -3.96438628e-01 -4.33934957e-01
-1.36313188e+00 -8.87277663e-01 3.22252214e-01 -3.12899262e-01
-6.94650888e-01 9.32066083e-01 3.54918957e-01 -5.29472709e-01
5.06843150e-01 -2.05560938e-01 -8.88280690e-01 -3.80314797e-01
1.82716161e-01 6.90605581e-01 -9.65301469e-02 -6.34808302e-01
-8.28790292e-03 1.50591552e-01 -8.17014396e-01 4.95856345e-01
1.08761060e+00 5.03998518e-01 8.53851438e-01 7.33968437e-01
-4.63591188e-01 7.09136844e-01 -5.96176982e-01 -7.10845053e-01
1.77884728e-01 -3.84010732e-01 -9.97362971e-01 -1.63529471e-01
3.11488509e-01 -5.54924458e-02 -2.93333232e-01 9.18628752e-01
1.00673413e+00 9.97449979e-02 4.26626623e-01 -1.06584847e+00
-9.66656864e-01 9.89209890e-01 -8.30229521e-01 4.72899795e-01
2.49047101e-01 4.06234056e-01 1.23537850e+00 -9.41909313e-01
4.50926781e-01 1.52838039e+00 3.79723549e-01 6.46071970e-01
-1.06647503e+00 -8.36914897e-01 5.15450001e-01 1.80832103e-01
-1.16452873e+00 -5.54381967e-01 4.06473637e-01 -3.57351035e-01
1.66653633e+00 1.59687862e-01 6.99727014e-02 1.21042562e+00
2.23943755e-01 1.15229440e+00 7.40607381e-01 -4.01341319e-01
-2.61480123e-01 1.62525445e-01 2.92001572e-02 3.62370849e-01
-1.77028418e-01 -1.94949135e-01 -7.52309620e-01 -8.03635940e-02
5.18886685e-01 -3.33841622e-01 -1.62782550e-01 2.67291814e-01
-1.38447583e+00 7.70756185e-01 1.07795671e-01 1.48614869e-01
-4.53142792e-01 -2.49373894e-02 7.85382867e-01 5.41723013e-01
6.26596034e-01 5.09382367e-01 -9.06009555e-01 -3.51415962e-01
-5.79573631e-01 7.40754604e-02 5.46191812e-01 1.52304244e+00
7.51754940e-01 3.75116378e-01 -3.30184400e-01 1.29709554e+00
5.69288358e-02 9.00104463e-01 6.84928417e-01 -4.82036829e-01
9.72882628e-01 3.43878061e-01 -1.99313313e-02 -3.36980045e-01
-2.76128113e-01 -4.97302771e-01 -7.45606720e-01 -4.84612256e-01
1.97579619e-02 -5.65999627e-01 -8.72223318e-01 2.31661320e+00
-7.70206973e-02 -7.74154440e-02 1.29609331e-01 8.11781883e-01
5.77486873e-01 9.01359737e-01 4.09831077e-01 -2.06744149e-01
1.41970575e+00 -1.24414504e+00 -6.21420979e-01 -8.33588064e-01
9.10702109e-01 -1.00840497e+00 1.38945246e+00 -1.15603730e-01
-1.36671793e+00 -5.23190618e-01 -6.71950340e-01 -3.11817855e-01
2.19018618e-03 4.83250841e-02 2.29312345e-01 4.16490406e-01
-1.07233763e+00 4.55869967e-03 -5.64398766e-01 -3.41255575e-01
1.40536994e-01 5.97650051e-01 -1.52782068e-01 -1.34396344e-01
-1.51149142e+00 1.14440465e+00 2.25820005e-01 -3.36041629e-01
-1.07440031e+00 -8.50984156e-01 -7.39289045e-01 2.05447838e-01
1.49566576e-01 -5.99102557e-01 1.41374421e+00 -1.14606977e+00
-1.33518016e+00 9.39199030e-01 -4.62439060e-01 -5.09771764e-01
2.31424659e-01 -4.45146501e-01 -3.62838060e-01 -3.97119135e-01
2.33977243e-01 8.54495287e-01 5.20028830e-01 -7.17645824e-01
-6.86957121e-01 -3.35957170e-01 -1.53931096e-01 6.02065206e-01
-2.46255070e-01 5.04433155e-01 -5.70334971e-01 -8.94606590e-01
-5.08080184e-01 -9.98731196e-01 -7.07419142e-02 -6.67697549e-01
-2.30344415e-01 -2.61092663e-01 6.62913501e-01 -1.15838277e+00
1.12775505e+00 -1.86279869e+00 5.25395393e-01 -4.58098084e-01
-1.79289669e-01 4.27165180e-01 -5.85441947e-01 5.01924336e-01
4.90783416e-02 1.48116961e-01 -2.93782234e-01 -3.78049076e-01
-6.13404950e-03 1.41334489e-01 -1.89923018e-01 1.73100829e-01
4.91386890e-01 1.19990218e+00 -7.52505481e-01 -3.10828269e-01
-1.18589640e-01 3.26715559e-01 -6.15270197e-01 4.03313875e-01
-3.65205586e-01 4.18540686e-01 -1.83005497e-01 6.24661148e-01
5.50653279e-01 -3.49197775e-01 5.14119685e-01 1.43174693e-01
1.63911283e-01 4.64061350e-01 -6.01437449e-01 1.94376385e+00
-8.81917059e-01 7.68628120e-01 2.82868147e-01 -9.17078078e-01
8.62459362e-01 5.40538251e-01 1.51619479e-01 -1.08162391e+00
9.77943242e-02 2.94643939e-01 3.94297749e-01 -2.80056030e-01
6.83292210e-01 -3.04151416e-01 -3.19618493e-01 7.14739621e-01
4.64511633e-01 2.06562936e-01 -2.86490731e-02 2.61582851e-01
8.34881485e-01 -1.94094643e-01 3.08883876e-01 -5.42223990e-01
5.91756582e-01 -1.28004506e-01 7.68837333e-01 5.87359548e-01
-2.98720717e-01 2.47277409e-01 3.23831856e-01 -1.16444610e-01
-1.48026311e+00 -6.16836190e-01 5.20542324e-01 2.16425347e+00
-7.00133741e-02 1.34750634e-01 -7.93265343e-01 -6.73780560e-01
-1.56646639e-01 5.74875176e-01 -8.43859836e-02 -2.08861068e-01
-8.80300283e-01 -7.95880675e-01 8.42914641e-01 7.14584947e-01
2.05332831e-01 -1.19543028e+00 -5.87927969e-03 4.11847740e-01
-6.92399561e-01 -1.11264336e+00 -1.28961980e+00 2.33919099e-01
-5.97025692e-01 -5.17508209e-01 -1.29935646e+00 -1.19933379e+00
2.01896667e-01 2.80178547e-01 1.44617248e+00 -2.88921446e-01
8.09775963e-02 4.73943911e-03 -2.02299252e-01 -4.49209005e-01
-4.71388429e-01 6.47090077e-01 1.96314797e-01 -1.89907208e-01
4.43581104e-01 -3.53578091e-01 -3.28598976e-01 3.48698497e-01
-5.53966105e-01 -2.03025360e-02 7.15270698e-01 1.08107579e+00
1.64794266e-01 -9.42984283e-01 1.10068035e+00 -7.18402743e-01
9.54724193e-01 -7.50474930e-01 -3.66116852e-01 6.08373702e-01
-3.04196060e-01 3.34717065e-01 7.90378392e-01 -5.51236868e-01
-1.09360254e+00 -8.64504650e-02 -2.79043585e-01 -2.88150042e-01
1.34620845e-01 4.51578230e-01 -4.57518160e-01 1.48892045e-01
6.91992402e-01 6.73331857e-01 -1.43761873e-01 -4.99239117e-01
4.41684693e-01 9.74595845e-01 2.19380960e-01 -7.52450883e-01
1.38711661e-01 -8.36585760e-02 -8.17045748e-01 -8.39638948e-01
-3.40713531e-01 -5.08802354e-01 -4.55357879e-01 1.79977447e-01
8.01176488e-01 -1.33162522e+00 -5.92308581e-01 6.14617467e-01
-1.40501201e+00 -6.25601292e-01 3.19452107e-01 3.69008183e-01
-6.24031544e-01 3.80436152e-01 -9.29876626e-01 -7.22562075e-01
-7.94411063e-01 -1.51207852e+00 1.06451571e+00 -1.20054185e-01
-2.60688126e-01 -9.75033045e-01 -3.52107137e-02 4.24799919e-01
5.75501740e-01 -6.72549844e-01 1.18633425e+00 -8.45496356e-01
-4.49509412e-01 2.93988764e-01 -2.57777959e-01 4.26240325e-01
6.22318983e-02 -5.77187061e-01 -8.70548964e-01 -5.61776400e-01
-3.56053263e-01 -5.10397315e-01 5.76757431e-01 4.00595188e-01
7.49877930e-01 -3.98426652e-01 -2.35208035e-01 5.47324121e-01
1.09082901e+00 5.67552686e-01 3.34971696e-01 1.29728198e-01
8.41215551e-01 6.11057103e-01 4.51036930e-01 3.60156804e-01
6.07032239e-01 7.28875101e-01 -1.93119217e-02 -1.57437190e-01
-7.24611282e-02 -1.44074038e-01 6.51508510e-01 1.03962064e+00
4.26724821e-01 -5.75501442e-01 -1.12747049e+00 9.66364086e-01
-1.68504298e+00 -8.02077293e-01 2.86613941e-01 2.05738306e+00
1.14671934e+00 -2.18542978e-01 4.41183954e-01 -5.89375734e-01
9.97084320e-01 1.08477585e-01 -7.89060891e-01 -6.95094228e-01
-4.02669072e-01 1.25540137e-01 6.35896206e-01 7.23497391e-01
-6.94190979e-01 1.44779718e+00 6.51358509e+00 1.03747737e+00
-1.48891497e+00 3.74885619e-01 8.01432550e-01 -2.10969791e-01
-4.98571038e-01 -3.45250040e-01 -9.93938863e-01 6.61628664e-01
1.24838877e+00 -5.93258977e-01 4.85866934e-01 7.38647163e-01
7.37292692e-02 4.57358569e-01 -1.01763880e+00 8.61321509e-01
-7.71673545e-02 -1.16880238e+00 2.46746838e-01 -3.16744037e-02
8.74180496e-01 6.39167368e-01 1.06292985e-01 5.81342876e-01
6.32464170e-01 -1.06268251e+00 6.54137433e-01 -1.36787549e-01
9.82019305e-01 -9.75521326e-01 5.48679352e-01 5.73702812e-01
-1.03853285e+00 2.03460827e-02 -2.86389172e-01 1.66742176e-01
5.11592448e-01 4.56127673e-02 -1.12574565e+00 4.07015681e-01
4.22162801e-01 5.50456285e-01 3.17320228e-02 3.88734698e-01
8.78263935e-02 6.36353016e-01 -1.80081818e-02 -1.39372066e-01
4.99444783e-01 5.99245727e-02 5.33906102e-01 1.82280946e+00
4.03898180e-01 -1.70937747e-01 2.41234988e-01 5.24560511e-01
-4.47396278e-01 3.99629086e-01 -5.46534896e-01 -3.32131863e-01
6.76908731e-01 6.75417364e-01 -1.29245743e-02 -5.29719174e-01
-6.11132264e-01 1.22008717e+00 6.09342813e-01 5.70287347e-01
-9.52000022e-01 -3.65099937e-01 1.19522285e+00 -1.97657749e-01
2.32137635e-01 -2.74406612e-01 -4.46267456e-01 -1.22619832e+00
-1.08271524e-01 -1.38352084e+00 4.47389573e-01 -5.75544298e-01
-1.30891895e+00 8.80514085e-01 -3.99002373e-01 -8.34959686e-01
-4.15911496e-01 -4.55815256e-01 -2.22603947e-01 1.60908043e+00
-1.50158775e+00 -1.22446322e+00 3.14290911e-01 6.90535784e-01
1.19117808e+00 -5.80155313e-01 5.46970427e-01 6.43332720e-01
-5.87291121e-01 1.17282438e+00 3.04540813e-01 2.22450629e-01
1.00078619e+00 -9.96246397e-01 1.00219750e+00 8.71637821e-01
-1.04061842e-01 6.56449854e-01 4.19025451e-01 -6.12149477e-01
-1.44169867e+00 -1.13628364e+00 1.14975393e+00 -2.86679566e-01
5.26348412e-01 -5.18064022e-01 -1.11633921e+00 1.04694557e+00
7.09232509e-01 -5.27576804e-01 7.54562736e-01 3.41831654e-01
-4.40904707e-01 1.81275576e-01 -8.53884339e-01 7.48952925e-01
1.14215875e+00 -8.37046623e-01 -3.73500735e-01 3.99407148e-01
1.16732192e+00 -4.08689409e-01 -7.00304449e-01 1.40077993e-01
3.10036391e-01 -4.13677037e-01 9.44574594e-01 -1.09376562e+00
4.57203776e-01 1.78244859e-01 -4.16904688e-01 -1.63715649e+00
-5.55958211e-01 -8.08464527e-01 1.64651215e-01 1.03561664e+00
8.61010551e-01 -6.03452504e-01 5.60321987e-01 2.35986173e-01
-4.78284806e-01 -3.79011214e-01 -7.74755120e-01 -8.35197330e-01
5.70276320e-01 -2.60789096e-01 6.26973033e-01 1.11222625e+00
1.18002072e-01 8.00386846e-01 -9.02722478e-01 6.18780255e-02
3.51075798e-01 -2.01149508e-01 6.11263990e-01 -5.14629245e-01
-4.58625585e-01 -5.26050806e-01 2.10462734e-02 -1.57690597e+00
5.64001262e-01 -1.13384628e+00 1.81014299e-01 -1.19814157e+00
4.06237870e-01 -2.77804792e-01 -2.98453629e-01 3.79274517e-01
-5.04703820e-01 -1.46202192e-01 3.95114511e-01 1.43555567e-01
-5.22675395e-01 5.34087539e-01 1.10901272e+00 -2.00824216e-01
-1.89575590e-02 -3.94554079e-01 -9.62433279e-01 1.31454796e-01
8.26163650e-01 -2.87483931e-01 -4.68199193e-01 -1.25522947e+00
-3.59670371e-02 3.37937534e-01 -2.67031252e-01 -4.13371861e-01
7.23894686e-02 -4.05642182e-01 -7.41268620e-02 -4.67248112e-01
2.90786952e-01 -3.14545363e-01 -2.61683255e-01 3.42061698e-01
-5.97172737e-01 3.18134904e-01 5.81468761e-01 4.04263616e-01
-3.12604964e-01 1.56886920e-01 9.20595169e-01 -3.82145852e-01
-8.67478371e-01 4.10803586e-01 -4.37422574e-01 5.85759759e-01
8.43043029e-01 1.90742910e-01 -3.99265379e-01 -5.80027461e-01
-6.72761977e-01 7.13606119e-01 2.39783272e-01 8.07027876e-01
3.06824207e-01 -1.23749614e+00 -1.23260391e+00 1.60559326e-01
1.40441984e-01 -3.04413706e-01 6.35531425e-01 6.44048691e-01
-2.89687961e-01 9.65140462e-01 -3.36793154e-01 -6.41316473e-01
-1.44171584e+00 3.89075488e-01 4.44796592e-01 -3.41484547e-01
1.10889986e-01 1.34109604e+00 7.68582106e-01 -8.78989279e-01
1.49328724e-01 -5.67051992e-02 3.41796905e-01 -1.53221399e-01
5.39471865e-01 3.57001215e-01 1.93759948e-01 -7.80856967e-01
-5.67825139e-01 3.21232259e-01 -5.71644008e-01 -3.75229567e-01
1.07095158e+00 -4.30239052e-01 1.27637267e-01 1.81785062e-01
1.34087002e+00 -2.61929959e-01 -1.26845360e+00 -6.27923071e-01
6.89328685e-02 -1.46281853e-01 -2.94856638e-01 -1.09702075e+00
-8.36128175e-01 1.25092816e+00 2.22065106e-01 -3.56004208e-01
9.79475677e-01 3.00277453e-02 1.04879141e+00 1.65868551e-01
3.63783360e-01 -1.08454633e+00 -4.62424830e-02 1.15362394e+00
9.98352587e-01 -1.50493431e+00 -6.90725386e-01 -8.42720494e-02
-1.27305996e+00 4.82374966e-01 8.35459411e-01 2.36780033e-01
1.19904540e-01 3.85777265e-01 2.82399148e-01 2.97430694e-01
-1.20892060e+00 -8.75326693e-02 2.26537794e-01 6.45016491e-01
1.03486466e+00 2.29382902e-01 -3.33373487e-01 5.49432933e-01
8.12308043e-02 -3.29775065e-01 1.64117411e-01 8.42231393e-01
-4.73104179e-01 -1.38880348e+00 -2.92326927e-01 1.39441714e-01
-7.70577371e-01 -6.60052896e-01 -4.65411425e-01 4.74667758e-01
-3.19067597e-01 8.25757384e-01 -3.27900276e-02 -3.30692559e-01
2.64866650e-01 5.15391052e-01 3.30100536e-01 -7.77163506e-01
-1.04442847e+00 2.41084129e-01 1.45375937e-01 -2.82203138e-01
1.47025824e-01 -5.60607195e-01 -1.11334693e+00 -4.32323277e-01
-4.77921516e-02 1.68470703e-02 4.22275841e-01 7.41654277e-01
8.78302276e-01 4.02190268e-01 5.56345224e-01 -4.28610027e-01
-7.75550961e-01 -1.26604152e+00 -2.90379852e-01 3.29133600e-01
4.34607297e-01 -2.78489411e-01 8.21449533e-02 7.46168345e-02]
|
[11.68820858001709, 10.087145805358887]
|
308ef62c-32de-4a1a-8324-49be489c4b7d
|
socially-compliant-navigation-dataset-scand-a
|
2203.15041
| null |
https://arxiv.org/abs/2203.15041v2
|
https://arxiv.org/pdf/2203.15041v2.pdf
|
Socially Compliant Navigation Dataset (SCAND): A Large-Scale Dataset of Demonstrations for Social Navigation
|
Social navigation is the capability of an autonomous agent, such as a robot, to navigate in a 'socially compliant' manner in the presence of other intelligent agents such as humans. With the emergence of autonomously navigating mobile robots in human populated environments (e.g., domestic service robots in homes and restaurants and food delivery robots on public sidewalks), incorporating socially compliant navigation behaviors on these robots becomes critical to ensuring safe and comfortable human robot coexistence. To address this challenge, imitation learning is a promising framework, since it is easier for humans to demonstrate the task of social navigation rather than to formulate reward functions that accurately capture the complex multi objective setting of social navigation. The use of imitation learning and inverse reinforcement learning to social navigation for mobile robots, however, is currently hindered by a lack of large scale datasets that capture socially compliant robot navigation demonstrations in the wild. To fill this gap, we introduce Socially CompliAnt Navigation Dataset (SCAND) a large scale, first person view dataset of socially compliant navigation demonstrations. Our dataset contains 8.7 hours, 138 trajectories, 25 miles of socially compliant, human teleoperated driving demonstrations that comprises multi modal data streams including 3D lidar, joystick commands, odometry, visual and inertial information, collected on two morphologically different mobile robots a Boston Dynamics Spot and a Clearpath Jackal by four different human demonstrators in both indoor and outdoor environments. We additionally perform preliminary analysis and validation through real world robot experiments and show that navigation policies learned by imitation learning on SCAND generate socially compliant behaviors
|
['Peter Stone', 'Joydeep Biswas', 'Justin Hart', 'Alexander Toshev', 'Soeren Pirk', 'Garrett Warnell', 'Xuesu Xiao', 'Anirudh Nair', 'Haresh Karnan']
|
2022-03-28
| null | null | null | null |
['social-navigation']
|
['robots']
|
[-3.10206175e-01 2.16285333e-01 2.42876694e-01 -4.29935366e-01
-2.60754347e-01 -4.65766281e-01 5.47800720e-01 -3.07249099e-01
-1.03286910e+00 9.93669212e-01 -7.57581294e-02 -2.42546886e-01
-2.94027537e-01 -4.41094786e-01 -7.33040869e-01 -4.07631636e-01
-5.38054943e-01 9.10270572e-01 2.78126806e-01 -8.02221000e-01
9.42905694e-02 4.87773925e-01 -1.72298789e+00 -4.28417832e-01
1.02472532e+00 2.81060070e-01 6.55514181e-01 8.92352343e-01
6.11563087e-01 5.14998734e-01 -9.72114503e-02 3.75141054e-01
4.24998969e-01 -2.21207649e-01 -4.27036703e-01 -1.51380479e-01
-3.29033792e-01 -4.90587234e-01 -3.33096296e-01 7.33851254e-01
4.60065842e-01 6.28575861e-01 9.01152670e-01 -2.08262277e+00
-2.54247367e-01 3.34233701e-01 -9.17001888e-02 -2.98871607e-01
6.26582325e-01 7.22541630e-01 6.57584250e-01 -3.68329525e-01
9.21032369e-01 1.46671307e+00 7.04615295e-01 5.37008584e-01
-1.08923519e+00 -6.53051078e-01 -9.68475565e-02 2.12983772e-01
-9.94862080e-01 -2.28057817e-01 4.06435430e-02 -5.05109370e-01
1.22609985e+00 -4.96952742e-01 8.16004932e-01 1.56740940e+00
4.33881253e-01 6.27538025e-01 8.85638833e-01 1.41137421e-01
4.33057576e-01 4.67910506e-02 -2.43799970e-01 6.42320275e-01
1.43291309e-01 5.66673815e-01 -2.56566495e-01 -2.88551059e-02
2.34168634e-01 6.90741986e-02 -1.82554409e-01 -1.18897653e+00
-1.47175336e+00 8.43420088e-01 7.71465421e-01 -1.67167678e-01
-5.87606311e-01 4.13937867e-01 2.79470026e-01 7.15510547e-01
-5.33265054e-01 4.45032746e-01 -6.65288389e-01 -6.81041539e-01
2.35230967e-01 7.06034958e-01 1.09189689e+00 1.39407372e+00
6.49751365e-01 -7.70460889e-02 7.79859364e-01 6.63992941e-01
5.17236173e-01 1.11919272e+00 3.45405608e-01 -1.48844290e+00
5.39071262e-01 5.88462293e-01 6.65853262e-01 -8.21853042e-01
-1.06325495e+00 3.39839488e-01 -1.77934214e-01 7.62318909e-01
5.01345515e-01 -3.85124177e-01 -4.17446405e-01 1.85318887e+00
4.02584076e-01 -4.71632093e-01 5.62727094e-01 1.10814178e+00
3.82944524e-01 4.02960479e-01 9.08926316e-03 3.84126872e-01
1.00634372e+00 -9.46883321e-01 -1.81578100e-01 -3.13465714e-01
1.02591300e+00 -1.88264653e-01 1.13834918e+00 2.19380170e-01
-3.96314770e-01 -3.40308964e-01 -1.03235292e+00 1.28420472e-01
-6.04761600e-01 -2.96397626e-01 1.02078892e-01 2.33105913e-01
-9.96469975e-01 5.30728400e-01 -8.82843137e-01 -1.21155918e+00
4.27415743e-02 7.19447374e-01 -7.44839787e-01 -1.17500037e-01
-1.09305048e+00 1.36433744e+00 -6.46350300e-03 -2.38145620e-01
-1.12171435e+00 -8.43711570e-02 -1.25905728e+00 -3.47735882e-01
3.82500261e-01 -6.75906003e-01 1.31509972e+00 -2.46712938e-01
-1.45850468e+00 5.82915366e-01 3.55613440e-01 -6.78407550e-01
8.17273378e-01 -3.29681695e-01 -1.03007019e-01 -5.73516116e-02
7.01997876e-01 1.24476790e+00 6.14775717e-01 -1.28990483e+00
-1.01432896e+00 -2.85490245e-01 1.14254095e-02 6.07302070e-01
3.01955074e-01 -4.09650594e-01 1.63395137e-01 1.84391648e-01
-1.61138132e-01 -1.77491486e+00 -4.99632031e-01 4.38799150e-02
-1.87432125e-01 -1.84533730e-01 1.08999407e+00 -1.57533601e-01
-2.52091825e-01 -2.14301515e+00 5.11821449e-01 1.06244542e-01
-1.93617165e-01 -2.18671501e-01 -2.68095702e-01 5.77060103e-01
5.73009193e-01 -3.46972108e-01 -2.84685493e-01 -3.58610392e-01
5.42395234e-01 6.57796800e-01 -3.86091098e-02 6.04371369e-01
-2.34084442e-01 9.50318635e-01 -1.33983135e+00 -1.48214191e-01
3.82706136e-01 2.24158064e-01 -6.40277088e-01 1.45864189e-01
8.69460553e-02 9.08226013e-01 -2.83291429e-01 6.22611165e-01
6.88692555e-02 2.90655702e-01 6.53264225e-02 5.31660438e-01
-3.30811977e-01 -4.10097800e-02 -9.12121713e-01 1.72791278e+00
-5.13838470e-01 6.84386671e-01 5.05362511e-01 -7.05441594e-01
7.17092752e-01 -1.46537170e-01 5.79235971e-01 -9.12667215e-01
2.93019712e-01 5.84968805e-01 1.62359312e-01 -8.17007005e-01
7.63974607e-01 5.43685630e-02 -6.82085812e-01 6.06631696e-01
-1.46392062e-01 -6.24163985e-01 5.68682067e-02 3.69376689e-02
1.40071750e+00 6.13074481e-01 1.36373788e-01 -1.96743473e-01
2.43751258e-01 7.35011280e-01 4.18198913e-01 9.01668251e-01
-8.37547779e-01 1.52867168e-01 5.78320734e-02 -4.91792597e-02
-1.27945864e+00 -1.42174113e+00 2.52249569e-01 1.16107297e+00
5.99143326e-01 7.12073892e-02 -4.97377634e-01 -4.62392926e-01
4.96685743e-01 7.26153433e-01 -2.68247068e-01 -1.48582339e-01
-8.76260519e-01 -2.09528461e-01 4.48687643e-01 2.40142718e-01
5.98394632e-01 -1.47965837e+00 -1.25950694e+00 2.28611857e-01
-8.22861269e-02 -1.37363625e+00 -4.41974215e-02 4.46937293e-01
-3.71339858e-01 -1.39303648e+00 -4.41170901e-01 -1.01041877e+00
5.97363651e-01 5.94162464e-01 5.19312859e-01 -2.99906552e-01
-2.48063460e-01 1.14759409e+00 -5.99451900e-01 -3.19522977e-01
-6.20473087e-01 5.07450178e-02 8.47424448e-01 -6.80209219e-01
2.04478979e-01 -7.31338441e-01 -5.93640149e-01 8.71259809e-01
-1.50311172e-01 -1.50591180e-01 5.50014317e-01 8.78288150e-01
2.11849250e-02 -9.03809816e-02 6.66716278e-01 -1.48228928e-02
6.88571870e-01 -9.10930455e-01 -5.86604893e-01 -4.52138335e-01
-4.60840315e-01 -2.38269523e-01 6.67825580e-01 -2.92812109e-01
-6.47399664e-01 2.58594036e-01 2.09716409e-01 1.57602370e-01
-4.38351154e-01 1.75370634e-01 1.55854821e-01 8.47715661e-02
7.43090749e-01 1.19254231e-01 6.15980804e-01 5.65587021e-02
5.77626348e-01 8.97437394e-01 6.27399504e-01 -5.60830951e-01
8.64846170e-01 5.99025548e-01 1.03806570e-01 -1.10486138e+00
4.84907448e-01 -4.60149378e-01 -6.43948972e-01 -4.21573848e-01
9.58887815e-01 -9.76469994e-01 -1.30413115e+00 4.17783886e-01
-7.99562991e-01 -1.04842699e+00 -2.41936043e-01 6.76541328e-01
-1.38170314e+00 6.66519161e-03 -3.06983769e-01 -7.60061204e-01
2.18994036e-01 -1.49698973e+00 9.85944152e-01 2.29170725e-01
-6.81090415e-01 -5.50805807e-01 9.53503624e-02 3.37299794e-01
4.22754228e-01 1.51413843e-01 5.74425876e-01 -5.18874884e-01
-6.50591075e-01 -1.29034203e-02 -8.17275867e-02 -4.55942862e-02
1.25043616e-01 -5.03443003e-01 -3.20009232e-01 -5.28124154e-01
-3.22075456e-01 -8.00575435e-01 1.63062736e-01 1.73925146e-01
-2.81782657e-01 1.73457995e-01 -6.02343917e-01 1.59961045e-01
8.28682363e-01 6.16489172e-01 2.90021151e-01 8.94768059e-01
4.71462816e-01 1.18215442e+00 1.17181396e+00 2.01746061e-01
1.17402351e+00 6.38956904e-01 7.72093177e-01 4.66631621e-01
3.26524049e-01 -4.63512152e-01 6.98212147e-01 3.78240407e-01
2.27294043e-01 -1.33763179e-01 -1.04660571e+00 5.31196535e-01
-2.23469520e+00 -7.28891134e-01 -2.04325348e-01 1.86115801e+00
-7.15812519e-02 6.59054667e-02 3.38314295e-01 -1.72725707e-01
4.02498811e-01 -5.48005402e-01 -1.01972175e+00 -3.27602327e-01
1.10152386e-01 -7.33175814e-01 6.98446393e-01 6.00584030e-01
-8.77597988e-01 8.46834719e-01 5.28222036e+00 -8.68355855e-02
-5.96891284e-01 -8.12174529e-02 -2.19888911e-01 -6.21395409e-02
2.70961583e-01 -1.17104776e-01 -6.78436160e-01 4.82661068e-01
1.02658880e+00 5.09736687e-02 7.32524335e-01 1.24967992e+00
7.02144444e-01 -6.98912978e-01 -1.25556123e+00 9.31998372e-01
-2.64383227e-01 -5.13450861e-01 -7.61419237e-01 3.92760247e-01
5.13606489e-01 7.69013941e-01 5.33040538e-02 8.55089664e-01
1.06904352e+00 -9.45843279e-01 8.10915589e-01 3.52558136e-01
1.90921620e-01 -6.62726879e-01 5.93200147e-01 9.60102558e-01
-9.93653715e-01 -6.94530785e-01 -1.25061378e-01 -3.32776129e-01
6.84415936e-01 -4.94529694e-01 -1.20695150e+00 1.08127713e-01
1.08905625e+00 8.12883139e-01 -1.25447258e-01 7.58150756e-01
-1.64959013e-01 -3.40151370e-01 -5.49723804e-01 -6.24378026e-01
6.24770463e-01 -6.06519222e-01 8.97761583e-01 4.54084009e-01
4.93621916e-01 -1.29518613e-01 2.18390420e-01 6.87161446e-01
4.97930974e-01 -2.02086836e-01 -1.40170097e+00 2.54964441e-01
2.89450556e-01 9.51905847e-01 -5.89409232e-01 1.63050592e-01
-1.07924566e-01 9.94457603e-01 2.22064003e-01 3.38490695e-01
-9.38762903e-01 -3.41474086e-01 1.07887995e+00 -1.85068056e-01
4.81978245e-02 -9.29108739e-01 1.54034033e-01 -7.46916711e-01
-3.68223116e-02 -8.27105284e-01 -1.98756769e-01 -1.03739095e+00
-9.82205689e-01 2.39425242e-01 -1.17298579e-02 -1.62795603e+00
-8.13770711e-01 -6.27440214e-01 -1.80940360e-01 3.82854044e-01
-1.34930098e+00 -8.39806437e-01 -5.10037541e-01 4.58923250e-01
5.91510952e-01 -5.25911212e-01 6.10037386e-01 4.19438668e-02
-1.37500629e-01 -6.52063787e-02 2.41389081e-01 -3.00679386e-01
7.22878337e-01 -1.13118863e+00 4.57022280e-01 1.50508881e-01
-4.94896829e-01 6.39646828e-01 1.00290918e+00 -8.24333549e-01
-1.65238929e+00 -7.57043004e-01 4.12770510e-01 -5.84878385e-01
7.86923587e-01 -4.08297032e-01 -3.35583419e-01 7.38891721e-01
-9.72851291e-02 -3.56904805e-01 2.90356100e-01 -3.22765529e-01
2.39408195e-01 1.73678696e-01 -1.39933109e+00 1.31510568e+00
1.47991824e+00 -1.65043607e-01 -6.70792758e-01 2.63377249e-01
7.35059440e-01 -3.02520156e-01 -4.43071276e-01 3.88029337e-01
8.67551744e-01 -9.43406403e-01 9.30777490e-01 -3.67436856e-01
1.42046139e-01 -4.71308261e-01 -5.20710528e-01 -1.69606960e+00
1.80799410e-01 -4.87355232e-01 4.46799427e-01 5.54827929e-01
3.03145051e-01 -1.00440478e+00 8.27536047e-01 5.68040609e-01
-3.67685109e-01 -3.25934500e-01 -1.23560429e+00 -1.13123953e+00
-8.89392570e-03 -5.18498182e-01 2.98581809e-01 3.02600175e-01
3.80053222e-01 3.11426759e-01 -3.45156103e-01 1.27227068e-01
7.51867771e-01 -3.58251482e-01 1.60523784e+00 -1.07286191e+00
1.45492524e-01 -2.64452785e-01 -6.07949793e-01 -1.18302596e+00
5.89392781e-01 -7.84143031e-01 7.51548409e-01 -1.61108100e+00
-3.84500593e-01 -8.91396701e-01 5.42471826e-01 1.20761573e-01
5.68473816e-01 -1.23268493e-01 1.90147892e-01 2.82588989e-01
-7.94406295e-01 8.95061553e-01 1.11260843e+00 -1.41284138e-01
-3.64253193e-01 4.42740228e-03 -1.48044169e-01 8.25187266e-01
8.23673546e-01 -5.11618912e-01 -6.44548714e-01 -1.40412554e-01
5.68642206e-02 2.87303239e-01 5.92978060e-01 -1.22859061e+00
5.95681369e-01 -1.02332383e-01 -1.47418380e-01 -5.01731873e-01
6.66075706e-01 -1.20175743e+00 6.15825467e-02 1.15998745e+00
-6.56440035e-02 2.92355418e-01 -1.00354552e-01 9.64105904e-01
2.27776289e-01 3.58363218e-03 6.38395369e-01 -2.08899692e-01
-1.13565838e+00 -5.87316975e-02 -1.13197792e+00 2.09024679e-02
1.54072022e+00 -4.31928724e-01 -2.95930237e-01 -7.37694919e-01
-6.40680075e-01 1.17311120e+00 9.56157327e-01 8.36426973e-01
6.67856395e-01 -1.10509753e+00 -2.80341506e-01 2.35598013e-01
2.96593308e-01 -1.53986402e-02 1.47933468e-01 8.33115935e-01
-6.45862818e-01 3.95682335e-01 -5.65221429e-01 -8.98421586e-01
-8.89296353e-01 1.13895059e-01 3.22499216e-01 3.10517043e-01
-9.63255346e-01 3.71555597e-01 -1.19900323e-01 -1.52867568e+00
1.29788801e-01 -3.36083859e-01 -1.92931816e-01 -1.94194436e-01
-1.42610535e-01 5.84196746e-01 -6.08673036e-01 -9.89100039e-01
-3.93249989e-01 3.96631122e-01 3.41991812e-01 -4.86240298e-01
1.30174065e+00 -7.12832630e-01 2.92412817e-01 4.67719525e-01
1.07046008e+00 -5.99563241e-01 -1.56859553e+00 1.97331831e-01
1.44026384e-01 -1.16168477e-01 -8.63777459e-01 -4.11134690e-01
-1.22993320e-01 2.65406579e-01 6.95836604e-01 9.71124787e-03
2.20664859e-01 -7.62550533e-02 1.03419518e+00 1.10542691e+00
1.28718591e+00 -1.44434333e+00 4.10883874e-01 9.79514539e-01
9.49805856e-01 -1.78825748e+00 -4.22087729e-01 1.11643627e-01
-9.01603460e-01 6.86567903e-01 7.87442029e-01 -3.10516000e-01
4.91115212e-01 3.83586846e-02 2.72125870e-01 2.17427872e-02
-5.41172206e-01 -2.41442859e-01 -6.45215750e-01 1.26909924e+00
-6.02214754e-01 1.68785334e-01 1.75042987e-01 2.21020594e-01
-7.76905417e-01 -2.69139022e-01 7.88025260e-01 1.33279002e+00
-6.66339338e-01 -7.86104858e-01 -3.63549680e-01 1.72476783e-01
5.00089884e-01 6.33077979e-01 -9.46819317e-03 1.30259013e+00
-5.07904813e-02 1.30405915e+00 6.22521248e-03 -5.97083330e-01
7.39108086e-01 -1.19158261e-01 1.98271856e-01 -5.63428164e-01
-2.52299607e-01 -5.52945554e-01 2.55120724e-01 -6.94992423e-01
-1.98088005e-01 -1.16945672e+00 -1.84461665e+00 -3.12030315e-01
3.27953637e-01 5.96393831e-02 1.09619200e+00 8.75376046e-01
2.72356272e-01 3.12643647e-01 5.88663578e-01 -1.44627798e+00
-7.61364043e-01 -7.69104242e-01 -6.13516629e-01 4.26721305e-01
4.71336633e-01 -1.09127498e+00 -6.61812544e-01 -4.42694068e-01]
|
[4.8320207595825195, 0.96552973985672]
|
b1a8abcd-2326-44df-8c23-7efb21600477
|
constrained-crystals-deep-convolutional
| null | null |
https://www.nature.com/articles/s41524-021-00526-4
|
https://www.nature.com/articles/s41524-021-00526-4.pdf
|
Constrained crystals deep convolutional generative adversarial network for the inverse design of crystal structures
|
Autonomous materials discovery with desired properties is one of the ultimate goals for materials science, and the current studies have been focusing mostly on high-throughput screening based on density functional theory calculations and forward modeling of physical properties using machine learning. Applying the deep learning techniques, we have developed a generative model, which can predict distinct stable crystal structures by optimizing the formation energy in the latent space. It is demonstrated that the optimization of physical properties can be integrated into the generative model as on-top screening or backward propagator, both with their own advantages. Applying the generative models on the binary Bi-Se system reveals that distinct crystal structures can be obtained covering the whole composition range, and the phases on the convex hull can be reproduced after the generated structures are fully relaxed to the equilibrium. The method can be extended to multicomponent systems for multi-objective optimization, which paves the way to achieve the inverse design of materials with optimal properties.
|
['Zhang H.', 'Gutfleisch O.', 'Shen C.', 'Samathrakis I.', 'Zhang Y.', 'Opahle I.', 'Fortunato N.M.', 'Long T.']
|
2021-05-10
| null | null | null |
npj-computational-materials-2021-5
|
['formation-energy']
|
['miscellaneous']
|
[ 1.97368294e-01 4.11029868e-02 -1.21203333e-01 -1.56963989e-01
-6.54649317e-01 -1.81976557e-01 4.89296287e-01 -5.23610003e-02
-1.28926724e-01 1.04457974e+00 2.99331844e-02 -6.73719868e-02
-4.17136401e-01 -1.00710118e+00 -6.68866098e-01 -1.54961610e+00
-1.76383201e-02 9.45424020e-01 1.13289140e-01 -3.06487143e-01
2.26396844e-01 3.94002706e-01 -1.68899274e+00 4.77315113e-02
1.14742720e+00 1.03898180e+00 4.80763584e-01 2.57174075e-01
-5.88428713e-02 1.70157805e-01 -1.36053219e-01 -1.15922078e-01
2.47228779e-02 -5.00624120e-01 -3.30084622e-01 -1.50820434e-01
-2.50430197e-01 6.16072305e-02 -2.91321333e-02 1.01146579e+00
5.71461141e-01 3.67023833e-02 1.13330925e+00 -7.11174965e-01
-1.00029683e+00 5.74619710e-01 -1.35199532e-01 -6.52323723e-01
1.90525070e-01 3.20213944e-01 1.14510608e+00 -8.58293116e-01
5.80569625e-01 9.79724228e-01 1.55684948e-01 5.92154562e-01
-1.53387713e+00 -4.95563835e-01 -5.01309298e-02 3.79614443e-01
-1.22480500e+00 -3.02078635e-01 8.69160116e-01 -6.34705067e-01
9.59200084e-01 2.50630796e-01 9.61753190e-01 9.12008226e-01
5.83706200e-01 5.46090007e-01 1.09948611e+00 -3.51796329e-01
6.37167931e-01 5.39216474e-02 1.42969107e-02 6.49734199e-01
5.82097650e-01 4.06938076e-01 -6.10647559e-01 -1.22885562e-01
2.88100779e-01 -1.87892064e-01 -1.92573756e-01 -6.64643943e-01
-9.06186938e-01 9.25732732e-01 5.28505623e-01 2.45878220e-01
-6.40357792e-01 5.33900075e-02 -2.22914651e-01 -1.79476798e-01
4.75944310e-01 9.89112377e-01 -3.49885166e-01 1.00476287e-01
-9.02734935e-01 5.57193100e-01 6.75833583e-01 5.35197914e-01
9.09832835e-01 2.14886487e-01 -8.34076181e-02 5.39391041e-01
6.63853645e-01 6.33029938e-01 1.12193786e-01 -6.73546553e-01
-1.77665040e-01 7.09328353e-01 2.42005527e-01 -2.24854082e-01
-3.27363193e-01 -5.81613839e-01 -6.89738989e-01 4.23085004e-01
8.61737281e-02 -2.73167253e-01 -1.09376991e+00 1.45999372e+00
3.61825824e-01 -6.15128458e-01 6.46844730e-02 8.53171229e-01
8.52581322e-01 8.56406569e-01 1.11614592e-01 -4.00194198e-01
9.89868641e-01 -5.99556386e-01 -5.20302773e-01 7.26337880e-02
3.51163894e-01 -2.88490444e-01 7.07867563e-01 5.56617022e-01
-1.10902643e+00 -1.92772239e-01 -1.17958188e+00 3.16836029e-01
-5.04363358e-01 4.98026870e-02 9.12728012e-01 6.89971983e-01
-7.80845642e-01 1.05912912e+00 -9.77051437e-01 9.15712416e-02
3.70868206e-01 7.06212461e-01 -1.37179762e-01 9.90496278e-02
-1.12808776e+00 7.90223956e-01 6.38839722e-01 2.40981683e-01
-1.05464554e+00 -6.67092264e-01 -4.86894071e-01 1.34024350e-02
3.52432489e-01 -9.17078137e-01 6.78782225e-01 -7.32578456e-01
-2.11891055e+00 4.59427357e-01 -1.85183182e-01 3.78998667e-02
2.81842649e-01 3.35583882e-03 -1.69388011e-01 -2.88622320e-01
1.39736220e-01 5.70127308e-01 5.91898561e-01 -1.24043369e+00
-4.57552187e-02 -2.43416041e-01 -4.06158864e-01 1.40337953e-02
-3.46783429e-01 -2.97176540e-01 1.25443861e-01 -1.42139494e-01
2.99686342e-01 -1.08598101e+00 -5.95445931e-01 -5.12439072e-01
-7.24367321e-01 -7.63441771e-02 2.82577038e-01 -4.03107166e-01
9.61460173e-01 -1.58526385e+00 1.04068446e+00 3.89080852e-01
2.60033399e-01 4.29196954e-02 1.62452251e-01 7.34181345e-01
-1.18848614e-01 4.89791371e-02 -3.24070215e-01 1.32966071e-01
-1.49165059e-03 -2.49719396e-01 7.85100535e-02 4.10766512e-01
1.83948398e-01 1.13551950e+00 -5.87586403e-01 -1.74946208e-02
1.72266692e-01 2.32898831e-01 -6.52641416e-01 1.89089492e-01
-7.36303687e-01 6.21885121e-01 -7.16463923e-01 5.42090476e-01
6.08592749e-01 -4.30909663e-01 3.38620991e-01 -3.84853035e-02
-4.47693020e-01 2.89463639e-01 -8.15338612e-01 1.22923970e+00
-1.09893635e-01 1.27372622e-01 8.52083266e-02 -8.60695183e-01
9.64375138e-01 1.36948153e-01 7.85559297e-01 -8.19184721e-01
9.83046070e-02 3.78320396e-01 4.94539946e-01 -3.37780744e-01
2.47791737e-01 -6.10626519e-01 -2.28710882e-02 3.86717975e-01
6.75693154e-02 -2.46491581e-01 3.89174998e-01 -2.85154045e-01
6.16420925e-01 3.40940475e-01 2.73456220e-02 -4.80827183e-01
5.72367251e-01 1.53566882e-01 2.51262188e-01 5.11503398e-01
5.11945128e-01 4.22827303e-01 3.78807873e-01 -2.64765650e-01
-1.28382540e+00 -1.12041342e+00 -2.91420072e-01 8.12971711e-01
1.90499246e-01 -3.24632972e-01 -6.81986034e-01 -7.36431638e-03
1.74948946e-02 6.16329074e-01 -5.89564681e-01 -5.73741853e-01
-3.89905244e-01 -1.44347417e+00 -3.51468354e-01 1.79063648e-01
2.00376794e-01 -1.18091702e+00 -2.63330311e-01 4.49212015e-01
2.98425466e-01 -5.72641075e-01 2.81140119e-01 6.44916952e-01
-6.50637627e-01 -7.85709620e-01 -7.87147701e-01 -4.46953297e-01
5.64542055e-01 -2.85358518e-01 5.66345036e-01 -2.05432221e-01
-4.39600587e-01 -1.14241146e-01 -4.76690568e-02 -3.12038392e-01
-8.31015706e-01 3.79940927e-01 2.48115703e-01 4.28906418e-02
-3.56764346e-02 -7.54647017e-01 -6.21839643e-01 1.68972284e-01
-4.67604786e-01 4.86984581e-01 7.60358930e-01 7.58907139e-01
8.09177160e-01 1.22726545e-01 8.12243521e-01 -6.89057767e-01
5.98553300e-01 -4.55146372e-01 -7.98000097e-01 3.00280660e-01
-9.40462887e-01 7.50431120e-01 6.58512950e-01 -3.80298853e-01
-1.16838396e+00 1.62341520e-01 -4.62549090e-01 1.01899311e-01
-3.15222107e-02 5.42759657e-01 -6.58655107e-01 -1.27403915e-01
4.35119301e-01 6.18188679e-01 -8.55545700e-03 -5.22590578e-01
4.13984269e-01 3.49291056e-01 -1.29631147e-01 -7.74598718e-01
6.82028353e-01 2.82581747e-01 4.31773275e-01 -1.00130045e+00
-5.34516096e-01 5.58109991e-02 -5.44153392e-01 -4.09622610e-01
1.04401827e+00 -6.07730448e-01 -9.71599996e-01 5.45206428e-01
-6.60948634e-01 -2.51745671e-01 -3.26915443e-01 4.58947033e-01
-6.40258014e-01 1.26524314e-01 -3.56045514e-01 -8.17462623e-01
-5.27156770e-01 -1.46991849e+00 1.06786072e+00 3.68864089e-01
-2.10820913e-01 -7.49165773e-01 2.90301621e-01 1.71748847e-01
5.50728261e-01 3.99732850e-02 1.50062919e+00 -4.34122771e-01
-1.05371857e+00 -8.38515162e-02 2.50161946e-01 -2.83386968e-02
2.87201941e-01 1.76229700e-01 -7.10796595e-01 -2.04423666e-01
-1.49338573e-01 -1.56931385e-01 1.02192998e+00 9.10166860e-01
7.68602073e-01 -3.28659452e-02 -4.70328867e-01 6.27912462e-01
1.39407170e+00 5.39260030e-01 5.94900191e-01 2.87859619e-01
7.31752276e-01 4.76859719e-01 1.05638191e-01 3.66521478e-01
-4.08598371e-02 6.71368122e-01 5.31728804e-01 4.03135680e-02
1.36812791e-01 -2.38435462e-01 3.75320524e-01 4.77324426e-01
-3.39725405e-01 -2.11128801e-01 -7.54002690e-01 6.85314760e-02
-1.83676732e+00 -9.95710492e-01 -2.78191388e-01 2.19842339e+00
6.65029347e-01 1.75017416e-01 1.08080171e-01 -2.02487037e-01
5.54360509e-01 1.45967538e-02 -9.50058460e-01 -1.31076857e-01
-1.34888664e-01 3.84603143e-01 4.79876250e-01 4.19516385e-01
-7.78622746e-01 9.01313305e-01 7.48951292e+00 7.69297183e-01
-1.38224077e+00 -2.07209438e-01 5.46996772e-01 -1.84603959e-01
-5.73947608e-01 2.39080369e-01 -9.56770897e-01 5.75500667e-01
9.03387189e-01 -2.53278047e-01 4.35689926e-01 6.89813673e-01
3.01846236e-01 -1.64973184e-01 -9.18519676e-01 6.17926121e-01
-5.12430131e-01 -1.70452440e+00 -1.01012383e-02 4.46329892e-01
7.93308079e-01 -1.91331252e-01 1.26679257e-01 -1.75929535e-02
1.79450214e-01 -1.00542104e+00 8.64333153e-01 7.37058818e-01
7.47322381e-01 -6.14245772e-01 2.37809524e-01 3.58175397e-01
-7.83294916e-01 -1.41906189e-02 -2.34549895e-01 -1.97022110e-02
2.88854152e-01 6.38149977e-01 -8.70481372e-01 4.54018623e-01
4.88481075e-01 5.25598347e-01 -2.75575757e-01 8.77637088e-01
-1.16629258e-01 3.34406197e-01 -3.30789387e-01 -6.98105991e-01
-1.18750915e-01 -9.82704401e-01 5.74515641e-01 5.14385045e-01
3.18431556e-01 -2.14981079e-01 -9.42168757e-02 1.61674511e+00
2.09623411e-01 2.44433619e-02 -3.28777850e-01 -4.75895911e-01
-2.16902560e-03 1.05174553e+00 -8.92278552e-01 -1.48461282e-01
-2.73031425e-02 5.03115654e-01 1.97798580e-01 4.13167834e-01
-7.20595658e-01 6.04188852e-02 5.51257551e-01 4.23709095e-01
5.36086857e-01 -2.46664122e-01 -2.96722412e-01 -1.02699578e+00
-1.72146812e-01 -4.97385144e-01 -3.10460299e-01 -5.62109113e-01
-1.05070961e+00 2.49863222e-01 5.84943444e-02 -6.09994888e-01
-8.59510750e-02 -8.31496119e-01 -5.96495926e-01 7.98169076e-01
-1.13678074e+00 -8.35325181e-01 9.63458419e-02 4.99942712e-03
7.86256045e-02 -2.08135784e-01 6.75373971e-01 1.62413597e-01
-8.50103736e-01 8.57392773e-02 7.54144371e-01 -6.61706388e-01
1.85852259e-01 -1.18006504e+00 4.22372557e-02 6.35572851e-01
-1.52064055e-01 4.92947191e-01 9.65063870e-01 -9.02446687e-01
-1.77685010e+00 -7.26022124e-01 3.33899826e-01 -9.67997462e-02
6.00397646e-01 -6.43699646e-01 -8.95308495e-01 -4.58460450e-02
-2.63299812e-02 -5.62317789e-01 6.69256508e-01 1.10068031e-01
3.80290538e-01 -1.20640863e-02 -6.87957108e-01 5.90174437e-01
9.18997884e-01 -2.51125824e-02 1.59399975e-02 6.14605188e-01
5.47774851e-01 -1.88384607e-01 -8.56588006e-01 4.25811857e-01
6.85406625e-01 -7.88700759e-01 9.89101470e-01 -7.49405742e-01
5.26147783e-01 -1.18533246e-01 -2.99660442e-03 -1.16974950e+00
-6.52114034e-01 -5.33932924e-01 -6.43514693e-02 9.46469188e-01
8.23877513e-01 -6.21111751e-01 9.58333850e-01 8.07110429e-01
-2.63882697e-01 -1.05830753e+00 -9.16476727e-01 -7.26136744e-01
1.26269847e-01 2.38034986e-02 7.40311265e-01 6.07555658e-02
-3.14886510e-01 5.84675252e-01 -3.60010028e-01 -5.76440580e-02
5.37699997e-01 5.82883298e-01 3.39728177e-01 -1.39776194e+00
-5.45024395e-01 -5.86641252e-01 7.63791651e-02 -9.05018806e-01
2.30949342e-01 -8.46774101e-01 1.10060215e-01 -1.51861894e+00
3.93862069e-01 -5.86659431e-01 -4.50350009e-02 6.69502560e-03
-2.51712464e-02 -3.26902300e-01 -4.84611243e-02 1.72509700e-01
-3.71152222e-01 1.28590250e+00 1.41485512e+00 -2.17005789e-01
-3.96906257e-01 2.38331974e-01 -9.20166492e-01 2.84201980e-01
5.65023482e-01 -3.87297451e-01 -1.41858071e-01 2.60592431e-01
7.17363179e-01 5.97513244e-02 -1.02164387e-03 -1.03105628e+00
-2.36560583e-01 -3.53076905e-01 3.31200540e-01 -3.93865824e-01
5.51625609e-01 -6.63784027e-01 7.24902868e-01 6.02599502e-01
-2.37846211e-01 -5.30954182e-01 -8.11551586e-02 5.03027380e-01
6.38338104e-02 -3.47235322e-01 8.82042766e-01 -1.24028660e-01
-3.22272837e-01 5.33580244e-01 -5.29387832e-01 -5.89707911e-01
1.14810216e+00 -3.98772210e-01 5.85288256e-02 -1.15651891e-01
-9.40738738e-01 1.97255164e-02 5.85687697e-01 2.65866444e-02
5.51585674e-01 -1.26469219e+00 -5.59957743e-01 3.05677265e-01
-1.48844585e-01 -1.44531457e-02 3.55047435e-01 5.83780110e-01
-5.85928023e-01 6.03652298e-01 -1.95636928e-01 -7.16984987e-01
-9.62746978e-01 5.74915767e-01 6.59541845e-01 -3.69377583e-01
-2.46742308e-01 5.41437924e-01 4.01368886e-01 -1.44915879e-01
-4.33402121e-01 -8.56965408e-03 -1.39661357e-01 5.39058186e-02
-3.48301902e-02 2.82894671e-01 2.52066761e-01 -4.86497909e-01
-3.53486896e-01 6.36227250e-01 -1.72437250e-03 5.04759811e-02
1.99193299e+00 2.67604709e-01 -2.55605340e-01 8.10046643e-02
9.69387770e-01 -1.06143877e-01 -1.24920952e+00 2.22162858e-01
-1.35043189e-01 6.52969927e-02 1.83626130e-01 -6.12048686e-01
-8.38373244e-01 7.39783168e-01 6.65473759e-01 1.36267766e-01
8.71323705e-01 3.50650042e-01 5.02464652e-01 3.69756162e-01
3.55054170e-01 -9.84954357e-01 8.93578604e-02 2.47208804e-01
7.92620897e-01 -1.14066482e+00 2.51257420e-01 -3.17059070e-01
-3.35684121e-01 1.14683139e+00 5.41236103e-01 -8.67086500e-02
6.83203697e-01 1.81482986e-01 -6.45248055e-01 -5.44539750e-01
-8.02497923e-01 -1.57246172e-01 3.43229830e-01 4.34346050e-01
4.78486329e-01 3.26939225e-01 -3.27699989e-01 4.45195138e-01
-1.15695417e-01 -3.60694557e-01 2.32068747e-01 7.27454722e-01
-5.80151975e-01 -1.60011590e+00 -1.51644185e-01 5.90707302e-01
-1.08133323e-01 -1.20753512e-01 -5.04044533e-01 4.94590461e-01
1.13971993e-01 4.71288711e-01 -3.14351112e-01 -2.59872317e-01
9.92832556e-02 2.72233605e-01 7.16283679e-01 -6.77715480e-01
-2.39243582e-01 2.19904572e-01 -5.53834699e-02 -2.31247067e-01
-1.90102026e-01 -7.72966564e-01 -1.10943079e+00 -7.33962879e-02
-8.06758165e-01 3.34321260e-01 6.90939903e-01 1.00480831e+00
5.14322221e-01 6.55209243e-01 5.81461489e-01 -1.12710178e+00
-2.40498602e-01 -6.91094279e-01 -6.27333283e-01 -1.24243976e-04
1.75035074e-01 -8.98490250e-01 -1.36420250e-01 -3.05762768e-01]
|
[5.211695671081543, 5.301068305969238]
|
94fb02d3-a928-4805-bcb5-c098f4c44677
|
mac-mining-activity-concepts-for-language
|
1811.08925
| null |
http://arxiv.org/abs/1811.08925v1
|
http://arxiv.org/pdf/1811.08925v1.pdf
|
MAC: Mining Activity Concepts for Language-based Temporal Localization
|
We address the problem of language-based temporal localization in untrimmed
videos. Compared to temporal localization with fixed categories, this problem
is more challenging as the language-based queries not only have no pre-defined
activity list but also may contain complex descriptions. Previous methods
address the problem by considering features from video sliding windows and
language queries and learning a subspace to encode their correlation, which
ignore rich semantic cues about activities in videos and queries. We propose to
mine activity concepts from both video and language modalities by applying the
actionness score enhanced Activity Concepts based Localizer (ACL).
Specifically, the novel ACL encodes the semantic concepts from verb-obj pairs
in language queries and leverages activity classifiers' prediction scores to
encode visual concepts. Besides, ACL also has the capability to regress sliding
windows as localization results. Experiments show that ACL significantly
outperforms state-of-the-arts under the widely used metric, with more than 5%
increase on both Charades-STA and TACoS datasets.
|
['JIyang Gao', 'Ram Nevatia', 'Runzhou Ge', 'Kan Chen']
|
2018-11-21
| null | null | null | null |
['language-based-temporal-localization']
|
['computer-vision']
|
[-1.57451674e-01 -5.97525001e-01 -7.19507217e-01 -3.27486128e-01
-1.05491543e+00 -7.02224910e-01 6.92136526e-01 -2.89377067e-02
-6.63677275e-01 5.36519766e-01 5.91067255e-01 3.36111784e-01
-3.83689627e-02 -2.31309637e-01 -7.08174407e-01 -5.44852376e-01
-6.07754469e-01 -1.58760086e-01 6.14880025e-01 1.55429915e-01
1.49584815e-01 2.81458795e-01 -1.47620070e+00 6.03224814e-01
3.63679379e-01 1.41294909e+00 2.66172141e-01 4.09896761e-01
-1.11460693e-01 1.41868544e+00 -3.36204529e-01 1.26426145e-01
4.74559516e-02 -4.90569502e-01 -5.96131206e-01 9.92433950e-02
5.65250337e-01 -9.62685868e-02 -8.27313662e-01 7.69026101e-01
1.56326175e-01 4.18414652e-01 4.03167099e-01 -1.50881898e+00
-3.91348749e-01 2.09574938e-01 -5.18193662e-01 7.24126577e-01
8.81005466e-01 5.63424081e-02 1.23431623e+00 -1.16081655e+00
7.81510353e-01 1.16015065e+00 5.14055371e-01 2.38398567e-01
-1.02643788e+00 -6.85847104e-01 4.47794378e-01 6.85518026e-01
-1.82731140e+00 -5.45955598e-01 6.70631647e-01 -5.73190331e-01
1.08012664e+00 1.50851458e-01 8.07158351e-01 1.52884972e+00
-7.02709258e-02 1.07245064e+00 8.65629554e-01 -2.43683867e-02
3.31237525e-01 -1.63775235e-01 -2.44318098e-01 9.04060304e-01
-1.98985919e-01 -7.50219747e-02 -1.26199245e+00 -3.73189524e-02
5.36984980e-01 2.55985379e-01 -3.47577095e-01 -6.90107107e-01
-1.73758042e+00 7.07477272e-01 3.39625835e-01 4.66933161e-01
-3.04266661e-01 4.28758681e-01 4.80402857e-01 1.08178139e-01
4.96643841e-01 2.62446195e-01 -5.30597746e-01 -3.82265717e-01
-1.18934381e+00 3.62203754e-02 5.92299402e-01 1.07426763e+00
8.52614760e-01 -7.35380203e-02 -3.18327844e-01 5.90399563e-01
1.67107284e-01 4.48818773e-01 6.53118074e-01 -9.54651654e-01
5.94904602e-01 4.05764490e-01 -5.88872237e-04 -1.03480601e+00
-1.62057295e-01 -2.94738799e-01 -2.25873694e-01 -4.31154996e-01
2.27657512e-01 2.26950362e-01 -8.19913149e-01 1.79392004e+00
1.10483035e-01 7.71773517e-01 -1.13075934e-01 1.04040146e+00
6.40248716e-01 7.47449756e-01 3.28633249e-01 -3.41218501e-01
1.24638009e+00 -1.11291802e+00 -7.86436319e-01 -3.45988601e-01
7.57401884e-01 -4.04576540e-01 1.16428697e+00 1.55786574e-01
-6.66842818e-01 -5.35909951e-01 -9.75253344e-01 -6.10474832e-02
-4.44505334e-01 2.78918177e-01 5.95797479e-01 8.36161822e-02
-9.02028620e-01 6.42887577e-02 -1.11449599e+00 -7.58212566e-01
4.51054692e-01 3.58325019e-02 -6.51619554e-01 -1.88211158e-01
-9.65510666e-01 5.55198848e-01 4.75491315e-01 -2.58392155e-01
-1.26168478e+00 -6.14424288e-01 -1.20505822e+00 -5.15537746e-02
8.89843166e-01 -1.58516809e-01 9.67487931e-01 -9.71567869e-01
-1.07393527e+00 5.71005523e-01 -3.68177235e-01 -6.14019573e-01
3.73648435e-01 -3.87517899e-01 -5.52913845e-01 6.38739765e-01
3.89139801e-01 7.59030044e-01 7.53139853e-01 -8.05066407e-01
-9.04983699e-01 -1.02477826e-01 2.35062420e-01 2.33241454e-01
-6.14028633e-01 5.18429875e-02 -1.09931898e+00 -9.29173172e-01
3.09247941e-01 -8.69165599e-01 1.13735534e-01 3.80688995e-01
2.08081063e-02 -1.77575231e-01 1.13820446e+00 -7.21392214e-01
1.37381208e+00 -2.26396871e+00 2.12457135e-01 7.46829761e-03
1.02444746e-01 -3.05101991e-01 -3.54531020e-01 3.60593498e-01
2.25357771e-01 -8.15158635e-02 8.20083022e-02 -2.35764548e-01
-9.70266685e-02 2.55242825e-01 -2.90367365e-01 6.75397336e-01
5.63997403e-02 9.98120785e-01 -1.15743446e+00 -8.45999181e-01
2.46937335e-01 4.11432058e-01 -6.67736173e-01 2.09749699e-01
-3.65922689e-01 4.65737700e-01 -6.41737759e-01 1.08061028e+00
1.26480656e-02 -3.58328074e-01 1.91858605e-01 -4.27029312e-01
-6.06805831e-02 1.46509066e-01 -8.66945982e-01 2.36562490e+00
-4.57974672e-01 9.46067274e-01 -1.98534682e-01 -1.15449417e+00
4.01356131e-01 3.34952533e-01 1.06602859e+00 -7.99053788e-01
-3.46404105e-01 -1.85587816e-03 -5.33950090e-01 -8.46520066e-01
3.02297503e-01 2.89371908e-01 -4.05074447e-01 1.21028967e-01
6.28697455e-01 4.68111843e-01 2.58929491e-01 3.52075756e-01
1.45430744e+00 4.69205797e-01 3.37847650e-01 3.06138936e-02
6.78405523e-01 1.05307077e-03 7.26955295e-01 7.55854726e-01
-4.29379314e-01 4.14243519e-01 4.50892121e-01 -7.80889243e-02
-4.47062254e-01 -1.25522673e+00 3.45986485e-01 1.48563302e+00
2.95229852e-01 -8.96946013e-01 -2.97537565e-01 -1.08650506e+00
-1.62277833e-01 3.57671171e-01 -6.80430710e-01 -2.46565323e-02
-4.52477306e-01 -2.77650863e-01 4.91032451e-01 6.65621698e-01
5.27363300e-01 -7.88527191e-01 -5.25088787e-01 5.04955463e-02
-6.60300076e-01 -1.79226255e+00 -8.16350996e-01 -7.63124833e-03
-6.29431665e-01 -1.23454690e+00 -4.75547403e-01 -6.84820175e-01
3.64341050e-01 5.25024116e-01 8.69258761e-01 -2.58844852e-01
-3.42839301e-01 1.07107770e+00 -7.65392244e-01 1.42665803e-01
2.73722023e-01 -1.26132786e-01 8.22100565e-02 4.81060535e-01
6.67441487e-01 -4.07754928e-01 -7.61023641e-01 4.63037819e-01
-6.66065574e-01 -3.08241069e-01 5.26140988e-01 6.75919592e-01
7.57719338e-01 -2.15179175e-01 2.55704820e-01 -1.84037656e-01
1.17148094e-01 -7.68426061e-01 -4.75504398e-01 3.89620513e-01
-2.86657155e-01 -3.82840186e-02 1.74213260e-01 -7.08697379e-01
-8.09913099e-01 5.23801684e-01 4.42125350e-01 -9.77306485e-01
-3.73250172e-02 6.07965231e-01 -1.56595618e-01 -1.69476599e-01
4.53162581e-01 4.65224117e-01 -4.52264577e-01 -3.87997031e-01
3.40424985e-01 1.85182437e-01 6.03105605e-01 -4.03836787e-01
7.48745382e-01 9.08124447e-01 -1.86483204e-01 -1.01433051e+00
-9.42466199e-01 -1.16893435e+00 -6.80353820e-01 -3.80124211e-01
1.13275838e+00 -1.30987144e+00 -4.19732749e-01 -7.68388361e-02
-9.93394792e-01 3.12093943e-02 -5.72575368e-02 9.67359245e-01
-8.20994675e-01 4.91893023e-01 -5.24050534e-01 -6.43215358e-01
3.81229758e-01 -1.05260444e+00 1.18662953e+00 -9.49305743e-02
-1.48048446e-01 -8.40375721e-01 -4.82396781e-02 3.80175054e-01
9.72056538e-02 2.49586388e-01 3.88468653e-01 -6.61249042e-01
-1.01739907e+00 -2.34118566e-01 -1.64926052e-02 8.08928609e-02
-1.18305227e-02 -5.41921496e-01 -8.68775189e-01 -2.83465028e-01
-6.95938468e-02 -4.86158192e-01 1.03580284e+00 2.28259653e-01
1.38438058e+00 -3.88814300e-01 -5.70431173e-01 7.54688621e-01
1.42248237e+00 1.71962887e-01 3.52363557e-01 1.86633155e-01
6.24192774e-01 5.57083607e-01 1.05172634e+00 5.03571332e-01
2.32027650e-01 9.53040838e-01 2.71183580e-01 1.80641428e-01
1.68005936e-02 -5.38502216e-01 8.14387143e-01 7.33733833e-01
-3.56127955e-02 -1.81880176e-01 -8.74623775e-01 6.86188817e-01
-2.08299160e+00 -1.27898407e+00 4.76983577e-01 1.95190001e+00
5.51444769e-01 -6.11001030e-02 1.26431704e-01 -2.80113697e-01
3.00555855e-01 5.65941811e-01 -4.50762898e-01 6.04523659e-01
-2.12241858e-01 -6.01249225e-02 5.21006525e-01 1.49715036e-01
-1.50904179e+00 1.14824307e+00 5.61669159e+00 8.91448021e-01
-1.01896834e+00 6.74338996e-01 5.36330603e-02 -5.86562335e-01
2.32982948e-01 7.20458552e-02 -5.45639277e-01 5.03384471e-01
8.76094639e-01 -7.91757479e-02 3.61672729e-01 9.17433441e-01
5.41090786e-01 -3.63269329e-01 -1.32318175e+00 1.31071889e+00
5.36796987e-01 -1.29112840e+00 -7.88347423e-02 -5.54092638e-02
4.45005506e-01 2.69599497e-01 2.94254045e-03 5.24436176e-01
-2.32470348e-01 -7.91942716e-01 1.07150412e+00 6.61746383e-01
8.49361598e-01 -4.15421009e-01 3.18193138e-01 2.46363789e-01
-1.79681063e+00 -3.20431113e-01 -5.65448813e-02 2.84060359e-01
1.32878914e-01 -8.70954320e-02 -5.92144847e-01 2.94171691e-01
1.08951604e+00 1.35960913e+00 -7.31964886e-01 1.09785903e+00
-8.74810442e-02 7.12041378e-01 -1.37466609e-01 9.36859101e-02
5.18394947e-01 -3.76738310e-02 5.58218241e-01 1.45759368e+00
5.37165165e-01 -4.69751470e-02 6.55549228e-01 5.66257775e-01
2.49019191e-02 1.12068333e-01 -7.04470932e-01 -4.53656524e-01
3.97871166e-01 9.72545266e-01 -7.74982035e-01 -2.94468343e-01
-9.17572856e-01 1.28928256e+00 1.18763819e-01 7.74937630e-01
-1.16050196e+00 -1.91849560e-01 8.60224426e-01 2.60688961e-01
4.01202649e-01 -6.46235943e-01 5.02357960e-01 -1.58741951e+00
1.32362783e-01 -6.25579476e-01 5.55066109e-01 -8.48433495e-01
-9.98483539e-01 2.28982553e-01 2.73976445e-01 -1.70099187e+00
-1.92767128e-01 -4.54982877e-01 -7.41252005e-02 1.96748450e-01
-1.35753870e+00 -1.41570473e+00 -4.16357189e-01 9.25362945e-01
1.06077588e+00 -3.60852748e-01 4.52745289e-01 5.64451635e-01
-1.60203189e-01 3.33221585e-01 -1.07937604e-02 3.70819569e-01
7.66519666e-01 -8.47875178e-01 -3.07584733e-01 9.06194806e-01
8.28480244e-01 3.37930471e-01 4.14722353e-01 -5.27341127e-01
-1.67651844e+00 -1.28264523e+00 8.00582647e-01 -6.64806187e-01
1.08809972e+00 -6.58442318e-01 -6.57546401e-01 8.17497015e-01
-1.02669030e-01 4.58632946e-01 7.46319711e-01 -1.97325781e-01
-6.13842964e-01 -1.40496805e-01 -6.28057063e-01 4.55158323e-01
1.48788381e+00 -1.14970303e+00 -5.95938265e-01 5.03427267e-01
7.64672041e-01 -2.73670435e-01 -5.77908635e-01 2.05784574e-01
5.47192633e-01 -7.57719100e-01 1.22352910e+00 -6.19654775e-01
1.87446624e-02 -5.30508578e-01 -6.26186788e-01 -7.71966398e-01
-7.03582987e-02 -2.59228259e-01 -4.91370589e-01 8.57645035e-01
1.69770986e-01 -8.01326483e-02 8.02229464e-01 -4.16142382e-02
1.11838311e-01 -4.69839841e-01 -1.19796753e+00 -1.06594682e+00
-7.11749554e-01 -7.83765018e-01 6.78527728e-02 1.03301692e+00
8.32940713e-02 2.44223967e-01 -5.97134590e-01 2.50542045e-01
5.43818533e-01 -1.06393933e-01 4.27764267e-01 -8.29314590e-01
-2.52677113e-01 -1.16161093e-01 -8.44146550e-01 -1.19315612e+00
5.84387720e-01 -9.02525365e-01 1.80871591e-01 -1.27028775e+00
2.32833385e-01 -4.09872853e-04 -6.56333327e-01 5.70309103e-01
3.70405883e-01 3.89675945e-01 1.86368108e-01 4.49521393e-01
-1.53311563e+00 6.41103864e-01 6.91074848e-01 -3.81992310e-01
-8.80964473e-02 -3.13861489e-01 -9.12007093e-02 6.52419746e-01
3.10773164e-01 -5.17760456e-01 -7.18105972e-01 -3.14256340e-01
2.17235070e-02 1.44105315e-01 6.86340630e-01 -1.37857699e+00
3.94114822e-01 -3.57520759e-01 4.12713408e-01 -7.28008449e-01
7.38571227e-01 -8.78959715e-01 -9.41267088e-02 2.05960706e-01
-5.36584079e-01 -1.71584194e-03 -1.59229971e-02 1.27995121e+00
-6.65095806e-01 2.46498302e-01 4.11378115e-01 -1.50615990e-01
-1.47037911e+00 6.13978326e-01 -5.19053102e-01 1.26511365e-01
1.28038061e+00 -3.45638692e-01 -1.66825112e-02 -5.76038718e-01
-8.60491812e-01 3.66779834e-01 2.76228547e-01 8.29680800e-01
8.08816075e-01 -1.48337114e+00 -2.08030790e-01 -7.57002365e-03
6.82650685e-01 -5.64542770e-01 1.62299648e-01 1.20518720e+00
-2.71645695e-01 7.31254935e-01 7.99203143e-02 -8.79125416e-01
-1.25024354e+00 6.36050403e-01 2.87579238e-01 1.55334681e-01
-5.38942039e-01 7.85264730e-01 4.87567931e-01 9.61785391e-02
5.94396174e-01 -3.44081849e-01 -2.62588024e-01 2.97101557e-01
4.45807606e-01 1.60993055e-01 -1.86163396e-01 -1.01857007e+00
-6.60371542e-01 5.20506024e-01 2.41342157e-01 -3.63762915e-01
1.16332364e+00 -3.01596195e-01 2.47157186e-01 5.79113126e-01
1.54349554e+00 1.84783507e-02 -1.37237394e+00 -6.09678745e-01
5.44409692e-01 -6.24995589e-01 7.05071464e-02 -6.20003223e-01
-9.28065777e-01 8.77201438e-01 6.94792747e-01 -3.91329765e-01
1.13534617e+00 4.88908082e-01 5.54763436e-01 5.91708302e-01
5.50361037e-01 -1.34295511e+00 8.71949494e-01 5.53712308e-01
9.80408728e-01 -1.25234163e+00 -6.99867755e-02 -2.18141094e-01
-7.36412108e-01 8.45725715e-01 6.76499784e-01 2.43371084e-01
6.49599493e-01 -2.84112487e-02 -1.91929936e-01 -1.82075843e-01
-9.06055033e-01 -4.58904445e-01 5.81444323e-01 6.02522731e-01
2.87948787e-01 -1.86701074e-01 -1.64290413e-01 6.16276622e-01
4.50830042e-01 4.43622842e-02 3.42990234e-02 1.16361523e+00
-4.14805412e-01 -7.17079401e-01 -2.44220525e-01 2.93545991e-01
-4.30166960e-01 8.43134429e-03 -2.10536316e-01 8.05518568e-01
2.76841134e-01 7.93165565e-01 1.74753904e-01 -5.35850227e-01
8.68103877e-02 1.69266015e-01 3.43808115e-01 -6.64743721e-01
-7.85182193e-02 4.07270223e-01 -2.99932938e-02 -1.32303870e+00
-9.25894558e-01 -8.62906456e-01 -1.19388974e+00 3.13442469e-01
-1.52860004e-02 2.50647098e-01 4.57825005e-01 9.32260096e-01
3.48717958e-01 2.65731066e-01 4.02385682e-01 -7.74353981e-01
-2.70002455e-01 -7.77109265e-01 -5.29753506e-01 3.57004642e-01
4.78987038e-01 -1.05514097e+00 -4.54411983e-01 3.37719887e-01]
|
[9.550318717956543, 0.7174724340438843]
|
e0b06a18-8ec3-447c-973b-1d8438b88449
|
dialogueein-emotion-interaction-network-for
| null | null |
https://aclanthology.org/2022.coling-1.57
|
https://aclanthology.org/2022.coling-1.57.pdf
|
DialogueEIN: Emotion Interaction Network for Dialogue Affective Analysis
|
Emotion Recognition in Conversation (ERC) has attracted increasing attention in the affective computing research field. Previous works have mainly focused on modeling the semantic interactions in the dialogue and implicitly inferring the evolution of the speakers’ emotional states. Few works have considered the emotional interactions, which directly reflect the emotional evolution of speakers in the dialogue. According to psychological and behavioral studies, the emotional inertia and emotional stimulus are important factors that affect the speaker’s emotional state in conversations. In this work, we propose a novel Dialogue Emotion Interaction Network, DialogueEIN, to explicitly model the intra-speaker, inter-speaker, global and local emotional interactions to respectively simulate the emotional inertia, emotional stimulus, global and local emotional evolution in dialogues. Extensive experiments on four ERC benchmark datasets, IEMOCAP, MELD, EmoryNLP and DailyDialog, show that our proposed DialogueEIN considering emotional interaction factors can achieve superior or competitive performance compared to state-of-the-art methods. Our codes and models are released.
|
['Qin Jin', 'Ruichen Li', 'Jingwen Hu', 'Jinming Zhao', 'Yuchen Liu']
| null | null | null | null |
coling-2022-10
|
['emotion-recognition-in-conversation']
|
['natural-language-processing']
|
[-7.17393637e-01 1.26894221e-01 1.92780375e-01 -7.62336254e-01
9.36518684e-02 -2.87812501e-01 6.16307497e-01 6.64280429e-02
-2.31879920e-01 5.01690507e-01 5.48267841e-01 4.17388648e-01
3.34235251e-01 -3.89043689e-01 2.40916327e-01 -4.94763166e-01
-8.86812210e-02 2.81056345e-01 -2.70141333e-01 -6.26853108e-01
1.11638963e-01 1.36991665e-01 -1.51823485e+00 3.20932835e-01
1.05836976e+00 1.07977760e+00 -3.85655344e-01 8.15496325e-01
-5.53667426e-01 1.20478368e+00 -5.88710845e-01 -6.69836760e-01
-6.50073647e-01 -7.53353953e-01 -1.01014650e+00 2.56680429e-01
-7.66816676e-01 2.19303712e-01 -2.99048033e-02 1.09343147e+00
8.75073016e-01 4.83180016e-01 6.14021480e-01 -1.39244473e+00
-4.16856349e-01 5.80371022e-01 -4.70248342e-01 1.19946778e-01
7.05975950e-01 -2.73198485e-01 7.93832660e-01 -8.67218018e-01
4.68078136e-01 1.51714218e+00 5.40256679e-01 9.19767916e-01
-5.72714627e-01 -5.18428743e-01 3.51729184e-01 7.09206402e-01
-9.64136004e-01 -5.64407349e-01 1.24667215e+00 -2.95972049e-01
8.91879261e-01 4.56767946e-01 9.23477948e-01 1.23467708e+00
1.92427844e-01 1.04892778e+00 1.17379522e+00 -3.52529407e-01
3.23135138e-01 7.68441081e-01 4.84826624e-01 2.07856342e-01
-9.24166501e-01 -3.21425050e-01 -5.70346594e-01 -1.83646798e-01
-4.74784039e-02 -4.48535591e-01 -2.71833181e-01 2.33410075e-01
-4.27452952e-01 8.84594500e-01 6.16089965e-04 5.08816659e-01
-6.02361917e-01 -3.64473015e-01 9.97683942e-01 4.12721872e-01
9.32719648e-01 1.03814334e-01 -5.01842856e-01 -8.81651163e-01
-1.24198191e-01 -2.37045869e-01 1.34830999e+00 5.56862831e-01
5.57388484e-01 -8.92400965e-02 -6.57278150e-02 1.54151046e+00
4.33885694e-01 7.77169466e-02 7.51076519e-01 -6.31283343e-01
-6.35119155e-02 7.24913359e-01 1.59979574e-02 -1.43059206e+00
-6.72349989e-01 1.01101071e-01 -9.82583642e-01 -4.22614813e-01
-4.55322981e-01 -7.70638704e-01 -1.58597082e-01 1.90333557e+00
6.60954058e-01 2.62539923e-01 5.99582434e-01 8.25794876e-01
1.27209532e+00 9.20423448e-01 4.37492818e-01 -6.45111680e-01
1.62294459e+00 -1.09666014e+00 -1.59307396e+00 -2.81901062e-01
3.58292907e-01 -9.10171330e-01 8.97345662e-01 2.19181478e-01
-1.05346620e+00 -4.46525097e-01 -5.81759572e-01 2.87877768e-01
-5.51099241e-01 -2.47025918e-02 8.01551759e-01 6.59951925e-01
-7.82007635e-01 1.06215611e-01 -5.02342165e-01 -4.54315722e-01
-1.45860806e-01 9.46220160e-02 -3.90438996e-02 7.16974616e-01
-1.92510509e+00 1.14755952e+00 1.12320088e-01 5.71814120e-01
-3.75574082e-01 -4.84364554e-02 -6.91489637e-01 1.31610274e-01
9.23229456e-02 -1.13534197e-01 1.38628316e+00 -1.49750161e+00
-2.24028492e+00 7.59269536e-01 -1.88652083e-01 1.21265374e-01
1.82254419e-01 -1.09387018e-01 -9.75719810e-01 -1.52315557e-01
-6.93926692e-01 4.87960517e-01 2.98257172e-01 -1.16418004e+00
-1.99319094e-01 -3.49770784e-01 -1.68424189e-01 9.25367475e-01
-5.64222217e-01 4.65224355e-01 -6.75896466e-01 -1.95649654e-01
-3.72675359e-01 -9.49160337e-01 -2.12691098e-01 -6.16566777e-01
-2.69703329e-01 -7.14874804e-01 8.00979316e-01 -5.26752651e-01
1.29880893e+00 -2.22981882e+00 3.65877301e-01 -6.39840774e-03
-2.92868763e-02 2.09712848e-01 1.07448354e-01 5.30089974e-01
-6.85413629e-02 -5.32701164e-02 1.68835998e-01 -5.75060248e-01
3.58867735e-01 2.67640263e-01 -1.29599139e-01 6.26386702e-02
-1.13615371e-01 7.30271757e-01 -7.74309516e-01 -6.81827307e-01
1.72870368e-01 7.81744719e-01 -4.66042191e-01 8.99039030e-01
3.47217657e-02 4.37907934e-01 -5.37304580e-01 1.99836165e-01
5.39053917e-01 1.33919224e-01 3.07643920e-01 -1.75175518e-01
-3.75717431e-02 -1.77643970e-02 -8.54974329e-01 1.32439530e+00
-7.94107020e-01 6.32237971e-01 5.06902337e-01 -7.33023524e-01
1.20984423e+00 7.04388857e-01 3.70101392e-01 -7.14422941e-01
6.47906423e-01 -1.39317170e-01 9.87004414e-02 -9.48880970e-01
6.05700970e-01 -4.92603153e-01 -4.89121109e-01 4.50383842e-01
-3.48385498e-02 -1.95059866e-01 -1.66476265e-01 1.35341227e-01
4.82454866e-01 -2.66983032e-01 2.85121828e-01 -1.82046801e-01
1.01229322e+00 -7.23340213e-01 7.66322017e-01 1.59729254e-02
-7.59672284e-01 -2.38696247e-01 8.08636963e-01 -7.70230070e-02
-2.51523048e-01 -4.49195534e-01 -1.51039129e-02 1.42912233e+00
4.21478510e-01 -2.55256265e-01 -1.05692506e+00 -4.64823544e-01
-5.71749747e-01 8.44831407e-01 -7.66087174e-01 -5.09987831e-01
-3.05621643e-02 -1.05212200e+00 4.62971568e-01 2.08466366e-01
7.42152095e-01 -1.63803995e+00 -4.03818697e-01 3.44679356e-01
-7.14796901e-01 -1.22392082e+00 -4.01865274e-01 4.09834534e-02
-2.86020368e-01 -5.61136067e-01 -2.28934944e-01 -5.57166159e-01
3.65822822e-01 -6.01815701e-01 1.05766726e+00 -2.98140764e-01
-3.31530482e-01 6.77568018e-01 -5.62154770e-01 -5.13684094e-01
-4.76483583e-01 -7.64789283e-02 1.08599812e-01 4.21491146e-01
6.52074635e-01 -5.19381464e-01 -6.34814858e-01 3.77922297e-01
-6.73187077e-01 8.42358470e-02 2.08069105e-02 5.34120202e-01
-6.31599575e-02 2.23755077e-01 9.20028389e-01 -8.76222968e-01
1.33521974e+00 -8.32814991e-01 3.08563024e-01 3.59176964e-01
-4.04597759e-01 -1.88993528e-01 6.30474031e-01 -4.96305585e-01
-1.93298495e+00 -3.89203280e-01 -5.09522617e-01 4.52586589e-03
-2.97846645e-01 6.35075808e-01 -3.87401193e-01 2.52752721e-01
-9.60726291e-03 8.42917413e-02 -2.23483920e-01 -1.88501403e-01
3.16600710e-01 1.31125116e+00 1.62125349e-01 -6.91977620e-01
-1.48277134e-01 1.00111410e-01 -7.41385520e-01 -8.91212642e-01
-7.53297269e-01 -4.83215332e-01 -2.59778857e-01 -8.52564633e-01
1.08149755e+00 -8.48969102e-01 -1.25639677e+00 8.96834731e-01
-1.27829993e+00 -1.40337527e-01 9.43359286e-02 4.00647104e-01
-3.89418274e-01 3.54364812e-01 -1.00281680e+00 -1.38368940e+00
-7.71272719e-01 -1.04930985e+00 6.80118501e-01 7.45433569e-01
-6.71166420e-01 -1.54313183e+00 3.14465612e-01 2.48749673e-01
6.11052275e-01 1.53061554e-01 7.45127797e-01 -7.83508778e-01
5.40282249e-01 3.83354090e-02 1.20418005e-01 2.92771637e-01
1.13537863e-01 2.13029742e-01 -9.50891376e-01 4.55456108e-01
3.54579628e-01 -7.54062891e-01 1.13716997e-01 -9.92385671e-02
1.01641679e+00 -2.52260178e-01 6.29145801e-02 1.16397530e-01
8.35458100e-01 6.61968887e-01 8.79994512e-01 -2.21254349e-01
2.39033580e-01 1.09658372e+00 7.60776579e-01 1.00085115e+00
8.51649582e-01 4.79468167e-01 1.96939543e-01 -2.19997168e-01
6.59375966e-01 1.03690840e-01 5.99184692e-01 1.75758362e+00
-1.94943398e-02 -4.45411086e-01 -5.66946805e-01 3.21757495e-01
-2.10576534e+00 -9.54354405e-01 -2.41962701e-01 1.38698709e+00
1.14515698e+00 -3.24600309e-01 -1.89222708e-01 -2.73272008e-01
8.44790876e-01 4.17914063e-01 -6.70495272e-01 -1.33949339e+00
-1.88800305e-01 -1.28543586e-01 -5.23311436e-01 6.49723887e-01
-7.91072488e-01 1.07823741e+00 5.61921740e+00 5.99486589e-01
-1.00281453e+00 1.57089725e-01 1.04184353e+00 8.80454406e-02
-3.10614407e-01 -4.30458456e-01 -3.88463378e-01 5.02719998e-01
1.19384837e+00 -3.44670475e-01 5.57796538e-01 7.80564249e-01
3.82916808e-01 -9.99603719e-02 -8.14004183e-01 1.12646174e+00
3.28172117e-01 -2.45096251e-01 -5.61875045e-01 -3.33627135e-01
4.50167626e-01 -3.18047315e-01 -5.87027445e-02 8.53121579e-01
3.49447668e-01 -7.77543008e-01 1.30821407e-01 9.55957592e-01
3.37883234e-01 -1.05710995e+00 1.02772880e+00 2.94275284e-01
-1.13859427e+00 3.30726653e-01 -7.23049343e-02 -9.11801159e-02
2.85548508e-01 3.97784591e-01 -4.56455231e-01 3.49039972e-01
7.55640268e-01 6.72308087e-01 -2.17249334e-01 1.46645354e-02
-2.22209483e-01 6.34436786e-01 -7.68126771e-02 -5.87241590e-01
1.22456305e-01 -6.36335731e-01 2.46742815e-01 1.55228794e+00
9.82190967e-02 6.35195255e-01 -1.54269814e-01 5.19550383e-01
-5.06482460e-02 6.55678570e-01 -6.43368438e-02 -4.24345553e-01
3.46726686e-01 1.80899978e+00 -3.76073867e-01 -4.06747103e-01
-1.52855307e-01 1.38352811e+00 2.91752040e-01 2.06422463e-01
-1.27953351e+00 -5.41130424e-01 8.10365260e-01 -8.87293339e-01
-2.24177420e-01 3.47688258e-01 1.01610519e-01 -1.22748840e+00
-2.74542391e-01 -8.90008867e-01 3.24430197e-01 -8.43086541e-01
-1.38928986e+00 9.78008509e-01 -4.69655305e-01 -6.16833627e-01
-2.74068147e-01 -1.97682112e-01 -1.01961911e+00 6.91233933e-01
-1.23209381e+00 -6.26550734e-01 -5.78449607e-01 5.67411065e-01
5.79953372e-01 4.22718702e-03 1.10447860e+00 3.82511735e-01
-1.12504566e+00 4.25127745e-01 -2.94466496e-01 -6.96370155e-02
7.67277360e-01 -1.12107694e+00 -3.20881873e-01 5.66590726e-02
-6.45615458e-01 4.05111700e-01 8.86735201e-01 -2.03836143e-01
-1.21750844e+00 -6.50838196e-01 1.08822703e+00 6.12801015e-02
7.90760815e-01 -3.96989286e-01 -9.33975518e-01 3.58981043e-01
1.00519097e+00 -3.94765019e-01 1.24568808e+00 4.29932177e-01
2.11573943e-01 1.00963324e-01 -1.15465260e+00 7.53251314e-01
7.16801941e-01 -6.53949380e-01 -5.75643122e-01 1.20178811e-01
4.49989229e-01 -3.27989250e-01 -1.11991608e+00 3.33256751e-01
5.35794079e-01 -1.16607606e+00 5.47979593e-01 -5.80275297e-01
5.29788375e-01 2.41807297e-01 8.69431049e-02 -1.72401845e+00
1.16065433e-02 -6.79501355e-01 -9.41983517e-03 1.76050401e+00
8.52125585e-02 -7.16536522e-01 2.57754356e-01 1.16040540e+00
-2.22828183e-02 -9.19300020e-01 -6.83462977e-01 9.13508162e-02
-1.33973911e-01 -4.03506786e-01 6.00581169e-01 1.24684560e+00
8.62835884e-01 8.59131098e-01 -7.37641037e-01 -2.41404489e-01
-7.57134557e-02 -3.86054702e-02 5.64981461e-01 -1.23863375e+00
-1.31506160e-01 -5.49301267e-01 -9.16677713e-02 -7.16990650e-01
7.50445962e-01 -4.54006225e-01 3.10018301e-01 -1.36319327e+00
1.58809319e-01 -1.13232613e-01 -3.18331838e-01 1.72408611e-01
-4.32816565e-01 -2.49837428e-01 -7.65780583e-02 -3.07469010e-01
-1.03746200e+00 1.40588033e+00 1.02279532e+00 2.64981806e-01
-3.84307712e-01 -3.68167490e-01 -6.56851768e-01 1.02305913e+00
8.33666801e-01 3.30388695e-02 -3.70593697e-01 1.58298299e-01
3.82470012e-01 3.97163272e-01 -3.27183366e-01 -6.17643416e-01
2.90712684e-01 -1.52278483e-01 -2.50091567e-03 -4.02452886e-01
7.56884754e-01 -9.79274690e-01 1.32270694e-01 1.12464838e-01
-6.26610279e-01 -1.61277913e-02 2.39577159e-01 3.93361896e-01
-6.05882883e-01 -1.79541156e-01 8.43424797e-01 2.22790986e-02
-5.78510046e-01 2.38037676e-01 -8.62861633e-01 1.73850894e-01
1.14399147e+00 3.28085423e-01 -9.38005820e-02 -9.19059455e-01
-1.04349208e+00 6.46225452e-01 -1.45423412e-01 7.38399446e-01
5.09250939e-01 -1.31879640e+00 -5.69845140e-01 -1.97303101e-01
8.09928030e-03 -5.80123067e-01 9.92455125e-01 9.98149514e-01
3.95956784e-02 -3.00552100e-02 -4.34869304e-02 -1.08685434e-01
-1.74581051e+00 1.45653382e-01 6.28670394e-01 -4.50655729e-01
9.69491601e-02 9.85753834e-01 1.82131782e-01 -9.28850949e-01
3.78539294e-01 2.82691389e-01 -8.00015748e-01 7.06718147e-01
3.06292027e-01 4.32636380e-01 -3.82450432e-01 -9.14229512e-01
-5.12538731e-01 4.27074850e-01 -2.03350689e-02 -1.55921206e-01
1.19121861e+00 -6.63051426e-01 -5.19212961e-01 1.00569046e+00
1.33569098e+00 7.58699104e-02 -6.42809153e-01 -1.68853641e-01
-8.60012099e-02 8.01311731e-02 2.89066676e-02 -9.53598022e-01
-1.00201356e+00 1.03750324e+00 5.91797352e-01 4.75540429e-01
1.22394753e+00 -1.18095607e-01 9.30289626e-01 3.06240320e-01
-2.32948344e-02 -1.86882889e+00 2.79389650e-01 8.62767696e-01
9.41791296e-01 -1.18406713e+00 -5.05609751e-01 -3.76527816e-01
-1.45142710e+00 1.02593219e+00 1.03590226e+00 3.76786113e-01
9.31575119e-01 2.86980361e-01 4.74419028e-01 -3.00325423e-01
-1.32676578e+00 3.41351107e-02 1.55015290e-01 5.69535419e-02
9.13353741e-01 1.75975502e-01 -4.22975123e-01 1.05710888e+00
-2.64052153e-01 -2.56601453e-01 2.58468390e-01 4.52610165e-01
-2.09294707e-01 -9.66376960e-01 -6.64736629e-02 -9.11073014e-02
-4.82580006e-01 6.51552752e-02 -9.93936658e-01 3.63798320e-01
1.13417879e-01 1.40809810e+00 2.51169860e-01 -4.98868823e-01
4.03209597e-01 5.70236504e-01 -1.21342830e-01 -2.12358743e-01
-1.04910243e+00 1.35137379e-01 6.28485858e-01 -3.78969491e-01
-7.72437453e-01 -5.48937142e-01 -1.61339223e+00 -2.73842454e-01
-2.81267077e-01 8.09135616e-01 7.90399432e-01 9.52680409e-01
6.32638097e-01 7.55991101e-01 1.18530929e+00 -4.46650326e-01
-9.77717191e-02 -1.36934793e+00 -6.57953441e-01 7.18901396e-01
-2.02555344e-01 -4.41971123e-01 -7.54919589e-01 -2.00237259e-01]
|
[13.024286270141602, 6.0434746742248535]
|
af3e4ea6-1372-40e9-ac87-70ae743f3074
|
unsupervised-meta-learning-via-latent-space
| null | null |
https://openreview.net/forum?id=-pLftu7EpXz
|
https://openreview.net/pdf?id=-pLftu7EpXz
|
Unsupervised Meta-Learning via Latent Space Energy-based Model of Symbol Vector Coupling
|
Meta-learning aims to learn a model from a stream of tasks such that the model is able to generalize across tasks and rapidly adapt to new tasks. We propose to learn an energy-based model (EBM) in the latent space of a top-down generative
model such that the EBM in the low dimensional latent space is able to be learned efficiently and adapt to each task rapidly. Furthermore, the energy term couples a continuous latent vector and a symbolic one-hot label. Such coupling formulation allows the model to be learned in an unsupervised manner when the labels are unknown. Our model is learned unsupervisedly in the meta-training phase and evaluated semi-supervisedly in the meta-test phase. We evaluate our model on widely used benchmarks for few-shot meta-learning, Omniglot, and Mini-ImageNet. Our model achieves competitive or superior performance compared to previous state-of-the-art meta-learning models.
|
['Ying Nian Wu', 'Bo Pang', 'Deqian Kong']
|
2021-09-30
| null | null | null |
5th-workshop-on-meta-learning-at-neurips-2021
|
['unsupervised-few-shot-image-classification']
|
['computer-vision']
|
[ 1.54577553e-01 -7.41625205e-02 -3.43661249e-01 -5.45076072e-01
-8.62036288e-01 -2.17233792e-01 8.51597846e-01 5.70722669e-02
-3.88319671e-01 4.87967789e-01 1.53474689e-01 4.23260003e-01
-9.78554264e-02 -6.77857578e-01 -8.22328269e-01 -7.71121323e-01
7.62913236e-03 8.33538353e-01 8.18733275e-02 1.17807947e-01
1.36508569e-01 3.29667777e-02 -1.62642825e+00 4.64871347e-01
8.51277709e-01 5.59687018e-01 5.75309694e-01 5.53561866e-01
-7.58165568e-02 7.15181828e-01 -1.87808827e-01 -7.13059753e-02
-1.75946380e-03 -6.24637306e-01 -9.59574819e-01 2.73378700e-01
1.84876397e-01 3.85205716e-01 -1.78078637e-01 8.67533445e-01
2.80214041e-01 6.77163780e-01 1.11809278e+00 -1.16036117e+00
-5.24954200e-01 3.86442244e-01 -2.60310084e-01 1.18114660e-02
-2.45709956e-01 1.75399184e-01 7.67940342e-01 -9.89821434e-01
8.87249172e-01 1.12456799e+00 3.99833769e-01 8.72996211e-01
-1.54580736e+00 -3.80742073e-01 4.51958776e-01 4.51201260e-01
-9.74072695e-01 -3.62381876e-01 8.60445023e-01 -4.54934061e-01
1.22488463e+00 -2.77121723e-01 8.23775589e-01 1.48065841e+00
5.84689319e-01 8.53436768e-01 1.14719260e+00 -4.60908234e-01
7.41066635e-01 8.85966048e-02 -5.33308126e-02 8.72344553e-01
-1.65233597e-01 -6.10673018e-02 -9.03413951e-01 -2.28249133e-01
2.61998802e-01 2.88175941e-01 1.78397939e-01 -7.05981195e-01
-1.26196754e+00 6.99370801e-01 4.33539808e-01 3.74267787e-01
-5.01766860e-01 4.99994606e-01 4.82867926e-01 2.73615092e-01
9.39974487e-01 4.97067451e-01 -4.55315530e-01 -1.75356418e-01
-1.19164360e+00 4.10868004e-02 6.05844796e-01 7.66890168e-01
1.00167215e+00 -6.74333074e-04 -3.61425459e-01 1.08638370e+00
4.51194435e-01 1.13802917e-01 9.44403231e-01 -9.40159142e-01
4.06571567e-01 5.12332618e-01 -1.84416279e-01 -2.21899495e-01
-2.48111278e-01 -8.02171707e-01 -8.41640294e-01 9.95467156e-02
-3.26622635e-01 7.06310049e-02 -1.37626827e+00 1.99248815e+00
2.63095856e-01 8.13659847e-01 1.80187654e-02 3.19095016e-01
3.27250242e-01 8.67339849e-01 4.09924477e-01 -3.75135630e-01
8.40680599e-01 -1.55515599e+00 -5.52377880e-01 -6.31612897e-01
7.08016813e-01 -3.06258321e-01 1.11661398e+00 3.65065157e-01
-1.20059216e+00 -9.17721093e-01 -9.43979859e-01 8.91080350e-02
-4.60939348e-01 -1.21379629e-01 2.21948326e-01 2.85492271e-01
-9.97405946e-01 7.19826460e-01 -1.30532074e+00 -4.32988733e-01
3.91751528e-01 7.84055144e-02 -1.17575653e-01 -3.21580857e-01
-8.42983603e-01 8.77898693e-01 6.34871364e-01 -2.42168844e-01
-1.73672950e+00 -6.89492941e-01 -7.47657120e-01 1.17927879e-01
1.41731277e-01 -1.09702575e+00 1.24336553e+00 -8.91174316e-01
-1.48083556e+00 8.98518443e-01 -3.55305374e-01 -2.88339227e-01
3.16371202e-01 -1.52427750e-02 -2.85663933e-01 -1.10194817e-01
9.37805101e-02 7.75079191e-01 1.23368728e+00 -1.31740284e+00
-5.49224615e-01 -3.10097754e-01 -2.19640344e-01 4.98217225e-01
-4.83195901e-01 -5.87669492e-01 -4.05728698e-01 -7.52478182e-01
-3.51305865e-02 -1.12562275e+00 -2.70410031e-01 -5.65622091e-01
-1.61280841e-01 -2.89292544e-01 8.12446117e-01 -1.68949023e-01
1.09830403e+00 -2.04115009e+00 8.60527933e-01 -1.59345925e-01
5.19122891e-02 -4.92957700e-03 -3.57123554e-01 5.58239758e-01
-2.05422249e-02 -2.80516017e-02 -2.64576048e-01 -1.02376628e+00
-1.57927889e-02 4.10332203e-01 -6.44142404e-02 1.88518405e-01
-1.62758604e-02 1.14002109e+00 -1.22328568e+00 -3.67991447e-01
8.40770453e-02 3.79651397e-01 -4.65673894e-01 4.66412276e-01
-4.80652660e-01 5.56648016e-01 -3.65372539e-01 5.31994164e-01
1.66501343e-01 -6.33120537e-01 2.59525269e-01 -9.21536908e-02
2.28487611e-01 1.28266200e-01 -7.53580034e-01 2.41087961e+00
-1.01673067e+00 3.13389748e-01 -4.87767994e-01 -1.30518281e+00
7.23377585e-01 3.23305905e-01 4.83534873e-01 -6.47746980e-01
-1.80187091e-01 6.72285035e-02 -3.41633230e-01 -3.78978699e-01
3.17106396e-01 -1.42378971e-01 -2.04847604e-02 7.08291888e-01
8.65703404e-01 1.94373086e-01 1.26519397e-01 1.07015952e-01
9.94166136e-01 3.91427785e-01 7.06762746e-02 -2.21387878e-01
2.47859731e-01 -1.74663439e-01 4.44225907e-01 8.20970714e-01
2.40739360e-01 2.91202307e-01 -1.35937005e-01 -4.86284047e-01
-9.32406485e-01 -1.10578787e+00 1.60965413e-01 1.59935558e+00
-2.23103479e-01 -5.64972937e-01 -7.75253952e-01 -9.32199419e-01
-2.73139596e-01 9.13887560e-01 -8.39301288e-01 -7.68164396e-01
-3.09178114e-01 -8.01386714e-01 -1.05344191e-01 4.46651101e-01
5.15034318e-01 -1.17363989e+00 -6.59178495e-01 4.78682756e-01
-1.99451044e-01 -7.45462358e-01 -3.39587629e-01 6.24842942e-01
-1.36299491e+00 -7.48095870e-01 -6.67593360e-01 -8.85713696e-01
8.64877641e-01 1.04340561e-01 1.18069708e+00 -2.57617176e-01
-2.13504627e-01 5.82921386e-01 -2.50359744e-01 -3.36678959e-02
-5.63373148e-01 4.28212792e-01 1.01747401e-01 3.29897255e-01
1.31795928e-01 -8.07933986e-01 -4.90479350e-01 1.34530082e-01
-8.47142816e-01 4.30505335e-01 4.99441862e-01 1.09348607e+00
1.04059577e+00 -5.86026646e-02 7.67301679e-01 -1.13199329e+00
3.61986667e-01 -8.55889320e-01 -2.82684982e-01 6.37615502e-01
-9.87088919e-01 5.10087311e-01 6.57655954e-01 -7.05468655e-01
-1.29521441e+00 1.97439134e-01 1.95805460e-01 -6.72271192e-01
-8.17515627e-02 6.03527665e-01 -3.08302175e-02 9.09739062e-02
6.63825452e-01 4.97597635e-01 -5.33271730e-01 -7.69594848e-01
6.83985174e-01 3.32024455e-01 4.92868513e-01 -8.41619134e-01
5.86534739e-01 4.97978091e-01 9.13186744e-02 -4.55360442e-01
-1.18044031e+00 -5.11586666e-01 -8.37534845e-01 -3.16494107e-01
7.37094283e-01 -9.35597241e-01 -7.67064318e-02 5.47295272e-01
-7.08992481e-01 -7.58817017e-01 -5.90331376e-01 4.40740496e-01
-1.05178428e+00 -1.20109305e-01 -3.24126542e-01 -5.84535718e-01
-2.87177682e-01 -1.05623806e+00 9.81498659e-01 2.09042624e-01
2.31373142e-02 -1.60554767e+00 7.07133889e-01 1.42183408e-01
4.43354487e-01 1.93091467e-01 1.10891128e+00 -6.14551008e-01
-3.54030281e-01 -2.28692479e-02 3.89226973e-01 3.51532936e-01
2.94148088e-01 -3.34860027e-01 -1.11461067e+00 -5.96724272e-01
5.47819100e-02 -8.04433942e-01 1.38270652e+00 4.92221564e-01
1.26435256e+00 -2.31413767e-01 -6.03085220e-01 8.12295675e-01
1.40418220e+00 -1.42510802e-01 3.33001316e-01 2.96795309e-01
6.11973107e-01 2.71264851e-01 4.87491399e-01 2.29888558e-01
4.87397522e-01 7.63415873e-01 3.07325095e-01 2.90513813e-01
-8.83047283e-02 -5.43168724e-01 4.89689887e-01 1.08886337e+00
-2.66446583e-02 -2.39593834e-01 -8.81724000e-01 6.65671647e-01
-2.23870635e+00 -8.55710149e-01 3.96406472e-01 2.13570356e+00
8.95243049e-01 1.79587618e-01 -1.40229642e-01 -4.09597456e-01
3.19453359e-01 3.53209764e-01 -8.72407675e-01 -3.62103432e-01
2.45879605e-01 3.10580760e-01 -1.10558206e-02 2.98876435e-01
-9.65245724e-01 8.99951518e-01 6.53957176e+00 8.11543286e-01
-9.72112238e-01 8.25421154e-01 7.24717736e-01 -3.18023115e-01
-2.26638198e-01 7.10638165e-02 -7.56554067e-01 5.58914959e-01
1.28121805e+00 -2.63082683e-01 5.13300061e-01 8.75245571e-01
-1.81692943e-01 6.94608018e-02 -1.45498431e+00 8.31566751e-01
1.80913463e-01 -1.47176600e+00 7.00939968e-02 -1.34771159e-02
1.38165247e+00 4.24277693e-01 3.78316551e-01 7.63979673e-01
3.40643942e-01 -8.16320300e-01 5.80405474e-01 9.85505700e-01
6.60002947e-01 -6.77543879e-01 3.19603682e-01 8.65637183e-01
-1.17745030e+00 -8.02332461e-02 -3.84507000e-01 3.62983048e-01
9.70310792e-02 3.22285235e-01 -6.97950602e-01 3.86404663e-01
4.90684241e-01 1.05138183e+00 -6.21052384e-01 8.91002297e-01
-2.75109857e-01 6.73944712e-01 1.29376248e-01 2.84028381e-01
4.21896070e-01 -3.66247669e-02 4.80440348e-01 1.30726206e+00
3.94516110e-01 -2.74186164e-01 4.84705299e-01 8.87266457e-01
-4.03391063e-01 -5.23270965e-02 -4.62195307e-01 -5.42822480e-02
2.77564734e-01 1.36546218e+00 -5.36907613e-01 -5.07857621e-01
-1.66008219e-01 1.25618362e+00 7.25462794e-01 5.67757308e-01
-6.01348937e-01 2.03941450e-01 2.95318156e-01 -1.34497255e-01
1.59081861e-01 -2.48871699e-01 6.53164685e-02 -1.42184734e+00
-2.64458299e-01 -4.06928420e-01 5.38530946e-01 -8.08648109e-01
-1.40289629e+00 5.27795494e-01 2.49918267e-01 -1.13392878e+00
-7.98856735e-01 -1.58634767e-01 -7.38842905e-01 8.63478303e-01
-1.63149250e+00 -1.34876776e+00 -3.96761596e-01 5.92010736e-01
1.03508329e+00 -6.44661844e-01 1.07185960e+00 -1.35982588e-01
-4.29953486e-01 4.20972228e-01 5.22805035e-01 -4.19203341e-01
6.46738768e-01 -1.28960872e+00 5.98423481e-01 5.62577546e-01
5.52837789e-01 4.61055964e-01 3.92567366e-01 -6.18732154e-01
-1.07028317e+00 -1.55141783e+00 7.61019588e-01 -4.91052955e-01
3.34828496e-01 -4.43592519e-01 -1.13138473e+00 7.76223004e-01
1.44573897e-01 2.25751922e-01 7.10879922e-01 2.83447027e-01
-4.46633101e-01 -4.80713136e-02 -9.31391180e-01 1.81300148e-01
1.08340323e+00 -5.59455216e-01 -6.80076063e-01 6.53487027e-01
5.89452684e-01 -2.90575951e-01 -7.57232964e-01 7.52118900e-02
4.12337303e-01 -6.10863388e-01 8.83145630e-01 -1.08862805e+00
3.72851729e-01 1.13747157e-01 -3.32538448e-02 -1.88082898e+00
-6.25379443e-01 -4.61511165e-01 -8.40302825e-01 9.49078619e-01
3.15295368e-01 -3.19581091e-01 6.31349742e-01 1.63914546e-01
-3.29953879e-01 -8.97814870e-01 -9.72628295e-01 -9.74318326e-01
-1.00679569e-01 -3.48086596e-01 1.94540143e-01 9.71459091e-01
-1.59495756e-01 3.74362141e-01 -5.03595293e-01 -1.16846047e-01
9.78713214e-01 1.31063521e-01 4.46974576e-01 -1.38708687e+00
-6.72279358e-01 -1.52003363e-01 -1.12925030e-01 -6.82035506e-01
6.55351102e-01 -1.36589575e+00 2.23588869e-02 -1.61651349e+00
7.19836295e-01 -4.32197213e-01 -9.40975726e-01 7.33279645e-01
-1.00415759e-01 2.83862539e-02 2.35455818e-02 4.32879478e-01
-9.58538353e-01 9.05768692e-01 9.46094334e-01 -3.36578935e-01
-4.22304660e-01 7.42198667e-03 -2.90641069e-01 6.37902021e-01
6.97584867e-01 -1.10606420e+00 -7.11637795e-01 -4.54982549e-01
3.29008698e-01 -1.98593378e-01 3.01311940e-01 -9.93075550e-01
4.01790768e-01 -3.89136910e-01 3.57209206e-01 -2.94770330e-01
5.81064343e-01 -5.41825771e-01 3.27315599e-01 4.81217682e-01
-6.69772029e-01 -2.41474718e-01 -7.13804662e-02 9.41357255e-01
-9.09894854e-02 -4.51911241e-01 8.51768076e-01 -2.84524024e-01
-7.93946445e-01 5.47720730e-01 3.40591231e-03 2.10298717e-01
8.91829133e-01 3.84771079e-02 -1.89928070e-01 -8.54063109e-02
-1.16495121e+00 2.98784465e-01 4.99305278e-01 7.17271328e-01
5.86045980e-01 -1.58710051e+00 -5.88025689e-01 2.26863697e-01
4.42606837e-01 -1.63567677e-01 5.23896337e-01 6.04995072e-01
2.09666729e-01 1.27655432e-01 -3.98791254e-01 -7.19855964e-01
-7.59702504e-01 5.26553214e-01 4.90441412e-01 -7.76985824e-01
-6.60786092e-01 7.54023075e-01 1.79353178e-01 -3.81808788e-01
1.61613315e-01 -1.04959324e-01 -6.83929473e-02 -3.48872989e-02
3.16251725e-01 3.76318842e-01 5.23005147e-03 -4.37767625e-01
-1.40330300e-01 4.77066845e-01 -2.06257433e-01 -3.65865916e-01
1.62208295e+00 -1.04238413e-01 1.14636213e-01 1.16350353e+00
1.40524149e+00 -7.16911852e-01 -1.50323153e+00 -5.96645534e-01
3.35369669e-02 -2.29858562e-01 2.25737840e-01 -8.54846120e-01
-8.41910183e-01 9.78735328e-01 7.34068513e-01 -1.43352062e-01
1.06484532e+00 1.95798337e-01 5.10640979e-01 6.79503918e-01
3.79799426e-01 -1.52146733e+00 7.67739415e-01 4.91652608e-01
6.40041828e-01 -1.13109410e+00 -7.37508461e-02 5.30658603e-01
-7.33880639e-01 1.03210688e+00 4.47272927e-01 -4.34325933e-02
7.92143464e-01 -2.58819695e-04 -5.35211384e-01 -2.39761472e-01
-1.37404561e+00 6.04655668e-02 6.63846076e-01 5.27203560e-01
2.33258545e-01 3.81441824e-02 3.28113645e-01 4.78991449e-01
3.89513612e-01 8.79648477e-02 6.66559041e-02 1.09486294e+00
-6.46629155e-01 -1.30189121e+00 2.45274127e-01 3.72141302e-01
6.03770763e-02 1.54108778e-01 -1.68529134e-02 3.46956551e-01
2.00233221e-01 5.32837629e-01 -8.53465963e-03 -3.07272494e-01
5.88604882e-02 8.00033569e-01 6.01966679e-01 -1.21369064e+00
-4.47554737e-01 1.10158578e-01 -4.25880581e-01 -6.82477117e-01
-6.39637887e-01 -6.29311204e-01 -1.05749881e+00 3.55278254e-01
-3.45964171e-03 2.42549539e-01 8.88088048e-01 1.15710044e+00
4.84300792e-01 8.52705002e-01 8.56610417e-01 -9.10922587e-01
-5.45856416e-01 -1.06148219e+00 -5.10943353e-01 3.02555621e-01
1.43684775e-01 -8.32027376e-01 -1.05361126e-01 4.62621868e-01]
|
[9.863201141357422, 3.018612861633301]
|
d5fbadb1-a56d-4411-92d0-a9b4e175158c
|
differentially-private-topological-data
|
2305.03609
| null |
https://arxiv.org/abs/2305.03609v1
|
https://arxiv.org/pdf/2305.03609v1.pdf
|
Differentially Private Topological Data Analysis
|
This paper is the first to attempt differentially private (DP) topological data analysis (TDA), producing near-optimal private persistence diagrams. We analyze the sensitivity of persistence diagrams in terms of the bottleneck distance, and we show that the commonly used \v{C}ech complex has sensitivity that does not decrease as the sample size $n$ increases. This makes it challenging for the persistence diagrams of \v{C}ech complexes to be privatized. As an alternative, we show that the persistence diagram obtained by the $L^1$-distance to measure (DTM) has sensitivity $O(1/n)$. Based on the sensitivity analysis, we propose using the exponential mechanism whose utility function is defined in terms of the bottleneck distance of the $L^1$-DTM persistence diagrams. We also derive upper and lower bounds of the accuracy of our privacy mechanism; the obtained bounds indicate that the privacy error of our mechanism is near-optimal. We demonstrate the performance of our privatized persistence diagrams through simulations as well as on a real dataset tracking human movement.
|
['Jordan Awan', 'Jinwon Sohn', 'Sehwan Kim', 'Taegyu Kang']
|
2023-05-05
| null | null | null | null |
['topological-data-analysis']
|
['graphs']
|
[ 1.32379681e-01 2.77658731e-01 -6.67936280e-02 5.49463220e-02
-6.37942851e-01 -9.50573564e-01 2.18652084e-01 3.02712440e-01
-6.10037565e-01 8.22543800e-01 -1.90751538e-01 -5.41302204e-01
-4.80852872e-01 -1.01532710e+00 -1.04238355e+00 -9.68410850e-01
-7.22339749e-01 6.62738681e-02 3.86002153e-01 -1.28261939e-01
1.66684493e-01 4.84740376e-01 -1.44120860e+00 -3.68018866e-01
7.96691358e-01 1.01473808e+00 -5.87178528e-01 8.38251710e-01
3.46707821e-01 -1.51196364e-02 -6.03757560e-01 -6.30294025e-01
7.26171315e-01 -3.86867344e-01 -6.15888000e-01 -6.89290583e-01
3.11649382e-01 -2.93313950e-01 -5.47795177e-01 1.45591748e+00
4.08423007e-01 -1.82536021e-01 6.35414362e-01 -1.69481313e+00
-3.81626010e-01 6.00848317e-01 -5.38157105e-01 3.81962210e-01
1.42443553e-01 -1.27841502e-01 9.07811582e-01 -1.55919297e-02
7.44776368e-01 1.08935702e+00 8.49407077e-01 2.50035435e-01
-1.54278815e+00 -9.57387805e-01 -2.34898299e-01 -2.60125160e-01
-1.71563160e+00 -1.78593740e-01 5.91762304e-01 -3.94843757e-01
3.40011537e-01 5.20391643e-01 4.36674446e-01 7.99328268e-01
5.31897962e-01 3.17538828e-01 1.11933196e+00 -3.15943718e-01
3.30859542e-01 9.15728286e-02 3.05925816e-01 9.27191079e-01
9.18152273e-01 1.46939889e-01 -2.10741907e-01 -7.10313678e-01
5.78999102e-01 -1.02543645e-01 -3.06011617e-01 -7.34596133e-01
-9.10546601e-01 7.03723609e-01 2.11999953e-01 1.25693291e-01
2.04073727e-01 5.78176975e-01 2.86729515e-01 6.15122378e-01
2.01600567e-01 2.21322894e-01 -1.09572530e-01 -2.18166918e-01
-5.18640280e-01 3.50758761e-01 8.99536431e-01 1.27741992e+00
5.92203677e-01 -6.99642837e-01 -6.10451214e-02 1.59467489e-01
6.42764289e-03 8.27984035e-01 -1.99409381e-01 -1.09008038e+00
5.76771021e-01 5.60601950e-01 4.46364999e-01 -1.35399711e+00
-2.73563951e-01 -2.18693078e-01 -9.63160634e-01 8.49837717e-03
9.82568264e-01 -1.73461914e-01 -1.48329288e-01 2.30689240e+00
2.08247498e-01 -4.46365178e-01 1.41588137e-01 3.57161105e-01
-1.37149021e-01 4.59308743e-01 -1.56592578e-01 -4.15612072e-01
1.17875838e+00 -8.77858698e-02 -5.37323833e-01 4.61507559e-01
9.34783399e-01 -8.88369605e-02 9.96456504e-01 5.07527106e-02
-8.85816336e-01 1.34164244e-01 -1.19672740e+00 1.83933109e-01
-4.28987712e-01 -2.02014878e-01 2.66834289e-01 1.22205722e+00
-9.83931482e-01 5.22996128e-01 -7.03102767e-01 -6.66204512e-01
5.04040718e-01 5.52714288e-01 -2.62220740e-01 3.04704428e-01
-1.22583175e+00 2.98376441e-01 6.38061538e-02 -2.24360913e-01
-4.80490297e-01 -4.99364227e-01 -3.44677120e-01 5.07366285e-02
2.41441146e-01 -4.23042357e-01 5.04598320e-01 -2.18739837e-01
-1.01220465e+00 7.89069831e-01 1.83023825e-01 -7.87546515e-01
8.37787151e-01 2.17932552e-01 -1.85806260e-01 4.22364086e-01
2.46317126e-02 2.78938711e-01 5.09761393e-01 -1.05405736e+00
-6.76864982e-01 -7.12478578e-01 2.10484326e-01 -3.34316730e-01
-4.60880816e-01 -2.87781149e-01 1.91155002e-01 -3.87193292e-01
-1.29156023e-01 -1.18572640e+00 4.50106664e-03 4.11352903e-01
-2.62524158e-01 1.38852045e-01 1.00276530e+00 -3.11760426e-01
1.54083109e+00 -2.42524529e+00 -7.20767900e-02 7.90167511e-01
4.63172883e-01 3.68821844e-02 1.85663700e-01 5.21881998e-01
4.71288443e-01 5.70333362e-01 -3.79870653e-01 9.20988470e-02
2.24665955e-01 1.45309756e-03 -2.80305773e-01 1.08464241e+00
-6.09174192e-01 3.67091149e-01 -8.37342560e-01 -4.96756166e-01
-2.71219015e-01 1.52660027e-01 -5.63078523e-01 -2.77714133e-01
4.87090237e-02 1.60473496e-01 -4.57172245e-01 3.64063323e-01
1.05554366e+00 -3.62467438e-01 3.81598711e-01 7.32275620e-02
-1.52699932e-01 -2.88041562e-01 -9.67987776e-01 1.11272216e+00
1.10480376e-01 5.83567679e-01 1.60768643e-01 -8.11797261e-01
9.41777945e-01 -6.21768348e-02 6.38475418e-01 -3.78938228e-01
2.09967464e-01 2.50278324e-01 -9.09360796e-02 -6.51929677e-02
3.76662999e-01 -4.00815345e-02 -5.41153371e-01 6.44479454e-01
-4.39809173e-01 5.95878065e-01 -2.89472818e-01 2.41509810e-01
1.53383088e+00 -5.65151393e-01 1.70874238e-01 -8.68236840e-01
4.92755562e-01 -2.55650610e-01 5.55855870e-01 9.44892108e-01
-7.21551836e-01 6.77010566e-02 1.17412221e+00 -7.52984956e-02
-1.18272936e+00 -1.25355124e+00 -3.07140172e-01 6.43699765e-01
5.96440434e-01 -5.30076206e-01 -1.05684006e+00 -6.41927600e-01
4.56106722e-01 2.83100545e-01 -8.80508244e-01 -2.41623446e-01
-3.95111918e-01 -6.86987996e-01 1.12376547e+00 1.25866532e-01
8.83920908e-01 -1.68663234e-01 -8.81600499e-01 -1.90351427e-01
-9.64889899e-02 -8.14601541e-01 -5.63216269e-01 -1.57533169e-01
-6.94895267e-01 -1.15884495e+00 -6.33675694e-01 -3.39562744e-01
7.25011468e-01 2.11593553e-01 1.52814791e-01 -3.49002406e-02
1.05134081e-02 5.79850197e-01 -1.71869293e-01 -9.17506739e-02
-4.01609182e-01 1.83672041e-01 2.40385145e-01 6.95370957e-02
1.87360317e-01 -7.50863552e-01 -9.23448980e-01 5.24192154e-01
-8.36924911e-01 -5.37900150e-01 1.54907018e-01 5.52115619e-01
5.73656499e-01 2.92511553e-01 3.50257635e-01 -6.63921535e-01
6.87816083e-01 -4.30770457e-01 -1.16361451e+00 4.10264581e-01
-8.89090419e-01 3.01942587e-01 6.06897712e-01 -4.16440070e-01
-3.84021252e-01 6.29253499e-03 4.23361599e-01 -2.94126540e-01
4.44937319e-01 5.00970520e-02 -2.46637598e-01 -4.67459559e-01
7.06540227e-01 7.90059045e-02 3.26924533e-01 -1.86226517e-01
3.18323582e-01 6.83025658e-01 5.20103574e-01 -7.37558782e-01
6.32744551e-01 7.87049592e-01 6.80463672e-01 -1.00593448e+00
-9.76331234e-02 1.70497131e-02 -3.59094381e-01 -4.66952473e-02
5.86962044e-01 -3.66586000e-01 -1.65627217e+00 3.42784345e-01
-6.91787779e-01 3.71423103e-02 -3.06521386e-01 1.51518941e-01
-8.91971588e-01 6.79447353e-01 -4.16317910e-01 -1.17126024e+00
-2.71611929e-01 -9.76525187e-01 6.63253665e-01 -8.77137706e-02
-2.46636700e-02 -6.05222285e-01 8.54294747e-02 -1.53672010e-01
2.94777542e-01 8.92870069e-01 1.19558620e+00 -4.40153986e-01
-7.49245346e-01 -4.88272905e-01 -2.42867157e-01 5.19749476e-03
-4.88768779e-02 -8.68401304e-02 -6.02019548e-01 -6.92877769e-01
-1.41392082e-01 2.13013217e-01 6.43229604e-01 4.01345551e-01
1.18831778e+00 -8.72743368e-01 -6.61243021e-01 6.31201625e-01
1.37120724e+00 9.89235342e-02 7.79317021e-01 2.34568045e-01
1.52681142e-01 4.87936407e-01 5.42043030e-01 6.03128135e-01
2.72626579e-01 5.15377045e-01 1.50724798e-01 4.09517288e-01
4.48074937e-01 -5.34218907e-01 3.57201695e-01 3.29639763e-01
1.36735722e-01 -2.84407079e-01 -7.09223807e-01 3.97909671e-01
-1.78834939e+00 -9.93454337e-01 8.58673826e-03 2.73329568e+00
5.86072206e-01 1.60429165e-01 6.26198292e-01 1.17389247e-01
8.13848972e-01 5.64370528e-02 -4.89457726e-01 -4.44302559e-01
-2.12648839e-01 9.35573578e-02 1.30446863e+00 5.19429266e-01
-9.22604084e-01 4.62145358e-01 6.82702923e+00 7.16111362e-01
-5.66577911e-01 1.33781284e-01 4.46623147e-01 4.56229858e-02
-3.32747370e-01 2.13158101e-01 -5.41490674e-01 8.70123863e-01
1.00860953e+00 -7.93629646e-01 3.45478952e-01 7.53815532e-01
-2.51893312e-01 -2.08681613e-01 -1.08768249e+00 9.81206834e-01
-5.11230350e-01 -1.27900743e+00 -1.37601420e-01 8.99671316e-01
2.92795807e-01 -5.47589242e-01 2.99277961e-01 -2.12748259e-01
3.72133225e-01 -6.15030408e-01 2.99278855e-01 3.84988427e-01
1.05406737e+00 -1.16497219e+00 4.54605848e-01 3.69808495e-01
-1.29032147e+00 -4.03712004e-01 -2.92043358e-01 3.04417163e-01
-2.00196400e-01 3.32672387e-01 -4.13650960e-01 3.93353432e-01
7.68613636e-01 1.20217301e-01 -3.88273418e-01 7.27946758e-01
5.14080882e-01 2.73418903e-01 -7.87098467e-01 -2.26721197e-01
-2.07327083e-01 -3.17404777e-01 8.79769504e-01 9.51091290e-01
5.49689054e-01 4.41168040e-01 -3.17504376e-01 7.73458362e-01
-3.76630276e-01 -5.53280078e-02 -9.10347342e-01 -2.80919194e-01
9.39732909e-01 4.36503351e-01 -6.52092993e-01 7.93972239e-02
2.34133795e-01 7.86333501e-01 4.93880250e-02 1.18277617e-01
-6.27379060e-01 -8.64596248e-01 9.55534041e-01 3.27391356e-01
4.42491651e-01 -3.42732161e-01 -1.59280166e-01 -9.69921291e-01
2.26419643e-01 -4.86449152e-01 6.77322030e-01 1.62962690e-01
-1.17185652e+00 1.85590431e-01 2.50482023e-01 -1.24658430e+00
1.24773912e-01 -1.90999120e-01 -1.11667104e-02 4.99625951e-01
-7.85087407e-01 -6.08507156e-01 1.70357451e-01 6.94938362e-01
-7.05543756e-01 5.64648211e-02 7.50381529e-01 2.42298961e-01
-3.59397918e-01 1.32635462e+00 6.24657214e-01 -5.17122857e-02
2.82851785e-01 -8.67107451e-01 3.20855796e-01 8.89604747e-01
-5.81592321e-01 6.03787959e-01 9.31037128e-01 -7.83686697e-01
-1.67458856e+00 -7.28244543e-01 6.22146964e-01 -3.84919256e-01
5.88100076e-01 -7.24457979e-01 -5.76572359e-01 6.30847156e-01
-4.74034041e-01 6.00165389e-02 6.56605780e-01 -1.10101201e-01
-4.63399023e-01 -4.68059629e-01 -1.95977855e+00 5.72462976e-01
1.43298721e+00 -5.59370100e-01 8.64471644e-02 2.40030680e-02
8.08454692e-01 -5.28688356e-03 -1.16386664e+00 2.52689928e-01
9.69651580e-01 -8.97611856e-01 8.39762151e-01 -3.20061207e-01
-2.48969331e-01 -2.06244573e-01 -6.37081742e-01 -5.62449515e-01
-1.17947288e-01 -1.10557520e+00 -2.25364923e-01 1.01264536e+00
3.22004020e-01 -1.18106353e+00 8.07698667e-01 6.33930147e-01
8.03702533e-01 -5.68867981e-01 -1.49417639e+00 -1.15153265e+00
3.80535245e-01 7.08293989e-02 7.18405187e-01 7.55965412e-01
2.38920674e-01 -3.19092184e-01 -3.80945921e-01 3.75134557e-01
1.10117567e+00 -8.32627565e-02 9.83605981e-01 -1.32611704e+00
-2.57558316e-01 -2.82739520e-01 -7.69954979e-01 -9.88864660e-01
-1.17398292e-01 -6.75453842e-01 -3.63450557e-01 -7.05100238e-01
1.51742026e-01 -8.16635251e-01 -2.15468735e-01 2.11275592e-02
4.79151517e-01 -3.31670433e-01 8.64031017e-02 2.77407169e-01
-3.21468383e-01 5.08852482e-01 9.91704106e-01 9.59198922e-02
-2.79492050e-01 2.62538549e-02 -7.06138074e-01 2.87999660e-01
6.93286836e-01 -6.25541925e-01 -3.32983524e-01 2.30332091e-01
1.75613910e-01 2.85028785e-01 4.21652883e-01 -9.92777288e-01
8.92122015e-02 -1.08087048e-01 -6.73703998e-02 -6.56354368e-01
2.62311578e-01 -8.58253300e-01 4.88444120e-01 8.02530169e-01
-4.22562748e-01 1.29208922e-01 -5.36900610e-02 1.13661993e+00
5.35181046e-01 2.50850648e-01 8.58863056e-01 2.68957585e-01
1.69924021e-01 3.16957086e-01 -2.74666220e-01 -6.21748855e-03
1.32655084e+00 -4.28121239e-01 -8.33063960e-01 -5.33831894e-01
-4.93520319e-01 2.72545844e-01 9.50073123e-01 -1.69560641e-01
2.61374593e-01 -1.25057757e+00 -8.16034675e-02 1.89793169e-01
2.59170681e-01 -5.60765982e-01 2.39201471e-01 8.70134532e-01
-6.61708653e-01 3.91951799e-01 -4.00888503e-01 -5.64669251e-01
-1.32464826e+00 8.69044542e-01 3.67788434e-01 -6.14886731e-02
-7.69606948e-01 4.52622622e-01 -4.77163680e-02 2.52350583e-03
4.38216209e-01 -3.65859360e-01 7.02972710e-01 -3.35206091e-01
4.21864629e-01 7.63336062e-01 -4.17126507e-01 -5.09478092e-01
-4.78379846e-01 5.61113000e-01 1.22849401e-02 -4.20720339e-01
1.00617647e+00 -3.80019486e-01 -1.46338403e-01 -1.40875960e-02
1.55491614e+00 8.50891694e-02 -1.18855417e+00 1.86666735e-02
-4.68365699e-02 -8.25334787e-01 -5.00024796e-01 -1.01957142e-01
-7.11661041e-01 6.55857563e-01 9.91405904e-01 6.73114896e-01
1.00622880e+00 3.71049307e-02 7.58289039e-01 3.95182073e-01
8.59773219e-01 -1.02972090e+00 -3.71747524e-01 4.12404686e-02
4.66322750e-01 -5.18782198e-01 1.45633379e-02 -5.53376436e-01
-3.60979885e-02 6.78853869e-01 6.44945875e-02 -1.33613601e-01
1.07646275e+00 1.37153089e-01 -5.49183011e-01 -1.59969464e-01
-4.54142153e-01 2.84392029e-01 -3.59557539e-01 6.58990920e-01
-2.97281891e-01 3.82311046e-01 -8.98207903e-01 5.48291624e-01
-4.01666343e-01 -7.82887712e-02 6.06720030e-01 1.13804758e+00
-4.57602978e-01 -1.26740193e+00 -2.53414243e-01 3.76278013e-01
-3.07505131e-01 4.85171080e-01 -5.38780928e-01 9.74718213e-01
-2.02074558e-01 7.04454541e-01 -9.54463556e-02 -6.10062659e-01
1.58295512e-01 -7.19109252e-02 6.35342121e-01 3.07077408e-01
-1.94771260e-01 -4.43497658e-01 -6.48805648e-02 -4.67597216e-01
-1.13600388e-01 -6.55190110e-01 -1.06219125e+00 -1.25184262e+00
-6.78930059e-02 2.81768858e-01 3.46672863e-01 4.11189556e-01
7.41099477e-01 -4.85013276e-01 9.12448108e-01 7.39080682e-02
-7.07546294e-01 -3.88411105e-01 -1.00090456e+00 1.65380999e-01
6.01559162e-01 -7.71882355e-01 -6.87024891e-01 -4.11551207e-01]
|
[6.08095645904541, 6.639958381652832]
|
13ed9fd7-b43e-4b0b-b349-aa8081289b56
|
instance-aware-domain-generalization-for-face
|
2304.05640
| null |
https://arxiv.org/abs/2304.05640v1
|
https://arxiv.org/pdf/2304.05640v1.pdf
|
Instance-Aware Domain Generalization for Face Anti-Spoofing
|
Face anti-spoofing (FAS) based on domain generalization (DG) has been recently studied to improve the generalization on unseen scenarios. Previous methods typically rely on domain labels to align the distribution of each domain for learning domain-invariant representations. However, artificial domain labels are coarse-grained and subjective, which cannot reflect real domain distributions accurately. Besides, such domain-aware methods focus on domain-level alignment, which is not fine-grained enough to ensure that learned representations are insensitive to domain styles. To address these issues, we propose a novel perspective for DG FAS that aligns features on the instance level without the need for domain labels. Specifically, Instance-Aware Domain Generalization framework is proposed to learn the generalizable feature by weakening the features' sensitivity to instance-specific styles. Concretely, we propose Asymmetric Instance Adaptive Whitening to adaptively eliminate the style-sensitive feature correlation, boosting the generalization. Moreover, Dynamic Kernel Generator and Categorical Style Assembly are proposed to first extract the instance-specific features and then generate the style-diversified features with large style shifts, respectively, further facilitating the learning of style-insensitive features. Extensive experiments and analysis demonstrate the superiority of our method over state-of-the-art competitors. Code will be publicly available at https://github.com/qianyuzqy/IADG.
|
['Lizhuang Ma', 'Shouhong Ding', 'Ran Yi', 'Xuequan Lu', 'Taiping Yao', 'Ke-Yue Zhang', 'Qianyu Zhou']
|
2023-04-12
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Zhou_Instance-Aware_Domain_Generalization_for_Face_Anti-Spoofing_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Zhou_Instance-Aware_Domain_Generalization_for_Face_Anti-Spoofing_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['face-anti-spoofing']
|
['computer-vision']
|
[ 3.49755585e-01 -3.98465067e-01 -2.71970004e-01 -6.44100964e-01
-4.23476905e-01 -8.05727124e-01 5.86190701e-01 -2.06667051e-01
-8.62627402e-02 6.50669634e-01 1.63234055e-01 2.87236534e-02
-1.67483002e-01 -8.29730451e-01 -3.67501050e-01 -9.97763634e-01
2.29335502e-01 1.46351889e-01 1.82362229e-01 -4.12990332e-01
1.65324181e-01 4.50123191e-01 -1.41973829e+00 3.47705603e-01
1.15763891e+00 1.00318432e+00 -7.55680054e-02 4.23204601e-02
-2.96009153e-01 5.23458272e-02 -7.70688355e-01 -3.73989016e-01
3.62401038e-01 -5.12573600e-01 -3.13198090e-01 9.00788531e-02
4.05832231e-01 -2.57648170e-01 -3.35144371e-01 1.29406643e+00
4.41548854e-01 3.40924077e-02 7.52233684e-01 -1.37476826e+00
-1.05596817e+00 1.91505373e-01 -7.51171172e-01 2.28965953e-01
1.60751566e-01 3.29903007e-01 7.99007058e-01 -7.82144666e-01
4.47303325e-01 1.48248422e+00 2.45479599e-01 8.56637537e-01
-1.20350230e+00 -1.29921246e+00 4.55983639e-01 2.66389042e-01
-1.48278606e+00 -4.24244076e-01 1.52355301e+00 -2.34406173e-01
1.64802253e-01 1.01259626e-01 2.36272871e-01 1.61766040e+00
-3.10017437e-01 6.68518007e-01 1.39466763e+00 -2.29175776e-01
2.88468957e-01 4.92243469e-01 6.24852516e-02 4.55543548e-01
4.50568467e-01 2.48432606e-01 -5.54238439e-01 -2.68160760e-01
7.71927178e-01 1.66605294e-01 -1.89122856e-01 -5.81107557e-01
-9.34224606e-01 8.04605126e-01 4.20550555e-01 3.11497778e-01
-1.16500989e-01 -5.33353925e-01 6.55168831e-01 4.02276039e-01
5.33824027e-01 2.51409382e-01 -6.94500506e-01 1.59589291e-01
-6.73010170e-01 2.17668295e-01 4.05663460e-01 1.07927251e+00
1.08228004e+00 1.38392583e-01 -2.72050083e-01 9.98905540e-01
6.47946224e-02 6.75573826e-01 7.00532794e-01 -4.85553831e-01
4.78252321e-01 8.26553404e-01 -1.56402826e-01 -1.21638620e+00
-1.20846473e-01 -5.75140595e-01 -9.89924550e-01 9.37705953e-03
4.58718538e-01 -1.47513241e-01 -9.28666234e-01 2.12423921e+00
5.02620459e-01 4.61181611e-01 -4.44148667e-02 7.78346121e-01
6.68888569e-01 4.92512077e-01 2.40510851e-01 -2.67116696e-01
1.53090954e+00 -5.34351885e-01 -5.03289521e-01 -1.81118622e-01
4.86399680e-01 -6.66654646e-01 1.43235004e+00 2.96554297e-01
-4.21203196e-01 -7.54019499e-01 -1.09942698e+00 2.57191032e-01
-4.93726492e-01 1.05076663e-01 4.77129549e-01 1.10138953e+00
-4.54470634e-01 3.18046331e-01 -3.87878537e-01 -3.12545180e-01
7.04326987e-01 3.83155763e-01 -3.87690037e-01 1.06715644e-02
-1.43959427e+00 3.08643550e-01 6.03355408e-01 -1.24848098e-01
-8.50054979e-01 -6.77937567e-01 -8.16576481e-01 2.28262711e-02
3.15006167e-01 -4.61349487e-01 8.23649824e-01 -1.31783402e+00
-1.68135977e+00 8.69100273e-01 -2.09730625e-01 -2.97307167e-02
3.89755517e-01 1.81326270e-02 -9.13883507e-01 1.63967699e-01
8.44124481e-02 4.32741284e-01 1.39625394e+00 -1.35085559e+00
-5.19873083e-01 -5.33529103e-01 1.63348857e-02 8.70187357e-02
-9.85310495e-01 -1.26889616e-01 -8.88145268e-02 -9.43195105e-01
4.38159928e-02 -8.34304571e-01 1.05735213e-01 -1.35745436e-01
-1.98432788e-01 -2.50561029e-01 1.05494404e+00 -5.28469265e-01
1.21766663e+00 -2.43391395e+00 -1.13952793e-01 3.24900448e-01
1.19620875e-01 6.15223467e-01 -4.03338790e-01 1.21649675e-01
-1.80764675e-01 1.10421620e-01 -3.42822552e-01 5.12716249e-02
-7.63656050e-02 5.00205308e-02 -4.80455190e-01 4.57621723e-01
6.09702885e-01 4.59962219e-01 -9.69554484e-01 -4.80787188e-01
1.19381890e-01 4.01176274e-01 -7.37044692e-01 1.67740643e-01
-1.85724854e-01 8.44138682e-01 -7.33411074e-01 7.12035358e-01
1.28112161e+00 -5.46886064e-02 2.19643816e-01 -3.07656825e-01
2.98859447e-01 1.00142017e-01 -1.37349486e+00 1.50034523e+00
-5.24688303e-01 -3.32834460e-02 2.92231347e-02 -1.07286382e+00
1.36326909e+00 3.88477333e-02 1.52061969e-01 -6.53335631e-01
8.16681329e-03 2.70372421e-01 -8.88670981e-02 -2.66054273e-01
5.05526438e-02 -2.79937148e-01 -2.35511869e-01 2.42760837e-01
5.23850881e-02 1.51617065e-01 -1.47391781e-01 1.13562020e-02
6.62163019e-01 1.29327700e-01 3.46814215e-01 -3.63306850e-01
1.12921500e+00 -3.36538166e-01 1.03963315e+00 4.50083047e-01
-4.87934321e-01 4.81180400e-01 3.80746514e-01 -3.48761976e-01
-6.83755398e-01 -1.21818590e+00 -2.56577045e-01 1.38015366e+00
6.45388305e-01 -2.43267462e-01 -7.99249530e-01 -1.16667175e+00
2.15691719e-02 4.75588858e-01 -6.70814693e-01 -4.72912073e-01
-6.27063811e-01 -7.67737627e-01 4.89425927e-01 3.80629361e-01
7.96740830e-01 -9.77408826e-01 5.25921583e-03 -9.00199413e-02
2.01738570e-02 -8.56721401e-01 -5.46162486e-01 -2.04832464e-01
-7.01365530e-01 -8.58416677e-01 -7.21165657e-01 -7.23342836e-01
7.70249069e-01 3.14945281e-01 7.43859649e-01 1.14939928e-01
5.66572435e-02 -6.13745116e-02 -4.68158424e-01 -1.93760186e-01
-1.47983238e-01 2.90358186e-01 3.94814312e-01 5.78389823e-01
7.65295029e-01 -8.72340560e-01 -8.46099854e-01 5.17937660e-01
-1.13509810e+00 -3.39706808e-01 6.84149742e-01 1.09500229e+00
2.56354630e-01 1.26885802e-01 9.90090370e-01 -1.10836434e+00
7.32061207e-01 -5.65434098e-01 -3.19713891e-01 2.60263890e-01
-4.71195638e-01 1.26865268e-01 1.03152275e+00 -7.59497464e-01
-1.42297339e+00 -6.64776936e-02 5.66296093e-02 -4.43129987e-01
-6.17093742e-01 -1.27744973e-01 -8.53814840e-01 -1.00136846e-01
8.32742095e-01 4.60551322e-01 -1.15133487e-01 -3.92387986e-01
4.27595437e-01 7.18559206e-01 2.76267916e-01 -1.01126289e+00
1.24682105e+00 6.11359537e-01 -1.93952367e-01 -7.15663671e-01
-5.80120027e-01 -3.37130278e-01 -6.91398561e-01 2.03193039e-01
4.22535598e-01 -8.50930572e-01 -3.61890554e-01 4.79462653e-01
-9.66424823e-01 7.11965337e-02 -7.04188943e-02 1.88253447e-01
-3.15652728e-01 6.09676659e-01 -3.39100599e-01 -4.58325744e-01
-1.97430775e-01 -1.08415115e+00 1.08558428e+00 5.44809282e-01
-8.42841491e-02 -8.62615824e-01 -2.38678772e-02 1.18769266e-01
3.91793877e-01 1.83455110e-01 9.18647647e-01 -1.02477789e+00
-2.67405987e-01 1.60268489e-02 -5.05172133e-01 4.55166876e-01
5.71520627e-01 -1.91681221e-01 -1.23651481e+00 -5.03462255e-01
8.69856775e-02 -2.40805507e-01 7.49584138e-01 -2.36214884e-03
1.48437476e+00 -4.53751177e-01 -4.13229167e-01 8.10008705e-01
1.17383778e+00 1.10787287e-01 3.21663499e-01 3.08583528e-01
7.15529680e-01 6.74121797e-01 6.60697222e-01 5.90758324e-01
1.14760280e-01 7.50620127e-01 6.90547824e-02 9.82777681e-04
-1.18746027e-01 -5.46831250e-01 2.48801410e-01 3.30392182e-01
1.48489326e-01 7.32228309e-02 -4.62895185e-01 4.56709534e-01
-1.38480043e+00 -8.77263665e-01 4.27916586e-01 2.08006239e+00
9.67598677e-01 1.80810124e-01 3.10072809e-01 1.10386483e-01
1.07235157e+00 3.45295608e-01 -7.51945794e-01 -2.82882899e-01
-1.39117926e-01 3.26234460e-01 2.24808723e-01 1.82232171e-01
-1.17229140e+00 1.15919340e+00 4.32747602e+00 1.40226638e+00
-1.23310292e+00 1.29647210e-01 5.81410170e-01 8.39580595e-02
-5.03383934e-01 -1.31614715e-01 -9.20612872e-01 8.22829962e-01
3.82171631e-01 -2.57077813e-01 3.57760370e-01 1.07860696e+00
6.68439781e-03 6.90560043e-01 -7.48962224e-01 1.03124189e+00
-7.96451271e-02 -9.53367829e-01 3.23290497e-01 8.42472166e-02
7.89935172e-01 -6.68392479e-01 4.25854832e-01 4.53277051e-01
1.87933713e-01 -6.75484002e-01 3.26078057e-01 5.36453649e-02
1.01863968e+00 -9.84464467e-01 5.03096581e-01 9.96509790e-02
-1.15962076e+00 -2.72795588e-01 -5.95757961e-01 1.31171614e-01
-2.19612673e-01 6.21790707e-01 -7.19679713e-01 5.66495359e-01
6.51727974e-01 6.01380289e-01 -7.72733927e-01 5.48993051e-01
-2.88787305e-01 6.75707638e-01 -1.66058615e-01 1.37368321e-01
5.07356450e-02 -9.77972522e-02 5.35907269e-01 1.24399889e+00
2.57092744e-01 1.91056401e-01 2.41779804e-01 7.24863589e-01
-1.28259659e-01 1.94570795e-01 -6.92958891e-01 6.48036674e-02
8.59260559e-01 1.29210413e+00 -4.24338281e-01 -2.31648967e-01
-3.64889920e-01 1.06399179e+00 2.02805743e-01 4.67463851e-01
-7.24969566e-01 -5.11319697e-01 1.09623718e+00 1.42156601e-01
2.65110403e-01 -5.44633903e-02 -4.28985626e-01 -1.30456197e+00
8.38935897e-02 -1.22854900e+00 6.53665662e-01 -2.59577613e-02
-1.78723764e+00 6.80451095e-01 -6.77732602e-02 -1.50388026e+00
-7.63951913e-02 -6.39674485e-01 -6.49715245e-01 8.37457716e-01
-1.65616119e+00 -1.39367807e+00 -4.09114957e-01 9.19643998e-01
4.00516748e-01 -4.10803407e-01 6.65952384e-01 2.17244968e-01
-6.41749084e-01 1.25995612e+00 5.51701151e-02 2.28304669e-01
1.08097446e+00 -7.96666980e-01 1.94684327e-01 8.06008339e-01
-8.42450112e-02 1.05706966e+00 5.25139570e-01 -6.63357198e-01
-1.07604253e+00 -1.09686899e+00 3.86243105e-01 -3.66083533e-01
5.61313033e-01 -5.03185809e-01 -1.23752248e+00 3.22457194e-01
-9.30625796e-02 1.85980111e-01 7.45242119e-01 1.40287131e-01
-9.69984472e-01 -6.28576398e-01 -1.53686106e+00 5.62146783e-01
1.29343128e+00 -6.70184255e-01 -5.35155237e-01 2.19549075e-01
6.34630382e-01 -7.87821710e-02 -5.62872469e-01 7.01186180e-01
4.77517366e-01 -1.07301080e+00 9.32426274e-01 -6.07219696e-01
8.96556154e-02 -5.19078135e-01 -1.19386531e-01 -1.24265158e+00
-5.04512548e-01 -4.60317999e-01 1.30773187e-01 1.69387054e+00
-6.48901938e-03 -9.95249152e-01 9.08006728e-01 1.83127284e-01
2.12899730e-01 -5.01829267e-01 -8.68571937e-01 -9.93868291e-01
1.70982093e-01 1.07811563e-01 1.20747197e+00 1.17936695e+00
-2.32249707e-01 1.54899076e-01 -3.46627235e-01 4.90882844e-01
5.95212758e-01 2.82512814e-01 8.31137002e-01 -1.29151332e+00
-2.38119662e-01 -5.56144118e-01 -4.68422115e-01 -8.86153281e-01
5.17386436e-01 -8.07204127e-01 -3.86672705e-01 -5.90058923e-01
3.31444442e-02 -6.87069654e-01 -6.92828178e-01 2.74603516e-01
-3.87436956e-01 2.10057601e-01 4.63233627e-02 1.57150462e-01
-3.63141239e-01 6.96900070e-01 1.28574872e+00 -1.95775390e-01
-2.11451367e-01 -1.43066689e-03 -1.02774739e+00 5.28722405e-01
1.02978516e+00 -4.27761346e-01 -6.79481804e-01 -2.26219431e-01
-3.90979081e-01 -4.45852607e-01 3.02518338e-01 -1.07372117e+00
-1.55400574e-01 -4.81523871e-01 6.65701509e-01 -1.39824316e-01
1.88559875e-01 -7.67320931e-01 -1.98797345e-01 3.28398883e-01
-2.21646741e-01 -2.82936245e-01 1.73271626e-01 7.44958758e-01
-2.15165362e-01 7.47654289e-02 1.08480692e+00 -7.17975050e-02
-9.41230834e-01 6.19313180e-01 2.20897824e-01 2.03844979e-01
9.49096441e-01 -2.22855806e-01 -4.04263765e-01 -9.80914459e-02
-2.81835824e-01 -7.24208727e-02 7.01148212e-01 6.42814517e-01
4.91731048e-01 -1.43844485e+00 -7.96556473e-01 7.82840967e-01
4.68146265e-01 -1.86921149e-01 6.11601293e-01 1.84267208e-01
-8.92161578e-02 2.14011133e-01 -4.25726712e-01 -6.06534541e-01
-1.09771264e+00 8.44078958e-01 6.57514408e-02 -1.78531930e-01
-2.18074560e-01 9.79367256e-01 8.95431042e-01 -6.32089496e-01
-1.64526686e-01 1.17305905e-01 -1.78404763e-01 5.47068231e-02
7.20999360e-01 1.71893626e-01 -1.10051949e-02 -5.69139063e-01
-6.27307355e-01 7.32117891e-01 -3.53865594e-01 1.76776394e-01
9.92593825e-01 -1.73589185e-01 3.52334753e-02 -1.40667588e-01
1.20078325e+00 1.96666434e-01 -1.43046570e+00 -5.00605345e-01
9.55123175e-03 -7.64245033e-01 -4.17508453e-01 -6.70545459e-01
-1.04593837e+00 1.06667316e+00 8.04802477e-01 7.91100860e-02
1.57706261e+00 -1.78144768e-01 8.29994321e-01 1.32250180e-03
2.98911572e-01 -9.47414160e-01 2.73594946e-01 2.14905679e-01
6.83438599e-01 -1.23089361e+00 -2.13726491e-01 -5.36303699e-01
-7.07296371e-01 9.94428813e-01 1.00626993e+00 -1.38214767e-01
5.73170364e-01 -1.19120970e-01 1.47464961e-01 1.51377544e-01
-2.37750828e-01 -1.07457824e-01 3.31828862e-01 1.08073449e+00
2.00792119e-01 2.25180075e-01 -2.94806957e-01 7.97889769e-01
-9.64138955e-02 -2.42911875e-01 2.34779043e-04 5.68408430e-01
-2.89892137e-01 -1.46124768e+00 -4.33426619e-01 1.55668944e-01
-1.99193522e-01 -3.55274342e-02 -4.16640341e-01 5.53174198e-01
4.07206863e-01 8.88225555e-01 -1.35207981e-01 -5.95830142e-01
3.17260772e-01 1.02671258e-01 3.43101203e-01 -5.74740827e-01
-1.98726535e-01 -1.53290540e-01 -2.24197373e-01 -2.85826057e-01
-1.06017999e-01 -5.41414499e-01 -8.99334610e-01 -3.08080137e-01
-1.47396520e-01 9.62224752e-02 1.71057060e-01 8.22220802e-01
6.12382114e-01 6.58740848e-02 1.16810668e+00 -5.87761164e-01
-7.49991417e-01 -8.70539367e-01 -5.67960918e-01 8.36559415e-01
3.78236502e-01 -9.96388674e-01 -4.44253892e-01 -1.36952281e-01]
|
[13.220630645751953, 1.1286412477493286]
|
f4a24006-a24d-4bdf-af62-9124e7f4b4e0
|
a-deeper-look-at-3d-shape-classifiers
|
1809.02560
| null |
http://arxiv.org/abs/1809.02560v2
|
http://arxiv.org/pdf/1809.02560v2.pdf
|
A Deeper Look at 3D Shape Classifiers
|
We investigate the role of representations and architectures for classifying
3D shapes in terms of their computational efficiency, generalization, and
robustness to adversarial transformations. By varying the number of training
examples and employing cross-modal transfer learning we study the role of
initialization of existing deep architectures for 3D shape classification. Our
analysis shows that multiview methods continue to offer the best generalization
even without pretraining on large labeled image datasets, and even when trained
on simplified inputs such as binary silhouettes. Furthermore, the performance
of voxel-based 3D convolutional networks and point-based architectures can be
improved via cross-modal transfer from image representations. Finally, we
analyze the robustness of 3D shape classifiers to adversarial transformations
and present a novel approach for generating adversarial perturbations of a 3D
shape for multiview classifiers using a differentiable renderer. We find that
point-based networks are more robust to point position perturbations while
voxel-based and multiview networks are easily fooled with the addition of
imperceptible noise to the input.
|
['Jong-Chyi Su', 'Subhransu Maji', 'Rui Wang', 'Matheus Gadelha']
|
2018-09-07
| null | null | null | null |
['3d-shape-retrieval']
|
['computer-vision']
|
[ 4.11028750e-02 2.45820805e-01 2.71419078e-01 -3.65756810e-01
-8.90558660e-01 -1.15640485e+00 9.25338507e-01 5.01391701e-02
-1.22298360e-01 1.71941817e-01 -1.37327656e-01 -3.66010875e-01
1.95942849e-01 -9.43737566e-01 -1.17984927e+00 -4.82020080e-01
-1.31759062e-01 6.31270409e-01 -2.31598578e-02 -3.99971813e-01
1.71625033e-01 1.27624011e+00 -1.50762653e+00 2.27837950e-01
3.77167910e-01 1.03469419e+00 -6.12360001e-01 7.25650251e-01
9.06413123e-02 2.97137290e-01 -8.69205475e-01 -3.63936633e-01
6.58598959e-01 7.22547323e-02 -6.65423274e-01 1.51117846e-01
1.05941761e+00 -4.40044224e-01 -4.45486426e-01 7.95697629e-01
8.29017401e-01 1.23319253e-01 1.14132762e+00 -1.47694087e+00
-1.19696832e+00 -3.95093672e-02 -8.66881236e-02 -4.31899801e-02
5.37983358e-01 5.31192303e-01 5.05329311e-01 -8.52275372e-01
7.99347460e-01 1.65811408e+00 9.45496440e-01 6.11254811e-01
-1.67395997e+00 -4.40069884e-01 9.98205990e-02 -4.03661728e-01
-1.01560247e+00 -4.00546849e-01 1.02161372e+00 -5.93345761e-01
1.08253860e+00 3.61920834e-01 2.73768723e-01 1.38212621e+00
1.83763117e-01 4.97924477e-01 9.03974891e-01 -2.60094196e-01
5.77673651e-02 -9.08549353e-02 -4.24507737e-01 6.72566235e-01
-1.63052261e-01 4.20397133e-01 1.20809138e-01 -3.33816946e-01
8.92576039e-01 -1.43093571e-01 -9.76586416e-02 -7.76843309e-01
-1.10490799e+00 7.56563962e-01 7.75814891e-01 9.63122174e-02
-5.53952269e-02 2.99709648e-01 5.87644994e-01 5.85461259e-01
7.25255728e-01 6.28966331e-01 -3.82780880e-01 3.72037113e-01
-4.08808738e-01 4.06545907e-01 4.37100828e-01 1.07332444e+00
6.38564765e-01 5.82529306e-01 6.56373203e-02 6.49469137e-01
-4.27132957e-02 7.67651796e-01 2.76623785e-01 -1.08573210e+00
5.24741769e-01 5.81649780e-01 -1.07525952e-01 -1.15378070e+00
-4.49978560e-01 -1.96839139e-01 -7.30190873e-01 9.41782594e-01
4.45853770e-01 -1.68035440e-02 -1.11518288e+00 1.63440120e+00
2.44149327e-01 -3.45392749e-02 9.85035896e-02 6.56127393e-01
1.05226386e+00 2.88644075e-01 1.21964782e-01 5.11205554e-01
6.87281370e-01 -1.97394475e-01 -4.39800993e-02 1.59153149e-01
6.54325008e-01 -7.12089896e-01 1.05297005e+00 -2.07747415e-01
-1.29604208e+00 -7.14872420e-01 -1.12218630e+00 -1.46094069e-01
-7.31396437e-01 -3.56108457e-01 1.28344759e-01 7.38674462e-01
-1.04055905e+00 9.97258246e-01 -7.71010399e-01 4.78359777e-03
6.69040501e-01 4.24829781e-01 -8.00228655e-01 6.24403693e-02
-8.09622407e-01 1.03231144e+00 2.78866831e-02 -2.48626396e-01
-1.21303916e+00 -1.04870498e+00 -1.14808393e+00 -9.66404304e-02
-3.90781224e-01 -7.76009619e-01 9.77962375e-01 -1.02603686e+00
-1.27597404e+00 1.18984699e+00 2.15802819e-01 -2.13927701e-01
7.83538163e-01 1.98782429e-01 -2.12137401e-01 3.18459183e-01
-4.66953553e-02 9.18005228e-01 1.22324371e+00 -1.72152483e+00
9.75773856e-02 -4.09607440e-01 2.90063500e-01 1.05568975e-01
1.16347857e-01 -3.39042187e-01 3.49160641e-01 -8.59679282e-01
2.90031254e-01 -9.15129721e-01 -7.04886094e-02 3.79870296e-01
-3.72751594e-01 -6.92406520e-02 1.30036616e+00 -5.15293300e-01
-2.30223984e-01 -2.47486567e+00 1.87399387e-01 1.47740766e-01
5.00800684e-02 3.09326231e-01 -4.34438109e-01 3.39536190e-01
-5.64776838e-01 6.81423187e-01 -1.01855189e-01 -2.20112160e-01
1.17016263e-01 2.09335074e-01 -4.81039107e-01 7.03424990e-01
4.82814103e-01 1.22643101e+00 -7.49023795e-01 -5.48192747e-02
5.75347960e-01 7.49043107e-01 -5.74044287e-01 2.13225767e-01
-3.04104239e-01 6.88949585e-01 -2.13300943e-01 5.71803153e-01
5.86480856e-01 -7.13216737e-02 -4.89378124e-01 -2.76299983e-01
3.79426211e-01 1.32098034e-01 -8.16722572e-01 1.51369739e+00
-5.40495694e-01 7.30295599e-01 5.93451001e-02 -9.28398550e-01
1.19401312e+00 5.24860680e-01 3.74170840e-01 -4.90219593e-01
2.20201626e-01 1.41039908e-01 -3.13576370e-01 -1.73177764e-01
1.88770279e-01 -3.54084671e-01 -3.10179554e-02 5.27360857e-01
2.17316672e-01 -6.45867646e-01 -6.61368072e-01 5.71637042e-02
9.18635964e-01 2.32435092e-01 -1.73465267e-01 -9.48224217e-02
2.65459120e-01 -1.25222310e-01 4.28972058e-02 5.02969205e-01
-1.13204204e-01 8.78370941e-01 3.85979414e-01 -8.04371297e-01
-1.48145151e+00 -1.43133891e+00 -2.07448751e-01 6.40106797e-01
5.75726666e-02 2.32943505e-01 -3.80812407e-01 -9.20310795e-01
5.03110707e-01 8.62100601e-01 -7.73317516e-01 -4.07280087e-01
-6.64865017e-01 2.09505223e-02 8.56891096e-01 6.75322056e-01
1.72450319e-01 -9.12122548e-01 -3.28298032e-01 -2.30061412e-01
3.31318408e-01 -1.12027657e+00 -1.69865116e-01 2.96008177e-02
-1.02263677e+00 -1.06539810e+00 -7.35210299e-01 -8.82448673e-01
9.21116292e-01 -1.98274814e-02 1.08829594e+00 4.67883609e-02
-1.33177817e-01 9.69292223e-01 -1.70570850e-01 -4.16577548e-01
-9.23669577e-01 3.40552367e-02 1.55301809e-01 -3.41662079e-01
-2.54381061e-01 -1.05149412e+00 -4.17384833e-01 4.17257398e-01
-8.48827899e-01 -4.60934758e-01 8.72795582e-02 6.73722923e-01
3.49519163e-01 -4.11659271e-01 4.76269484e-01 -7.02815592e-01
5.27026534e-01 -1.53527677e-01 -4.05442953e-01 -1.99991371e-02
-5.97123094e-02 6.14168867e-02 9.14299369e-01 -6.25156581e-01
-6.05397642e-01 -3.00225103e-03 -2.71520078e-01 -1.10597456e+00
-6.91138566e-01 -1.41701281e-01 -2.15435788e-01 -7.53128707e-01
1.34208894e+00 3.45491990e-03 3.26169968e-01 -3.45777422e-01
7.30210006e-01 1.19143866e-01 5.92537701e-01 -7.61718810e-01
1.26651108e+00 7.37116754e-01 2.28089198e-01 -8.40695858e-01
-2.69607067e-01 2.24570274e-01 -9.21138883e-01 -1.45055279e-01
8.22530091e-01 -7.73579419e-01 -7.72424757e-01 5.37842631e-01
-1.21907425e+00 -3.39428723e-01 -4.37121660e-01 1.52451696e-03
-1.01883972e+00 4.26453292e-01 -5.44743001e-01 -2.46308252e-01
-2.29636073e-01 -1.39420044e+00 1.32498729e+00 -4.22941059e-01
-2.63578802e-01 -1.15909290e+00 -2.94571817e-01 5.91534190e-02
3.32486033e-01 1.06771111e+00 1.46781135e+00 -7.87252605e-01
-5.40217638e-01 -5.48258424e-01 2.67021023e-02 4.61437434e-01
2.80014038e-01 -6.18417971e-02 -1.23214114e+00 -4.95328873e-01
-2.08381414e-01 -6.80999756e-01 3.78933430e-01 7.73230121e-02
1.29396677e+00 -4.00722116e-01 -6.54580668e-02 1.05987000e+00
1.24955940e+00 -3.44993994e-02 5.75109959e-01 2.70338446e-01
8.53831947e-01 4.64545608e-01 -2.14370210e-02 -1.12217657e-01
-4.82819341e-02 4.43151921e-01 9.09894168e-01 -8.42913315e-02
-1.26813248e-01 -3.75958472e-01 -4.70414422e-02 3.26200545e-01
-1.12889312e-01 2.60580098e-03 -9.78948295e-01 3.75745118e-01
-1.07876325e+00 -8.90758693e-01 3.34677637e-01 2.12237072e+00
3.84904951e-01 1.93774849e-01 8.18996876e-02 2.50909299e-01
6.50666833e-01 3.04403156e-01 -6.15875185e-01 -6.65712774e-01
-1.28367826e-01 1.13238655e-01 5.09862661e-01 4.82683092e-01
-1.23167443e+00 7.84054041e-01 7.19012308e+00 4.14938211e-01
-1.39211810e+00 -1.38171405e-01 6.48610830e-01 8.29266980e-02
-6.14490986e-01 -5.29829979e-01 -2.48576239e-01 1.18920118e-01
5.53811789e-01 1.46120548e-01 2.76191086e-01 1.03807008e+00
-2.34817982e-01 7.33442664e-01 -1.46745026e+00 8.90855312e-01
1.79808006e-01 -1.47375166e+00 5.22146285e-01 -5.91963232e-02
8.02320242e-01 8.33625421e-02 4.10812080e-01 2.23968402e-01
4.79028344e-01 -1.34420836e+00 6.99195206e-01 2.62683660e-01
9.47351098e-01 -8.01315248e-01 2.76512086e-01 2.65564948e-01
-8.36004019e-01 1.23435326e-01 -3.13824236e-01 2.54506826e-01
-2.62151472e-03 -1.16894715e-01 -8.33347201e-01 6.18448317e-01
5.00310183e-01 5.63148737e-01 -4.32202905e-01 6.50072634e-01
-6.66612536e-02 7.58349150e-02 -3.28934819e-01 4.38874692e-01
2.97629476e-01 1.44837111e-01 9.22012091e-01 8.10713291e-01
1.80898160e-01 -1.67648092e-01 9.23747942e-02 9.89199698e-01
-1.94323659e-01 -1.69781640e-01 -1.38253021e+00 1.91139758e-01
3.26750308e-01 8.35378945e-01 -4.54654574e-01 -1.40882388e-01
-8.80929977e-02 8.33825707e-01 3.83243829e-01 6.91440940e-01
-5.70603251e-01 -2.31968805e-01 7.90588021e-01 1.41564578e-01
4.97039139e-01 -4.98479247e-01 -5.47861457e-01 -8.59204113e-01
1.28739923e-02 -1.08803487e+00 3.74322310e-02 -9.84814405e-01
-1.42276955e+00 5.97703993e-01 -4.96913530e-02 -1.33281910e+00
-3.66871744e-01 -8.24875534e-01 -7.45179534e-01 9.76143897e-01
-1.16263044e+00 -1.43316603e+00 -1.88848585e-01 6.31359100e-01
1.98321864e-01 -3.30676943e-01 1.06898069e+00 -4.98309731e-03
2.30686843e-01 8.69037211e-01 -2.19631582e-01 3.91853184e-01
5.16794741e-01 -1.17945302e+00 9.60015893e-01 4.03922737e-01
5.20274453e-02 2.79215187e-01 6.02520227e-01 -3.92426789e-01
-1.38287008e+00 -1.20432782e+00 2.78534830e-01 -9.58395541e-01
3.15795004e-01 -2.30012804e-01 -9.89352465e-01 1.03182197e+00
-1.20977789e-01 4.19466645e-01 4.01140451e-01 -4.15212028e-02
-9.40578103e-01 1.68532744e-01 -1.50432444e+00 6.45739913e-01
1.04669893e+00 -1.01696527e+00 -7.03426778e-01 2.62600511e-01
9.25995767e-01 -7.23668396e-01 -1.22436535e+00 6.27035141e-01
3.59065026e-01 -7.03247786e-01 1.58441520e+00 -9.53289747e-01
4.69183564e-01 -1.17670737e-01 -3.52335781e-01 -1.65602219e+00
-2.97548473e-01 -2.63821185e-01 1.96113214e-01 9.37014520e-01
3.81737769e-01 -7.35700190e-01 7.66825974e-01 5.90551078e-01
-2.76645243e-01 -5.91162384e-01 -1.03238595e+00 -9.53536153e-01
7.42865801e-01 -3.82560998e-01 7.49425530e-01 1.18359447e+00
-5.84341288e-01 -4.97813448e-02 -1.07478444e-02 4.45068955e-01
5.57928801e-01 5.90097383e-02 1.03375316e+00 -9.87901628e-01
-5.50604574e-02 -4.76632416e-01 -1.14361596e+00 -6.58520043e-01
5.34786522e-01 -1.32011688e+00 -3.81608397e-01 -1.19566298e+00
-6.65702105e-01 -5.75155437e-01 2.20060691e-01 5.07918060e-01
2.06320867e-01 4.93294835e-01 4.29754615e-01 -3.06462701e-02
1.62316412e-02 6.57863140e-01 1.51863563e+00 -5.07402658e-01
-7.88417384e-02 -3.99048720e-03 -2.03692764e-01 7.56238818e-01
7.81514823e-01 -2.46355250e-01 -3.71387303e-01 -6.78362310e-01
-1.45096600e-03 -1.60045661e-02 8.97173047e-01 -7.71329880e-01
-1.94341332e-01 2.62645781e-01 7.93304026e-01 -5.18643618e-01
6.41906023e-01 -9.81095314e-01 6.69571087e-02 2.80545056e-01
-4.00803953e-01 2.71556884e-01 6.10405624e-01 4.87902164e-01
9.42641720e-02 3.12349554e-02 9.57954884e-01 -3.10892850e-01
-2.26716816e-01 4.29277718e-01 -1.98493898e-01 1.53875068e-01
9.60739553e-01 -3.77283037e-01 -4.20674056e-01 -5.05347311e-01
-1.00617945e+00 -1.37565747e-01 7.97984660e-01 5.35816729e-01
6.63806915e-01 -1.63746548e+00 -5.59469104e-01 5.14888287e-01
-7.28207678e-02 1.17910653e-01 5.26957586e-02 9.78995562e-02
-7.05510855e-01 -4.54497058e-03 -4.90681350e-01 -1.04120100e+00
-1.11514807e+00 8.60303938e-01 7.83726096e-01 4.63292420e-01
-7.26679146e-01 6.74409151e-01 1.92580208e-01 -1.10472560e+00
3.64837646e-01 -2.97988951e-01 2.48787627e-01 -2.46637806e-01
5.66160753e-02 3.29214036e-01 2.73614079e-01 -7.58377910e-01
-1.93691030e-01 7.33182907e-01 3.24200362e-01 1.19083486e-01
1.25930297e+00 4.24850941e-01 2.01875493e-01 4.59663719e-01
1.65546179e+00 -3.16265643e-01 -1.24830914e+00 1.65695548e-01
-5.49527884e-01 -4.18896198e-01 -3.45202714e-01 -4.29955691e-01
-1.09314227e+00 1.01943707e+00 5.49966514e-01 2.53122330e-01
5.58335543e-01 -9.96432900e-02 3.84117067e-01 4.64617550e-01
1.75742850e-01 -4.14675802e-01 1.93247721e-01 5.13012528e-01
1.33106029e+00 -1.22124994e+00 -1.53292596e-01 -3.79991472e-01
-2.14315280e-01 1.00303018e+00 5.24619639e-01 -6.69273496e-01
9.28338051e-01 6.74479306e-02 4.28121418e-01 -2.84895599e-01
-3.13060284e-01 2.35302061e-01 4.54359770e-01 1.15152347e+00
2.43974179e-02 -4.73521724e-02 6.09549165e-01 -2.09391862e-01
-4.03781354e-01 -7.10617781e-01 1.18745804e-01 7.91435003e-01
-5.41406907e-02 -9.78610575e-01 -5.54134965e-01 2.67821252e-01
-3.24475884e-01 2.05062568e-01 -5.10478020e-01 1.09009528e+00
-1.89765856e-01 3.48684102e-01 1.95683643e-01 -4.77711499e-01
6.09822214e-01 3.50279123e-01 7.29980409e-01 -4.30213064e-01
-7.35632300e-01 -3.35212082e-01 -1.98866334e-02 -4.48083639e-01
-2.32163966e-01 -6.33856058e-01 -8.70177805e-01 -4.05038387e-01
1.75344408e-01 -3.60620558e-01 6.35344505e-01 7.99526036e-01
5.24833918e-01 3.97691786e-01 7.64560521e-01 -1.53828812e+00
-8.14601600e-01 -5.38906574e-01 -1.79420769e-01 9.47378933e-01
6.10874653e-01 -7.07200825e-01 -6.58839226e-01 -1.21473342e-01]
|
[8.26539421081543, -3.9105582237243652]
|
95852a07-a03c-4de5-b6b8-ba4d09703acd
|
arhnet-adaptive-region-harmonization-for
|
2307.01220
| null |
https://arxiv.org/abs/2307.01220v1
|
https://arxiv.org/pdf/2307.01220v1.pdf
|
ARHNet: Adaptive Region Harmonization for Lesion-aware Augmentation to Improve Segmentation Performance
|
Accurately segmenting brain lesions in MRI scans is critical for providing patients with prognoses and neurological monitoring. However, the performance of CNN-based segmentation methods is constrained by the limited training set size. Advanced data augmentation is an effective strategy to improve the model's robustness. However, they often introduce intensity disparities between foreground and background areas and boundary artifacts, which weakens the effectiveness of such strategies. In this paper, we propose a foreground harmonization framework (ARHNet) to tackle intensity disparities and make synthetic images look more realistic. In particular, we propose an Adaptive Region Harmonization (ARH) module to dynamically align foreground feature maps to the background with an attention mechanism. We demonstrate the efficacy of our method in improving the segmentation performance using real and synthetic images. Experimental results on the ATLAS 2.0 dataset show that ARHNet outperforms other methods for image harmonization tasks, and boosts the down-stream segmentation performance. Our code is publicly available at https://github.com/King-HAW/ARHNet.
|
['Rachel Sparks', 'Sebastien Ourselin', 'Alejandro Granados', 'Xi Ouyang', 'Yang Liu', 'Jiayu Huo']
|
2023-07-02
| null | null | null | null |
['image-harmonization']
|
['computer-vision']
|
[ 9.57158580e-02 1.36552960e-01 -1.66554555e-01 -3.63177657e-01
-6.83843851e-01 -1.91154331e-01 3.57292205e-01 -5.22383228e-02
-5.69648564e-01 6.31098330e-01 4.74047847e-02 -1.11559242e-01
3.94502312e-01 -5.15123188e-01 -5.17407119e-01 -8.29963386e-01
1.88622311e-01 2.05586806e-01 7.76123881e-01 -1.44182995e-01
7.29613975e-02 4.58207071e-01 -1.14127648e+00 1.90602407e-01
1.23833799e+00 8.84750605e-01 4.56339091e-01 2.98033088e-01
-2.40171552e-01 5.82542121e-01 -5.51314831e-01 -1.45337552e-01
3.33111465e-01 -4.67998117e-01 -7.80431807e-01 -3.19380499e-02
3.03110391e-01 -3.88325870e-01 -3.71506661e-01 1.20218825e+00
6.37452126e-01 5.76257482e-02 4.15506959e-01 -1.34906137e+00
-3.32197785e-01 6.87190831e-01 -9.65720177e-01 6.93404794e-01
-3.20714742e-01 3.31363827e-01 3.69701117e-01 -7.07571983e-01
5.05455256e-01 8.72099996e-01 5.36553204e-01 8.18378091e-01
-1.13383603e+00 -8.70382667e-01 2.83353209e-01 1.82066694e-01
-1.40389395e+00 -4.05024439e-01 7.72782922e-01 -3.98507923e-01
3.43900293e-01 3.36853445e-01 9.40301597e-01 8.51951778e-01
1.03496984e-01 1.05428290e+00 1.20516634e+00 -1.22811690e-01
2.18255699e-01 -2.93496847e-01 1.30658180e-01 5.78859448e-01
2.07637250e-01 -2.13569805e-01 -3.02471101e-01 7.21722469e-02
1.16970277e+00 3.54949869e-02 -6.36877060e-01 -3.66565287e-01
-1.37952161e+00 6.20648324e-01 8.72781098e-01 4.64364499e-01
-2.99024314e-01 1.18565120e-01 4.05541301e-01 -2.79353082e-01
6.27908111e-01 2.65385509e-01 -2.09145807e-02 2.68110335e-01
-1.07719874e+00 9.91045088e-02 1.88950658e-01 7.08032906e-01
3.57746720e-01 9.35987532e-02 -4.35112149e-01 9.83361900e-01
6.92203715e-02 2.98346698e-01 7.53873587e-01 -9.60767567e-01
1.82889625e-01 5.93331754e-01 -2.42273927e-01 -6.20098293e-01
-6.15656972e-01 -5.62764466e-01 -1.00824380e+00 2.93145120e-01
5.31768024e-01 -2.76558120e-02 -1.23476446e+00 1.63191450e+00
6.14527285e-01 3.45623434e-01 -2.66434461e-01 1.13325298e+00
8.93587470e-01 3.65333229e-01 1.91906795e-01 -1.97230175e-01
1.18708897e+00 -1.22408211e+00 -9.10166919e-01 -4.12158579e-01
5.46001315e-01 -5.29671609e-01 1.26129353e+00 1.08133972e-01
-1.21437240e+00 -2.73977369e-01 -8.22915971e-01 1.83695536e-02
3.28537333e-03 -1.38133198e-01 5.15700579e-01 5.28007030e-01
-9.70519066e-01 3.47408324e-01 -1.27946997e+00 7.73430169e-02
1.06618929e+00 3.10020477e-01 -1.27246320e-01 5.20265475e-02
-8.94525588e-01 6.94669843e-01 5.55488467e-01 2.39510491e-01
-6.01173580e-01 -9.58612025e-01 -7.12871730e-01 -2.16583893e-01
2.83244908e-01 -5.77145696e-01 1.35830045e+00 -9.98875320e-01
-1.32203329e+00 9.18012798e-01 2.01909002e-02 -4.44699287e-01
8.69234204e-01 1.32923156e-01 -5.70821017e-02 3.75654548e-01
1.44286707e-01 1.04984009e+00 6.66563988e-01 -1.23533285e+00
-4.01541770e-01 -4.22270715e-01 -2.94697195e-01 1.56545490e-01
-6.20630793e-02 1.53132826e-01 -7.18744338e-01 -1.04720140e+00
1.77700564e-01 -8.13742340e-01 -5.82246423e-01 2.98061132e-01
-4.33592111e-01 2.33450547e-01 8.84491384e-01 -8.30283344e-01
1.12925541e+00 -2.09974432e+00 -1.89191531e-02 1.89507946e-01
4.82691795e-01 3.78665626e-01 8.86796787e-02 -4.78745908e-01
-1.37917757e-01 -8.63455807e-06 -8.09649944e-01 -3.02036166e-01
-4.09047514e-01 2.32633606e-01 2.64355037e-02 5.03889263e-01
1.29693702e-01 1.21975493e+00 -7.90067315e-01 -7.99061358e-01
2.75401682e-01 5.23130655e-01 -5.47328293e-01 1.64707989e-01
-1.17142662e-01 9.60736632e-01 -3.74243945e-01 6.71140134e-01
7.87772655e-01 -2.56366044e-01 -1.61695883e-01 -2.36765221e-01
-2.42468845e-02 -2.74404734e-01 -8.35981727e-01 1.73719978e+00
-1.61126226e-01 4.59638208e-01 2.00836867e-01 -8.62319648e-01
5.64696908e-01 9.05304328e-02 6.97303355e-01 -8.93992841e-01
4.27934676e-01 2.50404418e-01 2.68886536e-01 -3.25202018e-01
-3.08230538e-02 -2.86427233e-03 3.47874314e-01 2.64421850e-01
-3.67385119e-01 -2.77757317e-01 2.29200304e-01 1.67813867e-01
7.13417590e-01 -9.59929675e-02 1.18657254e-01 -3.12727213e-01
3.62564206e-01 8.64761397e-02 8.22388411e-01 4.86120880e-01
-5.98018169e-01 1.01117110e+00 5.15479863e-01 -3.28657061e-01
-1.04647028e+00 -9.98263478e-01 -2.65609086e-01 6.74248934e-01
3.87842089e-01 -5.24967127e-02 -1.32639992e+00 -6.65816784e-01
-3.42945188e-01 5.73793888e-01 -7.50231504e-01 -1.52211264e-01
-8.02890301e-01 -1.11568475e+00 4.47364450e-01 7.38589227e-01
8.10837269e-01 -1.06414568e+00 -8.67154956e-01 1.65074810e-01
-5.96366704e-01 -1.16936171e+00 -8.02301168e-01 -1.07433334e-01
-9.86625075e-01 -9.93002892e-01 -1.13612115e+00 -7.66253948e-01
9.51495826e-01 2.51305312e-01 8.70162725e-01 4.13716406e-01
-6.04644120e-01 -1.65282726e-01 -3.40058237e-01 -4.93785173e-01
-3.97621274e-01 -3.65089439e-02 -3.10453385e-01 -1.36029273e-01
-3.27184618e-01 -4.88329679e-01 -1.05167007e+00 5.13883352e-01
-1.25557113e+00 6.01590216e-01 5.38395941e-01 7.81463206e-01
8.76333296e-01 -2.88281202e-01 4.58681494e-01 -8.88801694e-01
4.02505606e-01 -2.71048725e-01 -6.25716925e-01 9.36467499e-02
-4.02495474e-01 -1.58201113e-01 3.59313548e-01 -6.48298621e-01
-9.82176304e-01 1.15465261e-01 -2.78572500e-01 -4.30847913e-01
-3.22471187e-02 1.57255054e-01 -1.17493875e-01 -3.50212604e-01
4.76958394e-01 1.26305163e-01 1.76343754e-01 -2.14448258e-01
2.43542165e-01 3.36606741e-01 8.43631804e-01 -3.13313514e-01
5.02931178e-01 7.37051070e-01 -2.78174520e-01 -4.54122543e-01
-7.69425511e-01 -2.90386885e-01 -7.26499915e-01 -5.05304039e-01
9.42900479e-01 -6.34332061e-01 -1.84000760e-01 6.44733727e-01
-9.09589589e-01 -8.47261608e-01 -3.25035751e-01 3.66443604e-01
-5.18173993e-01 3.33268940e-01 -6.48058653e-01 -3.22469652e-01
-6.43593252e-01 -1.66115320e+00 9.40153480e-01 4.71102715e-01
1.78223960e-02 -7.37593293e-01 -2.21741900e-01 6.39588892e-01
5.28522432e-01 6.16385639e-01 8.22149456e-01 -4.46679413e-01
-5.19560397e-01 7.77277499e-02 -3.44898194e-01 2.61311471e-01
3.88050899e-02 -1.36281177e-01 -9.17601705e-01 -1.97319031e-01
-1.03914253e-01 -4.07290310e-02 8.57278764e-01 7.57112086e-01
1.66400146e+00 -1.06472537e-01 -3.26675117e-01 8.90602767e-01
1.06531644e+00 2.01435685e-01 7.60602772e-01 4.62005585e-01
8.05298269e-01 4.44360882e-01 4.78575557e-01 2.16130748e-01
2.91515946e-01 7.25697994e-01 4.87963319e-01 -8.14561844e-01
-6.48603559e-01 2.33224466e-01 -6.93167299e-02 7.09951460e-01
1.02352966e-02 -2.01375093e-02 -1.20544517e+00 5.94574392e-01
-1.85699689e+00 -6.20880842e-01 -2.35473961e-01 1.98266625e+00
1.06826580e+00 1.28335148e-01 2.95052439e-01 -7.57249296e-02
8.73100340e-01 -9.06193256e-02 -7.69055009e-01 1.94574311e-01
-1.88350216e-01 1.05894782e-01 5.24281979e-01 4.72707212e-01
-1.09649038e+00 1.03030765e+00 5.59996223e+00 9.44779158e-01
-1.26387155e+00 5.05326450e-01 1.03704381e+00 -2.56116301e-01
-1.24180935e-01 -4.04129326e-01 -3.67762834e-01 6.51704431e-01
2.77213007e-01 -1.31360024e-01 2.31221050e-01 6.10633194e-01
3.81441653e-01 -2.18266577e-01 -6.61073804e-01 9.33934033e-01
-2.20869984e-02 -1.41716349e+00 -1.18396752e-01 -7.20058829e-02
8.54258060e-01 1.65422857e-01 5.91553375e-02 4.04067757e-03
4.48477222e-03 -8.88099015e-01 8.45693231e-01 3.24859619e-01
7.26918817e-01 -7.42542088e-01 6.94619894e-01 8.71928036e-02
-1.01517892e+00 2.10841149e-01 -1.43050596e-01 5.94206989e-01
2.64671803e-01 5.83929837e-01 -7.30948627e-01 2.62533218e-01
7.92864561e-01 2.79883415e-01 -7.65230775e-01 1.50049388e+00
-1.55023426e-01 4.28812534e-01 -2.35082239e-01 4.33359563e-01
1.90487280e-01 -2.22035229e-01 4.97879863e-01 1.10593355e+00
-9.96997431e-02 3.19263309e-01 1.94321468e-01 1.04756653e+00
-9.70486726e-05 3.10353100e-01 -1.79613680e-01 3.06425393e-01
2.56343782e-01 1.32883072e+00 -1.37686777e+00 -4.23593909e-01
-3.36233824e-01 9.14567232e-01 2.46249914e-01 3.51006597e-01
-1.03004444e+00 -1.10327281e-01 4.51693922e-01 3.10623169e-01
1.14954738e-02 -3.90694402e-02 -5.27265668e-01 -1.14619744e+00
-1.59021504e-02 -7.99947560e-01 3.91147524e-01 -8.02586854e-01
-9.29386675e-01 7.94110954e-01 1.42485961e-01 -1.04952848e+00
2.60384470e-01 -2.46221468e-01 -6.97222471e-01 5.48257530e-01
-1.52650213e+00 -9.86935318e-01 -7.21808791e-01 6.53498709e-01
6.69618011e-01 2.31176421e-01 2.62760133e-01 4.45004880e-01
-9.68384385e-01 5.39540827e-01 -2.28506386e-01 1.80569589e-01
6.31882906e-01 -1.08315372e+00 3.95864904e-01 1.10580420e+00
-2.03262269e-01 2.69538641e-01 6.10427499e-01 -6.91893160e-01
-6.95608199e-01 -1.24267304e+00 -2.34175622e-02 -2.57293973e-02
5.69023669e-01 -1.84734091e-01 -1.33173740e+00 5.19156992e-01
2.68853724e-01 4.05284196e-01 4.75511193e-01 -7.27457941e-01
1.84627827e-02 9.67460796e-02 -1.27053773e+00 8.59145164e-01
8.80778730e-01 3.05558927e-03 -3.41421127e-01 4.72646296e-01
7.28634536e-01 -8.11784208e-01 -8.78835738e-01 6.52120113e-01
2.58888185e-01 -8.46470773e-01 9.99851584e-01 -1.83487445e-01
3.53521705e-01 -4.61765587e-01 2.53110945e-01 -1.28470135e+00
-5.42832799e-02 -4.04224336e-01 -7.55747734e-03 9.08730090e-01
3.01446021e-01 -6.68525398e-01 8.17877412e-01 8.53509724e-01
-3.17352325e-01 -1.02731550e+00 -9.71594870e-01 -6.18257880e-01
2.17881113e-01 -2.89730132e-01 5.87951660e-01 9.38256562e-01
-3.30566168e-01 -2.19420403e-01 1.21276066e-01 1.55933917e-01
7.19008923e-01 -1.58697404e-02 5.64042270e-01 -9.32687521e-01
6.85011894e-02 -9.15158570e-01 -2.87502229e-01 -5.22652686e-01
7.57824033e-02 -1.01768744e+00 -5.59394024e-02 -1.56108844e+00
3.38996351e-01 -4.15856540e-01 -2.91014940e-01 7.06314087e-01
-4.37109649e-01 6.23302460e-01 2.32563987e-01 2.84120530e-01
-3.87182385e-01 6.45148396e-01 1.76838338e+00 -1.67385519e-01
-1.52339876e-01 -5.94619513e-02 -5.52671313e-01 8.78708005e-01
1.19669068e+00 -4.53182131e-01 -3.33755285e-01 -5.03250599e-01
-4.56146836e-01 -1.63864657e-01 5.76601803e-01 -9.89960551e-01
1.86364353e-01 3.15672643e-02 5.20494461e-01 -5.09855032e-01
3.98813672e-02 -5.31167805e-01 -6.10166648e-03 7.26033807e-01
-2.07279921e-01 1.24328025e-01 4.90873247e-01 2.00106695e-01
-1.00211687e-01 -6.10167086e-02 1.33243930e+00 -1.43905252e-01
-6.36178315e-01 5.78767657e-01 -1.20054342e-01 3.34337115e-01
1.17239213e+00 -2.44209051e-01 -2.15890124e-01 -3.73304039e-02
-7.05133021e-01 4.48498636e-01 5.87546349e-01 3.13873589e-01
8.17544401e-01 -1.24202061e+00 -6.45354450e-01 2.96043754e-01
7.32677951e-02 3.90461594e-01 3.69275510e-01 1.44643784e+00
-9.69874501e-01 -6.82349578e-02 -3.47084641e-01 -7.75816441e-01
-1.26046824e+00 2.34595701e-01 6.99264824e-01 -8.22976753e-02
-1.06144917e+00 8.55046391e-01 5.99267364e-01 -1.90286919e-01
3.61112416e-01 -6.06169999e-01 -1.60871342e-01 -2.33744010e-01
7.47358143e-01 9.28039029e-02 1.42632753e-01 -6.29775703e-01
-4.20545340e-01 4.15154755e-01 -3.61650825e-01 -9.11440253e-02
1.16885197e+00 -1.49747699e-01 -6.82206973e-02 4.63449359e-02
8.58146787e-01 -3.35843951e-01 -1.42635834e+00 -2.66037673e-01
-1.48727447e-01 -5.70879042e-01 4.32143420e-01 -6.77528799e-01
-1.79229093e+00 7.10104167e-01 8.79464507e-01 -1.69791386e-01
1.28047740e+00 5.26105464e-02 1.15464044e+00 -2.69331753e-01
1.70077667e-01 -9.10572350e-01 9.87593010e-02 6.12427071e-02
9.23800170e-01 -1.22927237e+00 -2.16302931e-01 -5.67209899e-01
-8.93640637e-01 8.14731896e-01 9.94826794e-01 5.82353212e-02
3.78855675e-01 5.74788332e-01 4.00599301e-01 -1.59952328e-01
-2.72354871e-01 -2.44207144e-01 3.23550701e-01 4.75295484e-01
3.08771491e-01 -9.04906467e-02 -3.81626338e-01 4.29519296e-01
-2.99818013e-02 -2.16936409e-01 5.17093360e-01 9.11414802e-01
-3.94083083e-01 -9.33002591e-01 -5.34490108e-01 3.41821015e-01
-5.07878125e-01 -1.01907082e-01 -2.42141068e-01 8.52692485e-01
1.53071001e-01 5.56207240e-01 5.26500121e-02 1.01250403e-01
3.36436182e-01 -2.57353187e-01 6.08890235e-01 -4.49083239e-01
-4.91012573e-01 4.16412741e-01 -4.15961117e-01 -6.96418107e-01
-3.29604596e-01 -5.95783651e-01 -1.65817416e+00 -5.40804155e-02
-1.32570192e-01 -1.14493951e-01 6.22231185e-01 8.07794571e-01
1.47452340e-01 9.03399587e-01 3.54051143e-01 -9.09867525e-01
-1.09403439e-01 -7.09055603e-01 -3.61959040e-01 4.40871358e-01
1.70302168e-01 -7.37881541e-01 -6.31079972e-02 9.83861610e-02]
|
[14.489340782165527, -2.237053871154785]
|
0b48d66d-b80c-4f22-8388-c301aba6efed
|
improving-continuous-sign-language-1
|
2212.13023
| null |
https://arxiv.org/abs/2212.13023v1
|
https://arxiv.org/pdf/2212.13023v1.pdf
|
Improving Continuous Sign Language Recognition with Consistency Constraints and Signer Removal
|
Most deep-learning-based continuous sign language recognition (CSLR) models share a similar backbone consisting of a visual module, a sequential module, and an alignment module. However, due to limited training samples, a connectionist temporal classification loss may not train such CSLR backbones sufficiently. In this work, we propose three auxiliary tasks to enhance the CSLR backbones. The first task enhances the visual module, which is sensitive to the insufficient training problem, from the perspective of consistency. Specifically, since the information of sign languages is mainly included in signers' facial expressions and hand movements, a keypoint-guided spatial attention module is developed to enforce the visual module to focus on informative regions, i.e., spatial attention consistency. Second, noticing that both the output features of the visual and sequential modules represent the same sentence, to better exploit the backbone's power, a sentence embedding consistency constraint is imposed between the visual and sequential modules to enhance the representation power of both features. We name the CSLR model trained with the above auxiliary tasks as consistency-enhanced CSLR, which performs well on signer-dependent datasets in which all signers appear during both training and testing. To make it more robust for the signer-independent setting, a signer removal module based on feature disentanglement is further proposed to remove signer information from the backbone. Extensive ablation studies are conducted to validate the effectiveness of these auxiliary tasks. More remarkably, with a transformer-based backbone, our model achieves state-of-the-art or competitive performance on five benchmarks, PHOENIX-2014, PHOENIX-2014-T, PHOENIX-2014-SI, CSL, and CSL-Daily.
|
['Brian Mak', 'Ronglai Zuo']
|
2022-12-26
| null | null | null | null |
['sign-language-recognition']
|
['computer-vision']
|
[ 0.05365623 -0.27874994 -0.17789061 -0.5487478 -0.55654395 -0.34369957
0.6444366 -0.7131609 -0.55293155 0.35602242 0.4424501 -0.05175062
0.08628836 -0.24588093 -0.62766206 -0.97137797 0.15623601 -0.18218398
0.1722108 -0.24315293 0.06179138 0.44269308 -1.2754958 0.33945006
1.0031712 0.98715776 0.13638899 0.13851155 -0.06926739 0.9633838
-0.36710176 -0.16418435 0.2967403 -0.6913401 -0.26147455 0.2042395
0.7553034 -0.46579415 -0.7954218 0.86870486 0.7130926 0.05227699
0.39923313 -1.4542687 -0.90487325 0.10448461 -0.64177656 -0.18339404
0.13899693 0.6776374 1.1977438 -0.98420703 0.82806885 1.0649917
0.42014003 0.74561346 -0.92803013 -0.8151018 0.7533844 0.35198045
-1.2602026 -0.42623624 1.182292 -0.24076004 0.68738794 0.23389381
0.80646706 1.2845451 -0.14921078 1.4409409 1.230723 -0.20678532
0.07127267 -0.10938239 0.10263514 0.7698178 -0.08588234 0.19018564
-0.8567395 0.24165305 0.73255527 0.1978472 -0.5133454 -0.45908272
-1.0896976 0.49119803 0.7944072 0.3720104 -0.18572892 0.12196416
0.37475806 0.39687493 -0.01674746 -0.10513677 -0.25199598 0.09263961
-0.7656695 -0.05936655 0.26041976 0.9572219 0.25482506 0.07518099
-0.64635456 0.816796 0.6043081 0.60310644 0.4678203 -0.4098736
0.78079855 0.7221649 -0.10212447 -0.68206877 -0.36617747 -0.6088639
-0.9753287 0.27306226 0.31548816 0.2187636 -1.1715304 2.1422675
0.01974327 0.18095843 -0.07241057 1.3948263 0.9043685 0.33568606
0.14621292 0.11449013 1.2119403 -1.1498973 -0.5734008 -0.29630718
0.5934618 -0.36135215 1.4064294 0.14523378 -0.8012062 -0.42022285
-0.9786676 -0.25175384 -0.03126478 0.5921418 0.5749174 0.05827489
-0.6225576 0.2285671 -0.87960327 -0.3136241 0.6233871 -0.03963511
-0.45939004 -0.3316931 -1.1485467 0.8218642 -0.3071429 0.5676736
-0.55258864 -0.4198123 -0.98399055 -0.10237922 -0.10457548 -0.6119243
0.69806004 -1.1364926 -1.386465 0.9875429 -0.289686 -0.04588757
0.90871906 -0.05375547 -0.3135376 0.15589991 -0.11994831 0.6975549
0.97806823 -1.2527428 -0.35864776 -0.49233714 -0.13418633 0.24436125
-0.35109442 0.05970315 -0.6448335 -0.9591793 0.33464262 -0.88860947
0.08385237 0.6109935 -0.3629524 -0.22428277 0.7815649 -0.9514341
0.9493297 -2.4710162 0.25238544 0.2786578 -0.13855079 0.2849158
-0.7443436 0.13218139 -0.04909595 -0.21956162 -0.42479104 -0.6838338
0.10193651 0.4031211 -0.35807145 0.6711839 0.33492002 1.288575
-0.75402296 -0.3999391 0.11632616 0.5705451 -0.3759247 0.10247625
-0.02857813 0.76737094 -0.4322601 0.8692525 0.77925706 -0.20666441
0.12774728 -0.2948819 0.02791767 0.2803582 -0.93895197 1.8748767
-0.43759793 0.6192896 0.1260477 -1.0420814 0.8398957 0.16027048
0.44777074 -1.2739242 -0.03486736 0.17779174 0.03469233 -0.5410472
-0.10936356 -0.04642622 0.08224262 0.48396522 -0.07594959 0.39589572
0.00653392 0.18296506 1.1081634 0.3730739 -0.05621293 0.17847945
0.5147267 -0.40063232 0.85581225 0.54839665 -0.5278279 0.6487663
0.44129622 -0.15312403 -0.6579245 -1.2320017 -0.11596254 1.0325606
0.31682193 -0.12153713 -0.19687481 -1.073607 0.04302482 0.5382124
-0.6155372 -0.2748557 -0.91675746 -0.4269164 0.60048085 0.87149405
0.8309425 -1.2703007 -0.4945129 -0.14893562 -0.24313767 -1.1013695
-0.91714025 -0.22868547 -0.5687112 -1.0757908 -0.9654548 -1.0552723
1.0555633 0.25419858 0.5207125 0.27218682 -0.2009093 0.34473392
-0.46513766 0.01891812 0.10693011 -0.3558622 -0.23110527 0.35710418
0.26820856 -0.7450488 -0.690829 0.43232366 -0.8737088 0.26173496
0.9717974 1.2511647 0.42514646 -0.5767457 0.35817224 -0.11838662
0.23572503 -0.01086909 -0.38561085 0.55003977 -0.19791208 0.19335137
0.6738105 -0.59862536 -0.9266479 0.24937041 -0.15323849 -0.429277
0.01229381 0.3155344 -0.6141848 -0.14442776 0.15634544 0.8701031
0.14326087 -0.5775289 0.29759434 0.51954854 0.5223856 -0.3208805
0.96490484 0.7066273 -0.09892823 -0.69913095 -0.7160296 -0.38756108
-0.5851904 -0.19657972 0.63263416 -0.9028741 -0.6424524 0.93487847
-1.1121633 -0.42796868 -0.29566386 0.55045575 -0.42193407 0.62047803
-0.53674114 -0.5262256 -0.41604745 -0.95950645 1.2223288 0.12775593
0.09356889 -0.5787691 -0.01819564 0.3771736 0.28957343 0.03647257
0.78476155 -0.3432275 -0.81887776 -0.14547633 -0.538512 0.69429386
0.10405283 -0.31098273 -0.90080917 -0.51562107 -0.29004568 -0.49062058
1.3646619 0.19420244 0.9769818 -0.37633833 -0.09628526 0.86581004
1.048781 -0.11157255 0.8919988 0.19238168 0.70008177 0.47442776
0.49445963 0.17813723 0.6090784 0.9252207 0.24761246 -0.22058448
-0.5546701 -0.46938425 0.7411537 0.7375942 -0.07383642 0.0287515
-0.4279322 0.43640438 -2.006943 -1.0665015 0.25185418 2.0620303
0.8808 -0.15563445 -0.04980829 0.12211407 0.40199924 0.60301405
-0.7565405 0.12784316 -0.5017516 -0.01372689 0.09748523 0.20828128
-0.9667326 0.90612334 4.3353734 0.61953664 -1.5042235 0.15646164
0.06743496 -0.29086372 -0.40240887 0.03932825 -0.6382145 0.6525798
-0.16629212 0.4620714 0.3292914 0.5997238 0.42500326 0.19893894
-1.0094616 0.9729141 0.3083646 -0.985377 0.04042405 0.03353736
0.5388932 0.06505211 0.17528836 0.37920457 -0.05930556 -0.93140864
0.8370492 0.7489393 0.93892103 -0.43243048 0.55892885 0.28457493
-1.4015682 -0.11704854 0.07224405 0.16917485 0.34926027 0.3010129
0.06074547 0.40275812 0.43514225 1.0113873 -0.63645786 1.0159469
-0.9337316 0.37736636 -0.33597264 0.0320086 0.36318612 -0.13375433
0.67176145 1.1353636 -0.01835885 -0.06618739 0.10128251 1.0238302
0.05243545 -0.05238247 -0.391919 0.2251792 0.2471615 1.0604582
-0.05565711 -0.01218006 -0.5894134 1.4627014 0.3590734 0.8360112
-0.8794501 -0.47198322 0.8161718 -0.07287887 0.5447471 -0.4653523
-0.19972229 -1.4998518 0.80779296 -0.77333283 0.36277553 -0.7411588
-1.3594921 0.37491593 -0.46909097 -1.6087252 -0.04357201 -0.5877523
-0.95505446 0.90510476 -1.932412 -1.8126638 -0.573498 0.83524853
0.19669347 -0.13683546 0.45734578 0.31260896 -0.6863381 1.2336714
-0.07439365 0.6943436 0.77009064 -0.833196 0.13932861 1.1052251
0.28124282 0.66628194 0.02787511 -0.46966845 -1.4668303 -1.1835818
0.9657125 -0.15239696 0.4541024 -0.5326688 -0.86318856 0.6398405
-0.26606312 0.47351286 0.2612685 -0.17523697 -0.80182004 -0.28468487
-0.86038524 0.76231647 1.4949095 -0.8092456 -0.6349109 0.26772115
0.23217587 -0.24962044 -0.54811007 0.48621148 0.76147175 -0.76085645
0.8686612 -0.7697533 0.40314087 -0.40610072 -0.12993315 -1.1452342
-0.12463988 -0.40085277 -0.23497829 1.2598937 0.08296275 -0.69880986
0.5493627 0.35197902 -0.04514411 -0.8269466 -1.3073872 -1.16894
-0.09209158 -0.32613826 0.29166037 0.84573567 -0.16266102 0.32052284
-0.52515084 0.02599267 0.61459744 0.5009749 0.8129998 -0.8162774
-0.32483074 -0.70638037 -0.31046897 -1.5041147 0.3631219 -1.0061438
0.02320008 -1.6063302 0.21979533 -0.46716657 -0.50623304 0.95552343
-0.02874634 0.0507336 0.55294764 0.35139713 -0.6488777 1.1500143
1.5970731 -0.23580141 -0.21598357 -0.09125408 -0.47346354 0.62779206
0.4548246 -0.16131957 -0.18731709 -0.49635032 -0.16399732 -0.30601662
0.8855598 -0.58313876 0.44831508 -0.15468383 0.28672764 -0.5291011
0.31271097 -0.7445273 -0.45507056 0.51783895 -0.3696757 -0.37162665
-0.05162081 0.48179913 -0.3655106 0.22732526 1.047882 0.1471867
-1.0119679 0.514247 0.02733039 0.1126202 0.6946963 -0.19511224
-0.44127262 -0.39008832 -0.44932005 0.44873628 0.33609912 0.67398995
0.9988987 -1.560089 -0.773787 0.5760407 0.49593282 -0.16424637
0.50892895 1.3557628 -0.06263495 0.3251283 -0.30524158 -0.58860034
-1.2195516 0.21943654 0.43637872 -0.28673276 -1.0593735 0.78134876
0.37669942 -0.42977935 0.6139346 -0.31165037 0.07584603 -0.14587863
0.5617151 -0.04479322 -0.2929448 -0.684174 -0.7784234 0.83367753
0.03760296 -0.0804194 1.1479833 0.0776756 0.03362716 0.05813668
1.2653062 -0.05539812 -1.5463438 -0.6207697 -0.2699938 -0.40707275
-0.10437846 -0.9299765 -1.3182629 0.99241877 0.5596858 -0.8197931
1.41645 0.0356443 0.8504183 0.28131524 0.18750961 -1.0229057
0.25882486 0.6209833 1.3997223 -1.3838278 -0.1918397 0.00945555
-0.7656036 0.7937546 0.7558186 -0.00962723 0.48416469 0.11851757
0.07663703 -0.00574135 -0.5147169 -0.3634489 0.7074107 0.3972842
0.1409036 -0.10519768 -0.41493118 0.806967 0.24339043 0.02036462
-0.22742267 0.99102694 -0.10732182 -0.9083146 0.03831798 0.14692892
0.24054873 -0.11612988 -0.49040145 0.94540495 0.06851646 0.4991022
-0.0856775 -0.3834598 0.49981785 0.00886036 0.6169777 -0.11543637
-0.35564974 -0.0865041 -0.06219881 -0.80694544 -0.40853265 -0.62573624
-1.4325581 0.02730878 -0.07604495 -0.44884446 0.4555059 1.1571268
0.3687693 0.23604268 0.636369 -0.48415244 -0.85529244 -0.9056233
-0.68195057 0.74993443 0.3582252 -0.61130476 -0.36545277 -0.21905361]
|
[9.21527099609375, -6.512218475341797]
|
a0f9dce1-9bdc-425a-b8bf-b7d7ec3e7122
|
on-distillation-of-guided-diffusion-models
|
2210.03142
| null |
https://arxiv.org/abs/2210.03142v3
|
https://arxiv.org/pdf/2210.03142v3.pdf
|
On Distillation of Guided Diffusion Models
|
Classifier-free guided diffusion models have recently been shown to be highly effective at high-resolution image generation, and they have been widely used in large-scale diffusion frameworks including DALLE-2, Stable Diffusion and Imagen. However, a downside of classifier-free guided diffusion models is that they are computationally expensive at inference time since they require evaluating two diffusion models, a class-conditional model and an unconditional model, tens to hundreds of times. To deal with this limitation, we propose an approach to distilling classifier-free guided diffusion models into models that are fast to sample from: Given a pre-trained classifier-free guided model, we first learn a single model to match the output of the combined conditional and unconditional models, and then we progressively distill that model to a diffusion model that requires much fewer sampling steps. For standard diffusion models trained on the pixel-space, our approach is able to generate images visually comparable to that of the original model using as few as 4 sampling steps on ImageNet 64x64 and CIFAR-10, achieving FID/IS scores comparable to that of the original model while being up to 256 times faster to sample from. For diffusion models trained on the latent-space (e.g., Stable Diffusion), our approach is able to generate high-fidelity images using as few as 1 to 4 denoising steps, accelerating inference by at least 10-fold compared to existing methods on ImageNet 256x256 and LAION datasets. We further demonstrate the effectiveness of our approach on text-guided image editing and inpainting, where our distilled model is able to generate high-quality results using as few as 2-4 denoising steps.
|
['Stefano Ermon', 'Robin Rombach', 'Tim Salimans', 'Jonathan Ho', 'Diederik P. Kingma', 'Ruiqi Gao', 'Chenlin Meng']
|
2022-10-06
| null |
http://openaccess.thecvf.com//content/CVPR2023/html/Meng_On_Distillation_of_Guided_Diffusion_Models_CVPR_2023_paper.html
|
http://openaccess.thecvf.com//content/CVPR2023/papers/Meng_On_Distillation_of_Guided_Diffusion_Models_CVPR_2023_paper.pdf
|
cvpr-2023-1
|
['text-guided-image-editing']
|
['computer-vision']
|
[ 3.71033996e-01 -1.81789312e-03 1.82707727e-01 -1.94952279e-01
-1.07260215e+00 -3.73764098e-01 8.74702573e-01 -1.96122065e-01
-5.66745460e-01 6.57431960e-01 1.38384908e-01 -1.64037332e-01
5.69121167e-02 -9.49403048e-01 -6.85781717e-01 -7.39072263e-01
8.05950686e-02 3.92748505e-01 4.22693402e-01 -8.88591781e-02
2.60329247e-01 4.53942299e-01 -1.32648027e+00 2.85822362e-01
9.90174174e-01 8.67813230e-01 3.59835982e-01 1.08013356e+00
-5.75703084e-02 1.09442973e+00 -5.40487826e-01 -1.05116494e-01
2.68786103e-01 -6.83105946e-01 -7.54413068e-01 2.21459851e-01
7.67066598e-01 -8.13600421e-01 -4.12834108e-01 1.10349572e+00
5.92712343e-01 1.32644489e-01 8.03212583e-01 -7.99766004e-01
-9.78722930e-01 4.46194798e-01 -7.02741206e-01 2.27816686e-01
4.53792922e-02 2.88842410e-01 5.59041977e-01 -9.30496275e-01
9.08906400e-01 1.34624755e+00 5.98579466e-01 7.97969222e-01
-1.74245739e+00 -6.07756317e-01 -1.70926955e-02 -1.24219730e-01
-1.20753860e+00 -6.08477771e-01 3.95176530e-01 -4.58839983e-01
1.08406627e+00 -1.20276347e-01 6.47790015e-01 1.04955769e+00
2.61611223e-01 5.59433997e-01 1.37299263e+00 -4.17685777e-01
3.07442665e-01 -5.51987477e-02 -2.23755941e-01 8.01369429e-01
-1.15934432e-01 2.36356124e-01 -3.31924945e-01 -1.40310571e-01
1.19699955e+00 -2.20964238e-01 -2.46358171e-01 8.56211260e-02
-1.35359144e+00 1.02470326e+00 4.98982221e-01 1.57937646e-01
-6.56229496e-01 6.58904552e-01 1.87956974e-01 3.18027675e-01
7.72687078e-01 1.73472658e-01 -6.65054610e-03 5.14433756e-02
-1.55997753e+00 5.51685333e-01 5.15142143e-01 8.37387145e-01
9.05552387e-01 2.97445595e-01 -3.44206035e-01 9.71616685e-01
1.26295313e-01 7.13385046e-01 4.62912083e-01 -1.56558216e+00
1.99913561e-01 -1.70483455e-01 1.26738921e-01 -8.19928944e-01
1.29312143e-01 -1.42861545e-01 -1.22847533e+00 6.17246807e-01
3.50872248e-01 -2.00925440e-01 -1.25902903e+00 1.64070463e+00
2.02038199e-01 3.06385964e-01 -8.80706161e-02 6.24485672e-01
3.25442702e-01 8.01135957e-01 7.07895309e-02 -6.57476485e-02
1.14869809e+00 -1.21088862e+00 -5.30473292e-01 -2.06506640e-01
6.19752347e-01 -9.55510259e-01 1.01745701e+00 5.25983751e-01
-1.46620739e+00 -5.58928072e-01 -1.05504060e+00 -2.73687840e-01
5.55083901e-03 -1.96375865e-02 5.30745804e-01 5.54502368e-01
-1.47368407e+00 8.57921958e-01 -1.02272737e+00 -2.50853181e-01
6.50433064e-01 1.69566333e-01 -2.94527948e-01 -4.39764738e-01
-7.98307955e-01 8.53294611e-01 -4.06663828e-02 -2.59064227e-01
-1.34652412e+00 -9.23976660e-01 -5.97769618e-01 -1.76603377e-01
-8.81805569e-02 -8.74985695e-01 1.10582662e+00 -9.50205505e-01
-1.57489610e+00 5.52021503e-01 -2.10957304e-01 -7.68277884e-01
7.53106892e-01 3.54148932e-02 -2.27252185e-01 4.21064019e-01
1.93395928e-01 1.23212278e+00 1.19649458e+00 -1.11819303e+00
-5.23724735e-01 -1.09852068e-02 8.05268735e-02 1.08472437e-01
-2.61082709e-01 -5.71281798e-02 -5.48196435e-01 -1.12354350e+00
3.32719311e-02 -8.33275080e-01 -4.68818247e-01 6.86638236e-01
-1.39701366e-01 2.56956398e-01 9.64471579e-01 -8.00591469e-01
9.38802242e-01 -1.99026501e+00 8.27778131e-02 5.98073192e-02
4.71388012e-01 1.72285452e-01 -4.92290974e-01 9.80600566e-02
1.35921955e-01 1.22272685e-01 -7.25745022e-01 -5.95520318e-01
-2.92955011e-01 3.30966681e-01 -2.77633488e-01 3.09793621e-01
1.32355630e-01 7.92903483e-01 -8.49432826e-01 -4.05214846e-01
2.70447105e-01 9.37114775e-01 -7.59981155e-01 9.31844208e-03
-2.09740222e-01 5.52939057e-01 8.51763785e-03 2.32860625e-01
7.92816281e-01 -2.51519561e-01 -1.58826441e-01 -1.76082402e-01
1.01862006e-01 3.97546776e-02 -1.11034346e+00 1.81531084e+00
-7.61426151e-01 8.19216549e-01 1.65338382e-01 -7.51065969e-01
7.49954641e-01 3.11749011e-01 2.97496885e-01 -7.55974948e-01
-2.73697764e-01 2.90275186e-01 -2.27580339e-01 -6.71341941e-02
4.43272412e-01 -2.51828402e-01 4.64214772e-01 8.84351790e-01
1.87162280e-01 -6.27436519e-01 2.95610040e-01 5.04554570e-01
1.21517432e+00 4.91002835e-02 -3.62170815e-01 -1.69163138e-01
3.58755171e-01 -1.38889328e-01 1.96921274e-01 8.46035480e-01
3.27215008e-02 9.97357309e-01 1.08582199e-01 -2.69911140e-01
-1.41637039e+00 -1.01572025e+00 2.47874800e-02 5.49469829e-01
-1.84132755e-01 -3.45182896e-01 -1.24464655e+00 -6.09546602e-01
-2.79963911e-01 7.95341194e-01 -5.96409738e-01 1.14376903e-01
-4.44281578e-01 -8.43447149e-01 9.04536068e-01 4.58007008e-01
9.09073114e-01 -8.84069920e-01 -2.44327664e-01 4.35140431e-01
-8.77493247e-02 -9.97171700e-01 -6.27382874e-01 -1.26151517e-01
-1.22232711e+00 -6.69876337e-01 -1.27763772e+00 -5.61347961e-01
9.80559230e-01 2.35156894e-01 1.08410180e+00 2.36898705e-01
-3.69736731e-01 3.13308001e-01 -3.18957232e-02 5.85358590e-02
-6.98102355e-01 -2.37903088e-01 -2.10531861e-01 -5.68375029e-02
-6.98726624e-02 -6.38011098e-01 -9.96954083e-01 1.96230903e-01
-1.18574941e+00 1.31397039e-01 4.70408380e-01 9.50517297e-01
7.78256059e-01 2.51437336e-01 4.68766958e-01 -9.82269168e-01
8.35993826e-01 -1.68746993e-01 -5.82971632e-01 -2.83395741e-02
-9.36530471e-01 1.40382648e-01 5.72198570e-01 -6.54090524e-01
-1.26949418e+00 1.11197658e-01 -4.78733629e-01 -5.43792844e-01
7.91117772e-02 1.71266466e-01 4.58638698e-01 -4.21432555e-01
1.00408125e+00 3.61389995e-01 1.49163082e-01 -4.03091013e-01
7.45015085e-01 4.02740240e-01 5.02848744e-01 -7.14006543e-01
8.28052521e-01 8.99461508e-01 -1.98970735e-01 -9.21269774e-01
-4.71419841e-01 9.05922428e-02 -4.74753231e-01 3.35471542e-03
9.80443537e-01 -1.01539397e+00 3.30409817e-02 8.24318767e-01
-1.11251152e+00 -7.86780417e-01 -5.38089156e-01 3.51565123e-01
-6.12730563e-01 5.48909307e-01 -1.02779818e+00 -4.47688907e-01
-3.89585733e-01 -1.32932973e+00 9.58503723e-01 -1.65962607e-01
-1.53004959e-01 -1.36825979e+00 -7.69592151e-02 1.49132475e-01
8.63499999e-01 6.04767017e-02 9.67971683e-01 1.90134466e-01
-6.93901777e-01 1.76208653e-02 -4.03037459e-01 8.17022383e-01
1.47045050e-02 -3.85127170e-03 -8.44914377e-01 -5.18717468e-01
-3.39964628e-02 -3.66439790e-01 1.13598943e+00 5.40211737e-01
1.09659493e+00 -1.26199976e-01 -6.74940720e-02 7.13667274e-01
1.40560055e+00 -9.25450772e-02 9.22844529e-01 5.26828021e-02
5.63208640e-01 3.22712511e-01 3.30165267e-01 1.02801181e-01
2.48291627e-01 5.04128218e-01 -4.14879173e-02 -3.93999666e-01
-8.98550034e-01 -1.83330759e-01 3.69173825e-01 7.78098702e-01
-1.05232984e-01 6.20247312e-02 -5.70403874e-01 6.73562586e-01
-1.53268349e+00 -1.04976416e+00 -2.12698787e-01 2.06262398e+00
1.05015934e+00 7.96256810e-02 -1.75652876e-01 1.00284092e-01
3.78424644e-01 1.92202538e-01 -6.05742037e-01 -2.78932869e-01
-1.36348009e-01 4.41495329e-01 5.07465541e-01 9.72960889e-01
-7.60698617e-01 1.07611144e+00 7.15173578e+00 1.20489669e+00
-1.19237018e+00 4.16090041e-01 8.92925084e-01 -1.89166501e-01
-5.18691361e-01 1.23789795e-01 -5.60565412e-01 3.84483933e-01
8.74065042e-01 -7.98136964e-02 8.43048930e-01 6.50451362e-01
3.12588483e-01 -3.26453567e-01 -9.73515332e-01 1.01441658e+00
1.27542829e-02 -1.62886822e+00 3.82251292e-01 2.59187102e-01
1.22108662e+00 1.62116379e-01 1.60900787e-01 -7.37326890e-02
6.00321114e-01 -1.10865152e+00 7.53271699e-01 4.86586571e-01
1.11216068e+00 -5.62724829e-01 1.82564303e-01 4.61898088e-01
-9.11372900e-01 2.28079513e-01 -6.31810129e-01 2.14309469e-01
2.87666619e-01 9.52738822e-01 -3.27488214e-01 8.97456333e-02
7.22614825e-01 6.19245350e-01 -3.63666981e-01 6.71396971e-01
-2.67723739e-01 5.84241748e-01 -3.59333634e-01 5.93726397e-01
2.79422641e-01 -2.18655795e-01 2.19131559e-01 1.12321532e+00
7.43127763e-01 1.29336417e-01 -2.70601958e-01 1.05194116e+00
-1.25053585e-01 -2.45258972e-01 -5.11440396e-01 1.15779527e-01
2.89919972e-01 1.14175761e+00 -7.50364125e-01 -7.85512269e-01
-3.40475827e-01 1.67512441e+00 3.07304859e-01 5.96000791e-01
-8.81429732e-01 -3.88078541e-01 3.53307754e-01 2.93527603e-01
4.65178460e-01 -4.93199259e-01 -1.94906771e-01 -1.13614392e+00
-3.35173756e-01 -9.91891503e-01 -8.02608579e-02 -9.60137367e-01
-1.30674481e+00 9.34144616e-01 -1.95885509e-01 -7.90953040e-01
-3.05440128e-01 -3.24400097e-01 -4.32489634e-01 1.36969781e+00
-1.62990987e+00 -1.11513305e+00 -3.76504779e-01 8.05299997e-01
6.40623927e-01 -3.62969819e-03 8.73287678e-01 5.46552658e-01
-8.78252387e-02 3.54007006e-01 2.30721533e-01 -4.22093570e-02
8.23378921e-01 -1.16344190e+00 8.11718941e-01 9.21190262e-01
1.51653871e-01 6.13552213e-01 4.42682266e-01 -7.34867811e-01
-1.02418995e+00 -1.19366241e+00 8.68282914e-01 -3.85609865e-01
4.01080638e-01 -1.06719151e-01 -8.68076384e-01 5.88757277e-01
3.82120579e-01 2.11924329e-01 1.20681390e-01 -5.04749775e-01
-4.12196696e-01 -5.47231548e-02 -1.32622051e+00 6.80033147e-01
1.04079163e+00 -7.06341147e-01 -6.86177015e-02 3.79093081e-01
3.59428734e-01 -4.10149485e-01 -9.06207800e-01 -5.46213351e-02
3.67927849e-01 -1.13485098e+00 1.08736598e+00 1.21087857e-01
5.60528815e-01 -2.97733843e-01 -1.10861063e-01 -1.41405261e+00
-4.27242547e-01 -6.93616390e-01 -1.14483677e-01 1.00243092e+00
3.64355385e-01 -5.79951882e-01 7.08704889e-01 4.24119264e-01
3.62048119e-01 -5.45233786e-01 -7.28199482e-01 -6.17006481e-01
3.62933367e-01 -3.94533038e-01 3.16795141e-01 8.36650848e-01
-8.53950799e-01 2.08965048e-01 -4.29189444e-01 -6.34966046e-02
9.28906441e-01 -3.93103600e-01 6.68885708e-01 -8.83013606e-01
-5.30351579e-01 -4.41367626e-01 9.48067755e-02 -1.64686072e+00
-1.04179762e-01 -7.35561550e-01 -7.21196905e-02 -1.69638193e+00
6.61507472e-02 -7.36709356e-01 8.44660699e-02 3.55797768e-01
-3.93640995e-02 8.17657351e-01 8.89562741e-02 5.23733258e-01
6.73229396e-02 4.22611505e-01 1.53340530e+00 -3.76250327e-01
3.33348177e-02 -3.92037988e-01 -6.68229580e-01 7.29380786e-01
4.34050083e-01 -6.38654768e-01 -7.53520906e-01 -8.96728158e-01
1.95417926e-03 2.15262193e-02 4.40975934e-01 -1.17382431e+00
4.02099431e-01 8.79419520e-02 5.04063070e-01 -2.72299707e-01
4.33864802e-01 -5.71653545e-01 1.63257167e-01 4.11471933e-01
-3.92683446e-01 -1.29527330e-01 1.16059653e-01 5.84690750e-01
-2.31609330e-01 -2.23288700e-01 1.12860167e+00 -2.77807862e-01
-5.52067220e-01 4.78903264e-01 -5.52547753e-01 -4.51474302e-02
8.68863463e-01 -1.88497335e-01 -1.94675013e-01 -7.22477734e-01
-7.86630392e-01 -1.59920350e-01 6.33598566e-01 1.47212923e-01
7.52392948e-01 -1.20746875e+00 -8.36635888e-01 3.25482965e-01
-4.28496271e-01 -6.83331862e-03 4.42417562e-01 6.99455440e-01
-8.84537697e-01 3.83941978e-02 -1.25639901e-01 -6.30764604e-01
-1.04643178e+00 3.25901866e-01 4.95486647e-01 -3.46970886e-01
-8.04569006e-01 9.57036912e-01 2.61927187e-01 -3.43752712e-01
-2.72305366e-02 -2.81420231e-01 3.10149699e-01 -3.20932269e-01
7.55366504e-01 3.98586452e-01 1.52562361e-03 -4.92309064e-01
1.74493138e-02 8.72586131e-01 -2.77015716e-01 -7.39383399e-01
1.21183670e+00 -1.59565732e-01 -1.51559740e-01 -1.30603090e-01
1.06623614e+00 9.39299017e-02 -1.79897797e+00 -1.57566145e-01
-3.68247002e-01 -5.15667200e-01 5.63194275e-01 -8.03700686e-01
-1.22560835e+00 1.19038415e+00 7.11310267e-01 -8.33360292e-03
1.24586380e+00 -3.52755129e-01 1.02817059e+00 5.71610220e-02
3.03512245e-01 -1.05103672e+00 2.24473551e-01 3.08170646e-01
9.35392499e-01 -9.56143796e-01 9.10223126e-02 -3.07451636e-01
-4.38581228e-01 9.23014581e-01 2.65849143e-01 -4.16250765e-01
7.67446339e-01 5.04134476e-01 3.09213877e-01 6.07033120e-03
-7.32289016e-01 1.16934158e-01 2.31825747e-02 8.82399678e-01
3.18400115e-01 -2.54661947e-01 -1.06514081e-01 -1.58144712e-01
9.82295573e-02 2.64667362e-01 4.66035962e-01 7.81961262e-01
-2.52873600e-01 -1.45282996e+00 -2.37563387e-01 3.34998757e-01
-3.78763914e-01 -4.48522866e-01 4.90445457e-02 4.25315917e-01
-6.77312762e-02 1.06810844e+00 2.03434452e-01 -1.02812380e-01
1.29668936e-01 -2.09292859e-01 6.40698195e-01 -4.93831307e-01
-4.07470703e-01 3.20948362e-01 -6.28440529e-02 -7.01723814e-01
-5.25061429e-01 -3.10705006e-01 -1.18987870e+00 -7.18236864e-01
-1.69393823e-01 -1.34776905e-01 7.60834992e-01 7.07349300e-01
5.89777827e-01 4.73731518e-01 2.91045129e-01 -1.12938070e+00
-5.52181959e-01 -8.96022558e-01 -6.16270781e-01 4.44197118e-01
2.07750663e-01 -2.71412879e-01 -3.62842202e-01 4.22019303e-01]
|
[11.273727416992188, -0.4743518829345703]
|
b79a5c60-2d06-4026-89be-71ecb57923b0
|
pids-joint-point-interaction-dimension-search
|
2211.15759
| null |
https://arxiv.org/abs/2211.15759v2
|
https://arxiv.org/pdf/2211.15759v2.pdf
|
PIDS: Joint Point Interaction-Dimension Search for 3D Point Cloud
|
The interaction and dimension of points are two important axes in designing point operators to serve hierarchical 3D models. Yet, these two axes are heterogeneous and challenging to fully explore. Existing works craft point operator under a single axis and reuse the crafted operator in all parts of 3D models. This overlooks the opportunity to better combine point interactions and dimensions by exploiting varying geometry/density of 3D point clouds. In this work, we establish PIDS, a novel paradigm to jointly explore point interactions and point dimensions to serve semantic segmentation on point cloud data. We establish a large search space to jointly consider versatile point interactions and point dimensions. This supports point operators with various geometry/density considerations. The enlarged search space with heterogeneous search components calls for a better ranking of candidate models. To achieve this, we improve the search space exploration by leveraging predictor-based Neural Architecture Search (NAS), and enhance the quality of prediction by assigning unique encoding to heterogeneous search components based on their priors. We thoroughly evaluate the networks crafted by PIDS on two semantic segmentation benchmarks, showing ~1% mIOU improvement on SemanticKITTI and S3DIS over state-of-the-art 3D models.
|
['Yiran Chen', 'Hai Li', 'Feng Yan', 'Mingyuan Ma', 'Tunhou Zhang']
|
2022-11-28
| null | null | null | null |
['robust-3d-semantic-segmentation']
|
['computer-vision']
|
[-2.09849700e-02 7.87808597e-02 -4.41621929e-01 -3.05653363e-01
-6.70913935e-01 -5.62470376e-01 5.01179039e-01 2.57626865e-02
-1.05255850e-01 -4.98957783e-02 -6.41425699e-02 -4.64491874e-01
-4.47676599e-01 -7.54799902e-01 -8.20296347e-01 -4.22157794e-01
-6.98205158e-02 1.12650132e+00 7.79512703e-01 -4.72567379e-01
5.36181808e-01 1.03925693e+00 -1.77256358e+00 1.93994626e-01
1.07562268e+00 1.20561802e+00 4.16069329e-01 2.40905389e-01
-7.28727043e-01 -1.25614345e-01 -4.87921268e-01 -2.09929749e-01
6.34059429e-01 1.96063623e-01 -8.13757837e-01 -1.95860490e-02
4.53737527e-01 -9.54008102e-02 5.77707849e-02 7.61559367e-01
5.15606046e-01 2.22422928e-01 6.49641395e-01 -1.27400756e+00
-5.18613398e-01 5.78309715e-01 -6.92170739e-01 9.25679877e-02
-9.70755517e-02 2.04577059e-01 1.28069770e+00 -8.40571821e-01
3.48264128e-01 1.44224274e+00 7.61019289e-01 5.85688770e-01
-1.11656797e+00 -5.44252276e-01 4.88753825e-01 8.96628127e-02
-1.27665591e+00 3.74870263e-02 1.10557818e+00 -2.93740630e-01
1.29857552e+00 5.39286494e-01 8.36235523e-01 9.11664307e-01
-1.16954945e-01 1.07057166e+00 4.87246454e-01 -2.30635330e-01
3.65892380e-01 3.24138142e-02 2.28926092e-01 4.70024675e-01
-2.86729224e-02 3.89502943e-02 -5.42278826e-01 -1.38075516e-01
9.97516930e-01 2.15872586e-01 1.50493190e-01 -7.51888156e-01
-9.82141018e-01 7.95028329e-01 7.64640570e-01 -1.41376480e-02
-2.45958552e-01 1.28486454e-01 2.84591168e-01 -9.94225889e-02
4.21441048e-01 1.06472552e+00 -8.27313960e-01 -1.71389729e-01
-1.04117084e+00 3.82042944e-01 6.16055071e-01 1.24748755e+00
7.53541291e-01 -2.08711624e-01 -1.81897804e-01 1.12502646e+00
3.57924968e-01 4.39592153e-01 1.61110200e-02 -9.98435736e-01
6.02529585e-01 1.18363047e+00 -1.01820022e-01 -7.73510993e-01
-6.23460472e-01 -6.71556413e-01 -5.97011268e-01 1.98656276e-01
-4.38669883e-03 4.70875651e-01 -1.35220575e+00 1.26579261e+00
5.46265185e-01 7.02467114e-02 -4.31614518e-01 1.02580249e+00
7.51186788e-01 5.30224264e-01 3.82451378e-02 5.30841887e-01
1.34789193e+00 -1.29468679e+00 5.46678379e-02 -3.61902088e-01
5.65026939e-01 -4.42593008e-01 1.30097556e+00 4.80823308e-01
-1.22916305e+00 -6.29612505e-01 -1.00955379e+00 -2.66784757e-01
-3.54000807e-01 -4.33268063e-02 8.42452526e-01 6.37417793e-01
-1.04454195e+00 6.70569301e-01 -1.01291347e+00 2.67944690e-02
7.42601097e-01 6.83714926e-01 1.79591537e-01 1.92915782e-01
-9.97770071e-01 7.49222517e-01 1.56082973e-01 -2.84583215e-02
-3.45724940e-01 -1.03275359e+00 -5.60272157e-01 1.59318179e-01
4.14122909e-01 -1.01203465e+00 1.03059065e+00 -1.41730532e-01
-1.28594363e+00 7.02833831e-01 -3.83374915e-02 -4.52391237e-01
4.97557908e-01 -3.87624025e-01 -1.31630123e-01 1.45829588e-01
1.72564834e-01 1.17337227e+00 6.59546912e-01 -1.45597136e+00
-7.16377616e-01 -6.56593442e-01 1.75857037e-01 3.93411636e-01
-1.91233203e-01 -3.33787024e-01 -1.06251526e+00 -5.03930449e-01
6.73565209e-01 -9.84096348e-01 -4.87944037e-01 1.29866032e-02
-4.79354769e-01 -4.54154134e-01 8.14757109e-01 -1.40711501e-01
1.12080193e+00 -2.08056450e+00 3.85279596e-01 3.96299332e-01
3.54801893e-01 -1.33616589e-02 8.21325835e-03 2.02487126e-01
2.16029316e-01 4.87053335e-01 -1.40424203e-02 -7.34479368e-01
4.33090150e-01 3.66788924e-01 -3.33123326e-01 3.54083888e-02
2.95305967e-01 9.74443257e-01 -6.04898810e-01 -5.01904368e-01
3.18529278e-01 4.66156185e-01 -1.01977134e+00 -1.75964251e-01
-7.09687412e-01 2.72101045e-01 -9.83494341e-01 1.02975523e+00
8.91255200e-01 -6.10116661e-01 -4.38300997e-01 -2.69075036e-01
-4.87626307e-02 3.32248867e-01 -1.09863210e+00 2.09115982e+00
-3.27666879e-01 -7.52741180e-04 -1.34366378e-01 -5.43379605e-01
1.32856786e+00 -1.95396468e-01 7.57867277e-01 -6.40192807e-01
5.96651882e-02 4.56233233e-01 -2.84758091e-01 -1.47391692e-01
7.72681117e-01 8.77182111e-02 -1.29948169e-01 1.54742658e-01
-4.49629314e-02 -6.74334407e-01 -2.39396483e-01 -5.45458309e-02
9.89334583e-01 2.04082742e-01 -3.79418612e-01 -2.39794433e-01
2.16743901e-01 2.54986316e-01 3.97828162e-01 8.62552166e-01
2.94358563e-02 7.50058174e-01 4.08571780e-01 -5.70738971e-01
-1.00739861e+00 -1.05850947e+00 -4.13343579e-01 1.24295151e+00
7.30295777e-01 -3.11795324e-01 -4.66986537e-01 -8.44143391e-01
1.96994111e-01 8.54523420e-01 -5.07901073e-01 -1.90004975e-01
-6.89514637e-01 -6.49435282e-01 3.99305880e-01 7.17019975e-01
1.61598727e-01 -8.17906022e-01 -7.80378520e-01 7.66897202e-02
2.64647633e-01 -8.73302519e-01 -2.49464020e-01 7.36649036e-01
-1.16243482e+00 -6.61959887e-01 -6.55567110e-01 -5.60828567e-01
4.33954149e-01 2.67739862e-01 1.55722046e+00 1.36876404e-01
-3.59145701e-02 1.24463268e-01 -5.89606225e-01 -4.67669874e-01
-4.75159697e-02 7.31194854e-01 -2.07125589e-01 -6.52483702e-01
3.86581123e-01 -7.39786386e-01 -8.92566979e-01 7.27007985e-01
-8.47157836e-01 2.19605714e-01 6.20209634e-01 6.73539639e-01
9.49854672e-01 8.42763111e-02 1.35302410e-01 -6.58755064e-01
2.13298261e-01 -4.43178713e-01 -5.45130014e-01 2.61130273e-01
-5.22405148e-01 3.64743918e-01 2.38637492e-01 -2.66023934e-01
-5.17062008e-01 -1.79514140e-02 -4.43976074e-01 -1.04040194e+00
-2.10758850e-01 6.65145442e-02 -2.13029742e-01 -1.10918775e-01
6.17792070e-01 -1.45169005e-01 -2.30521724e-01 -8.78858507e-01
5.58007598e-01 3.38145286e-01 1.50408730e-01 -9.76567388e-01
4.82553899e-01 4.92817730e-01 1.05591431e-01 -4.50078815e-01
-6.18571103e-01 -7.14562654e-01 -6.80243433e-01 1.39033929e-01
8.83967638e-01 -6.34332001e-01 -5.08923709e-01 1.77792445e-01
-1.15969229e+00 -1.44708127e-01 -4.16045010e-01 8.75076875e-02
-6.45636737e-01 4.82375324e-02 -2.80044049e-01 -6.48544192e-01
-3.00956219e-01 -1.73526049e+00 1.83190167e+00 9.58104208e-02
-1.17501311e-01 -7.52994716e-01 -2.26062208e-01 4.96953964e-01
2.32669115e-01 8.25150609e-02 1.08003485e+00 -7.97302723e-01
-9.58132386e-01 -9.65303853e-02 -3.44542682e-01 -1.57028794e-01
-1.54607967e-01 -1.53767988e-01 -7.65862226e-01 5.87432310e-02
3.90343852e-02 -5.40616699e-02 9.37373877e-01 5.04420102e-01
1.60836899e+00 1.44135088e-01 -7.32710779e-01 1.06167006e+00
1.17131400e+00 2.65380263e-01 3.67183328e-01 4.69591290e-01
9.45500851e-01 5.59521973e-01 5.82765341e-01 4.60407704e-01
4.77814585e-01 8.90128732e-01 7.89937735e-01 -1.64849803e-01
-4.98923808e-02 -3.68311763e-01 -4.26246375e-01 5.25122583e-01
-4.06095944e-03 -3.33323717e-01 -1.30607903e+00 4.49709356e-01
-1.69519567e+00 -3.94687593e-01 -1.01931449e-02 1.64868057e+00
6.20761454e-01 4.93030727e-01 1.52025729e-01 5.60566830e-03
5.17667949e-01 1.13584235e-01 -9.55547512e-01 -2.38220572e-01
-5.19810207e-02 2.82917649e-01 7.22061515e-01 1.56217799e-01
-9.76405621e-01 1.01586473e+00 5.49184036e+00 1.23642826e+00
-1.20143580e+00 -1.31184384e-01 7.66492367e-01 -3.81360501e-01
-7.58921742e-01 -4.76457216e-02 -1.25213194e+00 3.45226347e-01
5.34478009e-01 5.26960433e-01 2.61325806e-01 1.10722530e+00
-6.34904951e-02 2.51538664e-01 -1.19326532e+00 1.05104828e+00
-2.66008168e-01 -1.56907558e+00 2.63600469e-01 2.15066314e-01
6.04777992e-01 4.14988428e-01 1.78363368e-01 3.79892200e-01
1.75723314e-01 -1.10579789e+00 9.67748702e-01 3.25047851e-01
5.79805315e-01 -7.31334150e-01 2.92121947e-01 3.07577759e-01
-1.07262003e+00 -1.74454764e-01 -2.88969189e-01 4.35136676e-01
3.59620154e-01 4.05528218e-01 -8.12567055e-01 5.79437613e-01
1.02841580e+00 5.39589643e-01 -6.50252044e-01 1.11819530e+00
2.59962022e-01 3.12415093e-01 -7.86472440e-01 -1.21261381e-01
6.75943196e-01 -1.70008734e-01 7.46399283e-01 8.41499388e-01
4.47364360e-01 -1.98377013e-01 2.19659403e-01 1.23563099e+00
1.97886333e-01 -1.63136408e-01 -3.03946286e-01 1.33014917e-01
7.44568765e-01 8.37636471e-01 -1.11907208e+00 7.64080733e-02
-9.82038677e-02 6.11066401e-01 1.77186519e-01 1.38343811e-01
-1.02137768e+00 -5.11958115e-02 6.46146059e-01 3.41505468e-01
6.61338329e-01 -4.07956779e-01 -1.06349981e+00 -7.68385887e-01
5.79731986e-02 -6.33951545e-01 2.38903001e-01 -6.29722357e-01
-1.32051635e+00 8.89992237e-01 2.08301872e-01 -1.30657816e+00
1.19114444e-01 -5.94015956e-01 -3.17144752e-01 9.15775836e-01
-1.35722387e+00 -1.38758087e+00 -1.50216565e-01 1.92683578e-01
9.02872443e-01 -1.03113599e-01 4.46132243e-01 2.44765162e-01
-3.04423660e-01 5.32638252e-01 -2.17804775e-01 -5.14447510e-01
1.69045180e-01 -1.38698673e+00 1.01611292e+00 2.29892448e-01
2.14952320e-01 6.10449374e-01 4.98349994e-01 -7.54442155e-01
-1.45740163e+00 -8.21772277e-01 3.16392750e-01 -8.57778311e-01
3.66381407e-01 -4.48988974e-01 -8.91276598e-01 1.72923446e-01
-4.05707002e-01 -1.63616687e-01 4.76127326e-01 5.20136356e-01
-3.13842386e-01 1.47598699e-01 -1.06102812e+00 6.70933843e-01
1.62214994e+00 -3.04497927e-01 -4.74970967e-01 1.68026507e-01
1.39200175e+00 -6.42697811e-01 -8.86228144e-01 8.77223015e-01
3.75797898e-01 -9.55692708e-01 1.48842978e+00 -4.58301157e-01
3.40200782e-01 -2.98239797e-01 -3.78797561e-01 -1.03376627e+00
-3.55277508e-01 -4.13334459e-01 -2.69044966e-01 8.55568290e-01
6.47358835e-01 -4.39120501e-01 1.24329269e+00 7.66233802e-01
-7.31947780e-01 -1.43362832e+00 -9.39580798e-01 -6.28381550e-01
1.94883525e-01 -7.78823614e-01 1.33469307e+00 6.44449413e-01
-5.38046181e-01 1.09182857e-01 1.43201172e-01 2.32574090e-01
2.61973143e-01 3.35873216e-01 6.64038539e-01 -1.39648461e+00
-3.80962193e-01 -1.02943873e+00 -3.08385849e-01 -1.73761189e+00
-1.81715548e-01 -8.96607041e-01 -1.44247442e-01 -1.54490733e+00
-3.48110765e-01 -1.36810625e+00 -2.28325963e-01 3.84458929e-01
-3.92649136e-02 7.70126060e-02 1.93547323e-01 4.72882926e-01
-6.29642785e-01 7.63483405e-01 1.49507511e+00 -1.42824665e-01
-7.03815758e-01 1.75454780e-01 -7.94834256e-01 6.75585091e-01
6.57037079e-01 -2.53516763e-01 -6.98882818e-01 -1.05105162e+00
3.42733622e-01 2.91515496e-02 4.19326782e-01 -1.09198427e+00
3.05850744e-01 -2.33164698e-01 2.89106339e-01 -1.34710419e+00
8.07717144e-01 -9.75147843e-01 7.32792690e-02 -8.49340186e-02
-2.40339622e-01 2.58204900e-02 3.86120439e-01 5.54479659e-01
5.25277704e-02 -1.24118619e-01 4.38901544e-01 -2.68786043e-01
-8.58114421e-01 4.92243648e-01 6.09163761e-01 -1.36651993e-01
7.47165859e-01 -7.93473065e-01 -2.18832958e-02 3.45368117e-01
-7.72834420e-01 5.56536913e-01 7.41981149e-01 7.09662616e-01
6.14395142e-01 -9.91292834e-01 -9.05883014e-02 4.30969357e-01
3.50645408e-02 9.00485158e-01 4.62940305e-01 5.97770572e-01
-6.11619771e-01 4.00283694e-01 -2.11680913e-03 -1.19610345e+00
-7.15437472e-01 3.86703938e-01 2.23015875e-01 -2.08200410e-01
-7.44862437e-01 1.33944976e+00 2.03278899e-01 -7.02223003e-01
3.21886331e-01 -5.81648052e-01 -1.20964013e-01 -6.35423958e-02
-9.88399386e-02 3.28725755e-01 3.26261371e-01 -2.76530296e-01
-4.12448883e-01 7.49240994e-01 -3.00641060e-02 1.53729051e-01
1.54343152e+00 -1.96549334e-02 1.97638571e-01 2.72225738e-01
1.00835991e+00 -1.39845639e-01 -1.50300026e+00 1.79786328e-02
9.51501504e-02 -4.99668658e-01 1.88099727e-01 -8.30451131e-01
-9.05960858e-01 9.06432629e-01 3.97826612e-01 4.00605500e-01
9.41370964e-01 4.46454227e-01 9.71972048e-01 1.41595840e-01
5.63167751e-01 -1.10085285e+00 -5.00184409e-02 5.71307778e-01
9.02411699e-01 -1.00528300e+00 -5.48582785e-02 -7.08267748e-01
-6.25381410e-01 9.09248292e-01 7.45194614e-01 -6.27092347e-02
7.48018146e-01 7.74008706e-02 -2.68819273e-01 -6.24872148e-01
-6.20959520e-01 -1.14548966e-01 7.22883046e-01 3.96535814e-01
-3.84432003e-02 1.14631113e-02 2.48876289e-01 7.83594728e-01
-5.62158525e-01 -3.83012563e-01 -3.86914641e-01 7.83789515e-01
-4.55470860e-01 -1.12564635e+00 -3.68385613e-01 7.23472595e-01
1.75811711e-03 -2.61796974e-02 -2.92758197e-01 6.71387732e-01
3.44829857e-01 4.58192021e-01 3.20882052e-01 -5.85217714e-01
4.41253901e-01 -1.01226516e-01 3.22579712e-01 -6.37318969e-01
-5.89214861e-01 1.62891820e-01 -1.46364033e-01 -5.35901070e-01
4.41506272e-04 -4.59931731e-01 -1.34218335e+00 -5.82995303e-02
-5.23037016e-01 4.37897518e-02 9.74778116e-01 7.50020385e-01
7.95978963e-01 6.27282023e-01 4.89881426e-01 -1.26430118e+00
-6.45321965e-01 -4.98437643e-01 -3.51598322e-01 6.38092831e-02
1.37241557e-02 -1.04487765e+00 -1.77185148e-01 -5.46995461e-01]
|
[7.964248180389404, -3.3733766078948975]
|
e8a6149a-f192-46b7-ade1-a064fea64449
|
anticipative-feature-fusion-transformer-for
|
2210.12649
| null |
https://arxiv.org/abs/2210.12649v1
|
https://arxiv.org/pdf/2210.12649v1.pdf
|
Anticipative Feature Fusion Transformer for Multi-Modal Action Anticipation
|
Although human action anticipation is a task which is inherently multi-modal, state-of-the-art methods on well known action anticipation datasets leverage this data by applying ensemble methods and averaging scores of unimodal anticipation networks. In this work we introduce transformer based modality fusion techniques, which unify multi-modal data at an early stage. Our Anticipative Feature Fusion Transformer (AFFT) proves to be superior to popular score fusion approaches and presents state-of-the-art results outperforming previous methods on EpicKitchens-100 and EGTEA Gaze+. Our model is easily extensible and allows for adding new modalities without architectural changes. Consequently, we extracted audio features on EpicKitchens-100 which we add to the set of commonly used features in the community.
|
['Jürgen Beyerer', 'Rainer Stiefelhagen', 'Michael Voit', 'David Schneider', 'Zeyun Zhong']
|
2022-10-23
| null | null | null | null |
['action-anticipation']
|
['computer-vision']
|
[ 4.72897142e-01 -6.62235618e-02 -4.28924449e-02 -2.44607046e-01
-1.19924915e+00 -2.73437202e-01 8.05620372e-01 6.44111708e-02
-3.85550052e-01 5.59000552e-01 9.77184355e-01 4.57309753e-01
-4.05416191e-01 -3.04062814e-01 -3.73988837e-01 -5.64437807e-01
-4.16766316e-01 2.71532923e-01 1.42282531e-01 -5.46885490e-01
8.93341377e-02 -2.11040542e-01 -2.19363999e+00 1.12988997e+00
3.39043349e-01 1.22863555e+00 -4.74449486e-01 8.69836032e-01
4.29377198e-01 1.21047938e+00 -1.80099219e-01 -3.47075373e-01
1.46855041e-01 -2.34596163e-01 -8.00806224e-01 -5.01298726e-01
5.98738134e-01 -1.68990165e-01 -3.23287934e-01 4.30363923e-01
8.79922569e-01 2.06414849e-01 4.53511417e-01 -1.78135574e+00
-1.89166501e-01 1.01634634e+00 -3.36786032e-01 4.05715033e-02
1.20441926e+00 5.40829182e-01 1.23486423e+00 -6.49547040e-01
6.81509435e-01 1.26455891e+00 1.01325417e+00 4.96144146e-01
-1.17675495e+00 -6.35061681e-01 1.78307176e-01 7.71854162e-01
-1.00205183e+00 -6.62843406e-01 9.29340303e-01 -4.96998638e-01
1.41052496e+00 3.62830967e-01 1.04844403e+00 1.77486396e+00
1.46240056e-01 1.24554479e+00 1.08261096e+00 -4.11736727e-01
2.84757286e-01 -5.45850813e-01 6.52896464e-02 2.75956780e-01
-5.60661435e-01 2.36462265e-01 -1.68703902e+00 -2.27496460e-01
-7.12526292e-02 6.54651970e-02 3.99039872e-02 -2.07254738e-01
-1.81549001e+00 3.69983584e-01 1.10075884e-01 1.56341121e-01
-9.93220091e-01 5.61659694e-01 8.20448756e-01 3.24590176e-01
3.11287135e-01 1.58459425e-01 -3.67530376e-01 -1.10078657e+00
-1.00334799e+00 6.71960056e-01 5.15655637e-01 4.71816152e-01
3.09185863e-01 -2.19992384e-01 -5.46169043e-01 4.42379445e-01
3.72124553e-01 4.65699434e-01 3.50725710e-01 -1.56332946e+00
4.75851685e-01 7.70447493e-01 -2.39151455e-02 -6.14289105e-01
-6.36993766e-01 1.44208685e-01 -5.63328862e-01 3.67368370e-01
2.92931199e-01 -7.87444487e-02 -8.29944372e-01 2.06572938e+00
1.56029895e-01 2.58873940e-01 2.46280700e-01 6.74710453e-01
9.36314523e-01 8.19331482e-02 4.37411129e-01 -1.47307798e-01
1.24933827e+00 -7.13158369e-01 -7.28705525e-01 -2.13219553e-01
4.82188880e-01 -6.14322901e-01 8.88966084e-01 7.54721761e-01
-1.07432580e+00 -4.45339859e-01 -8.92378032e-01 1.77275240e-01
-2.81594902e-01 -4.90409508e-02 1.04327238e+00 3.69734436e-01
-1.07615709e+00 6.50191545e-01 -9.63608265e-01 -5.13541639e-01
4.76589948e-01 4.74644631e-01 -8.06113780e-01 7.42209777e-02
-1.34378850e+00 1.02694869e+00 3.98557991e-01 -8.48592296e-02
-8.60323370e-01 -7.70478547e-01 -9.25326824e-01 -1.43902540e-01
4.15774465e-01 -8.81309092e-01 1.55306399e+00 -7.37983763e-01
-1.69711137e+00 6.29821181e-01 -7.33699426e-02 -7.48109818e-01
4.85854387e-01 -6.31439865e-01 -8.19801569e-01 2.60650963e-01
-1.04249723e-01 1.06681120e+00 7.23800302e-01 -5.48232853e-01
-6.19277894e-01 -1.30057573e-01 1.42226309e-01 2.97173321e-01
-3.89494509e-01 1.14631139e-01 3.51133376e-01 -4.07914221e-01
-1.49394229e-01 -1.01662326e+00 4.71255779e-02 -2.04697207e-01
-3.71667624e-01 -5.62272847e-01 7.78810441e-01 -2.94117957e-01
1.48936260e+00 -2.23965693e+00 4.41082746e-01 -1.05219316e-02
2.21870944e-01 4.33236994e-02 -4.23002601e-01 9.56227422e-01
-3.79419953e-01 -2.53284961e-01 1.51069388e-01 -6.62668884e-01
4.82209593e-01 -6.48986176e-03 -2.16481745e-01 1.99381366e-01
1.82565480e-01 8.96669984e-01 -9.16466117e-01 -6.62427783e-01
4.96648341e-01 4.83648628e-01 -5.48121333e-01 -9.15444121e-02
-2.15578720e-01 2.92350352e-01 -2.57181793e-01 1.03551137e+00
-4.68690842e-02 1.30185738e-01 -4.80741151e-02 -2.37579271e-01
-1.47833213e-01 2.45802939e-01 -1.06341195e+00 2.37747812e+00
4.88599122e-04 7.69662142e-01 -6.32870674e-01 -5.00324011e-01
4.63079393e-01 6.65578604e-01 1.02455628e+00 -3.50300461e-01
3.19116890e-01 -6.55688047e-02 -2.99386382e-02 -7.40061462e-01
7.35479653e-01 1.28624856e-01 -4.71487999e-01 4.77424473e-01
3.45634788e-01 1.56942829e-01 3.60813826e-01 3.39048386e-01
1.64682674e+00 5.58275342e-01 3.14164281e-01 3.11329156e-01
2.77584285e-01 -1.00555070e-01 5.68000674e-01 5.48502386e-01
-5.96104085e-01 5.28768837e-01 4.99062121e-01 -3.10258865e-01
-4.57443923e-01 -1.19324446e+00 3.63631278e-01 1.55510724e+00
-4.53722000e-01 -1.08865285e+00 -4.84651476e-01 -5.01576841e-01
2.57360414e-02 6.66169941e-01 -1.06604719e+00 -1.35875240e-01
-2.46092737e-01 -2.95886308e-01 8.14961553e-01 8.75547409e-01
4.65707034e-01 -1.27270591e+00 -1.07752132e+00 1.78309470e-01
-7.32414842e-01 -9.41680729e-01 -2.27893703e-02 5.40076569e-02
-4.96287882e-01 -1.21162808e+00 -4.88065898e-01 2.34122127e-01
-2.28317425e-01 -2.24863783e-01 1.23908472e+00 -6.42657220e-01
1.86458807e-02 9.28326666e-01 -5.98173618e-01 -7.97838926e-01
-1.71379924e-01 -3.01325694e-02 3.58406097e-01 5.31350151e-02
6.52054131e-01 -9.84253526e-01 -5.78876972e-01 -6.07819110e-02
-6.49621665e-01 1.59912765e-01 4.28739339e-01 5.30450225e-01
3.70491773e-01 -5.11081636e-01 5.53604960e-01 -1.59726545e-01
6.40915394e-01 -3.57542038e-01 2.02688619e-01 1.16968542e-01
-2.04917863e-01 -4.22001742e-02 -9.59588140e-02 -5.50096929e-01
-9.32658911e-01 4.53481793e-01 2.70261280e-02 -6.59692466e-01
-3.01054984e-01 6.90563679e-01 1.97858855e-01 8.61913487e-02
8.32221091e-01 -1.86471760e-01 3.47966962e-02 -3.08601201e-01
5.39925814e-01 4.79012579e-01 8.54509473e-01 -1.78268313e-01
2.04771459e-01 4.50765729e-01 1.61359414e-01 -3.24704200e-01
-4.97047812e-01 -3.78147632e-01 -6.43973708e-01 -9.82830107e-01
8.15450907e-01 -1.12242734e+00 -1.42069709e+00 8.52754116e-01
-8.96855474e-01 -2.44477347e-01 -6.34725749e-01 5.76626956e-01
-9.88367021e-01 1.57246783e-01 -2.79211968e-01 -1.13849914e+00
-2.57682800e-01 -6.95860147e-01 1.21421635e+00 1.53829679e-01
-1.01166856e+00 -5.31048894e-01 6.33308470e-01 4.42369998e-01
3.75145018e-01 9.36716139e-01 1.01009369e-01 -6.09349787e-01
-1.92648724e-01 -3.10489386e-01 4.27649438e-01 8.30971003e-02
-3.85681331e-01 2.24656731e-01 -1.20992184e+00 1.11503527e-01
-6.02979243e-01 -7.85889924e-01 1.35352778e+00 3.78281146e-01
6.28373742e-01 2.01114595e-01 -3.70059490e-01 -7.14921281e-02
8.02589357e-01 -1.44701287e-01 7.32847273e-01 4.36207741e-01
4.38746005e-01 5.22437334e-01 7.44332850e-01 8.84845555e-01
8.32745433e-01 5.98499119e-01 6.74412012e-01 4.83823270e-01
6.25664666e-02 -1.16351075e-01 9.97895658e-01 6.21245503e-01
-6.94172800e-01 -2.71568865e-01 -9.71317530e-01 6.61701679e-01
-2.45073438e+00 -1.59560037e+00 -9.79314893e-02 1.74030983e+00
7.11475849e-01 9.62328613e-02 5.54433227e-01 6.22589231e-01
1.95511743e-01 4.77044553e-01 -3.98409009e-01 -3.22233200e-01
-2.92308688e-01 1.57138571e-01 -5.05867228e-03 -1.07075624e-01
-1.39291465e+00 4.28125829e-01 7.16169786e+00 7.14717627e-01
-7.38358140e-01 1.29369766e-01 -1.04536958e-01 -6.98802710e-01
-1.74118597e-02 -4.57553864e-02 -3.05084199e-01 4.31300759e-01
1.21118736e+00 -8.97117779e-02 3.49162161e-01 5.77131867e-01
7.00014755e-02 -6.07160985e-01 -1.55155182e+00 1.13710129e+00
9.01328027e-02 -1.10370588e+00 -5.56253374e-01 -3.35043073e-01
4.93396789e-01 5.14238775e-01 1.59321472e-01 4.30974424e-01
5.23962379e-01 -8.37043703e-01 8.95202458e-01 1.27592540e+00
6.71107650e-01 -5.28393567e-01 6.55478299e-01 6.95556626e-02
-1.15536892e+00 -3.26522291e-01 5.08710742e-01 -5.11402726e-01
5.88780224e-01 4.24356997e-01 -2.59823143e-01 7.55910456e-01
9.44341302e-01 1.26465118e+00 -7.47137904e-01 8.80234241e-01
-1.44151628e-01 3.49143684e-01 -6.03366196e-01 4.71757688e-02
1.08620219e-01 7.06694782e-01 7.65774608e-01 1.04811454e+00
4.95516837e-01 -1.79613695e-01 -5.15764095e-02 2.61791199e-01
2.20439449e-01 -3.54804307e-01 -6.62522435e-01 -3.57164770e-01
2.10063532e-01 1.20264578e+00 -8.63839835e-02 -3.38362724e-01
-2.00295463e-01 6.20728493e-01 -1.39781889e-02 5.36291897e-02
-8.34782362e-01 -9.41658765e-02 8.66726875e-01 -4.02667642e-01
1.66432440e-01 1.63821429e-01 -7.99331814e-02 -1.04887223e+00
-5.83628416e-02 -8.97937596e-01 7.26823688e-01 -1.18835318e+00
-1.20530283e+00 4.13270682e-01 3.41084957e-01 -1.80824959e+00
-1.00870180e+00 -2.58746862e-01 -7.20388710e-01 2.95121282e-01
-9.66216087e-01 -1.55329025e+00 -4.61757839e-01 9.06361818e-01
2.80585766e-01 -3.10620189e-01 1.26197755e+00 1.45343393e-01
-2.56041735e-01 3.45443785e-01 -2.91009754e-01 -3.30009222e-01
1.01232243e+00 -1.07889009e+00 1.38958991e-01 6.37811363e-01
7.40269125e-02 1.89965695e-01 9.94979203e-01 -3.55965346e-01
-1.19096553e+00 -5.74515998e-01 9.91056442e-01 -8.55696082e-01
7.91230083e-01 -9.96472612e-02 -4.14128423e-01 7.53399014e-01
8.41612637e-01 -2.62252539e-01 1.04949546e+00 4.52242404e-01
-4.33028758e-01 -5.38148023e-02 -9.50286269e-01 4.79314655e-01
1.34251225e+00 -6.13293648e-01 -8.15105438e-01 1.77730471e-01
4.43053007e-01 -4.61218923e-01 -1.12199187e+00 6.47107780e-01
1.21460187e+00 -1.43680859e+00 9.54665959e-01 -4.97610420e-01
7.60593414e-01 -2.71039963e-01 -3.91969353e-01 -1.41147554e+00
-1.61350444e-01 -9.71442282e-01 -5.56693137e-01 1.24508750e+00
9.79999900e-02 -4.53478068e-01 4.54146922e-01 5.70082605e-01
-1.89663216e-01 -6.19966686e-01 -1.31236732e+00 -4.87435251e-01
-5.79111159e-01 -1.11333740e+00 6.13343120e-01 5.72242379e-01
7.22895682e-01 1.20365560e-01 -6.13039970e-01 -5.02530456e-01
5.80085337e-01 -1.51832864e-01 7.31003582e-01 -1.44608712e+00
-2.27512360e-01 -5.11359632e-01 -7.60944128e-01 -1.16013870e-01
1.80964261e-01 -6.30634427e-01 -1.31316438e-01 -1.33094621e+00
1.70059785e-01 3.10519248e-01 -9.51381683e-01 1.12543845e+00
2.57299781e-01 4.57759559e-01 5.52566886e-01 2.83550341e-02
-1.25789535e+00 7.25672722e-01 7.78443038e-01 -1.90724984e-01
-1.15603417e-01 -1.34234071e-01 -6.75137103e-01 8.70660722e-01
6.46445096e-01 -3.58471125e-01 -3.66718262e-01 -1.34297788e-01
5.46027303e-01 6.64244443e-02 7.61147439e-01 -1.53735948e+00
3.98469210e-01 7.28623644e-02 2.23484948e-01 -9.84705627e-01
9.16191161e-01 -4.67928708e-01 5.85033596e-01 1.72814399e-01
-5.08435786e-01 1.64655857e-02 2.65833378e-01 6.52736425e-01
-3.06292236e-01 4.31285292e-01 4.43329737e-02 1.36819601e-01
-1.04925597e+00 -1.38634115e-01 -6.25026226e-01 -8.86833444e-02
1.00930119e+00 -2.59619504e-01 -4.83541042e-01 -5.57342589e-01
-9.52182889e-01 3.65049064e-01 1.34333625e-01 7.69772351e-01
5.34102261e-01 -1.76623881e+00 -7.32727468e-01 -1.38501942e-01
5.03823876e-01 -6.25571609e-01 7.72962034e-01 1.46672451e+00
3.37143809e-01 2.04122528e-01 -6.65764213e-01 -6.41520441e-01
-1.71656728e+00 2.33000323e-01 -1.56906903e-01 -3.51800263e-01
-4.07034189e-01 7.18015492e-01 -6.90848112e-01 -1.47318020e-01
3.39326233e-01 -2.30540827e-01 -4.11268592e-01 6.17005646e-01
5.00770926e-01 6.50770843e-01 -2.20439941e-01 -6.75211132e-01
-7.70765603e-01 3.12584639e-01 2.19862595e-01 -4.70259696e-01
1.56529796e+00 7.20088258e-02 1.24515101e-01 8.75592709e-01
7.40745962e-01 -5.01969814e-01 -1.11977589e+00 -1.74529478e-01
-1.93361938e-01 -2.11722806e-01 -4.35938686e-02 -1.21347690e+00
-6.56318724e-01 8.48601818e-01 9.20696318e-01 8.56389254e-02
1.56889629e+00 -5.57944626e-02 7.09643960e-01 5.04171133e-01
3.88765931e-01 -1.48617554e+00 1.99327365e-01 4.80106235e-01
1.02262986e+00 -1.35183501e+00 7.86842406e-02 1.79610938e-01
-1.00593889e+00 9.65987861e-01 3.31585497e-01 -7.87855871e-03
4.76442844e-01 2.53384352e-01 -8.81304964e-02 -1.89201728e-01
-1.35535097e+00 -4.71093088e-01 3.16923708e-01 6.79446816e-01
4.34677422e-01 2.64734566e-01 -2.22066671e-01 8.57999980e-01
-1.89947978e-01 6.42852187e-01 3.17435741e-01 1.14978778e+00
-7.53443986e-02 -1.02324784e+00 -2.68756002e-01 4.41455662e-01
-3.78895700e-01 3.35578807e-02 -5.14184594e-01 6.23714447e-01
3.11480463e-01 1.13074589e+00 -2.12220445e-01 -1.09172678e+00
4.64863509e-01 5.42426527e-01 5.94004810e-01 1.30506577e-02
-9.45672572e-01 -1.80038363e-01 5.64683259e-01 -1.23574114e+00
-1.26419127e+00 -1.42985654e+00 -9.22360420e-01 -3.66046429e-01
-2.13599689e-02 -5.04782438e-01 3.28314245e-01 9.54783022e-01
7.22036123e-01 8.07578802e-01 1.21367306e-01 -1.43017352e+00
-1.29652083e-01 -1.19602513e+00 -2.89014041e-01 3.84043753e-01
1.26345575e-01 -1.06830895e+00 -2.59071052e-01 9.97275263e-02]
|
[8.226682662963867, 0.5863378047943115]
|
7d5939b4-3909-4e9e-9486-11b520726768
|
wavepf-a-novel-fusion-approach-based-on
|
2305.17376
| null |
https://arxiv.org/abs/2305.17376v2
|
https://arxiv.org/pdf/2305.17376v2.pdf
|
DePF: A Novel Fusion Approach based on Decomposition Pooling for Infrared and Visible Images
|
Infrared and visible image fusion aims to generate synthetic images simultaneously containing salient features and rich texture details, which can be used to boost downstream tasks. However, existing fusion methods are suffering from the issues of texture loss and edge information deficiency, which result in suboptimal fusion results. Meanwhile, the straight-forward up-sampling operator can not well preserve the source information from multi-scale features. To address these issues, a novel fusion network based on the decomposition pooling (de-pooling) manner is proposed, termed as DePF. Specifically, a de-pooling based encoder is designed to extract multi-scale image and detail features of source images at the same time. In addition, the spatial attention model is used to aggregate these salient features. After that, the fused features will be reconstructed by the decoder, in which the up-sampling operator is replaced by the de-pooling reversed operation. Different from the common max-pooling technique, image features after the de-pooling layer can retain abundant details information, which is benefit to the fusion process. In this case, rich texture information and multi-scale information are maintained during the reconstruction phase. The experimental results demonstrate that the proposed method exhibits superior fusion performance over the state-of-the-arts on multiple image fusion benchmarks.
|
['Xiaoning Song', 'Zhongwei Shen', 'Chunyang Cheng', 'Yongbiao Xiao', 'Hui Li']
|
2023-05-27
| null | null | null | null |
['infrared-and-visible-image-fusion']
|
['computer-vision']
|
[ 4.04487818e-01 -3.11862826e-01 1.07093930e-01 -2.86665171e-01
-7.88951218e-01 5.92279807e-02 4.54833746e-01 2.74040792e-02
-1.99643865e-01 7.26817846e-01 5.10816813e-01 2.13595510e-01
-3.89885940e-02 -8.83952796e-01 -6.11757457e-01 -1.18362784e+00
4.84974474e-01 -7.45676875e-01 1.52046725e-01 -4.12555635e-01
9.59275588e-02 3.27724516e-01 -1.55827355e+00 5.54299653e-01
1.09337103e+00 1.36234915e+00 5.97564638e-01 2.29939688e-02
-1.60218477e-01 6.76881552e-01 -3.03503901e-01 -2.61216968e-01
2.72035599e-01 -5.07232845e-01 -2.12389231e-01 1.86490551e-01
1.33869752e-01 -5.10500193e-01 -5.26713669e-01 1.44610691e+00
6.52398288e-01 1.12989642e-01 6.98525980e-02 -1.06765258e+00
-8.20473850e-01 2.22118586e-01 -9.77971733e-01 2.52259135e-01
2.77711391e-01 2.26657167e-01 6.86148167e-01 -1.06032240e+00
3.43817651e-01 1.41405928e+00 4.40744132e-01 1.23894460e-01
-1.06488156e+00 -6.89398527e-01 2.40563273e-01 1.41437858e-01
-1.37901139e+00 -5.07106125e-01 9.92096245e-01 8.36721063e-02
4.25391793e-01 3.56736898e-01 5.21796823e-01 7.14341819e-01
5.13779581e-01 8.49441588e-01 1.29460287e+00 -7.82138482e-02
-7.56098852e-02 -7.55892918e-02 -2.18894988e-01 6.86056674e-01
2.95293689e-01 1.58568531e-01 -5.34635186e-01 2.12984994e-01
9.17444527e-01 5.26318967e-01 -6.71453118e-01 7.62444735e-02
-1.48867381e+00 4.30661917e-01 8.90720129e-01 2.52966046e-01
-8.75831902e-01 -1.77442729e-01 2.37663954e-01 3.03461462e-01
6.26725733e-01 -9.33805853e-02 -2.53534466e-01 4.38988686e-01
-9.24396217e-01 2.66464442e-01 6.34196680e-03 6.42295957e-01
1.15469170e+00 -7.51252919e-02 -7.06259847e-01 6.43984795e-01
2.37309217e-01 3.38184565e-01 4.90101188e-01 -6.75085485e-01
7.51119494e-01 8.03962886e-01 9.11874175e-02 -1.22732544e+00
-1.38449997e-01 -6.50529921e-01 -1.33378851e+00 3.84200454e-01
-8.70763436e-02 -1.37900501e-01 -1.01773643e+00 1.62661886e+00
2.84911811e-01 3.03310364e-01 2.92943507e-01 1.24661744e+00
1.03957140e+00 9.10486102e-01 4.02686372e-02 -3.92810643e-01
1.47442675e+00 -1.03247356e+00 -9.37166989e-01 -2.11453781e-01
2.04949658e-02 -9.68711376e-01 6.91197932e-01 6.68614879e-02
-1.18605888e+00 -8.60784292e-01 -1.19888246e+00 -1.65620685e-01
-1.98112130e-01 9.23484191e-02 5.05338490e-01 1.58606395e-01
-8.03968847e-01 4.24321562e-01 -5.90741217e-01 3.24015729e-02
6.00049078e-01 1.44094899e-01 -5.10898769e-01 -4.96368259e-01
-1.12750590e+00 6.22507036e-01 5.55753529e-01 4.53187466e-01
-3.92794043e-01 -6.35885179e-01 -9.51677024e-01 2.65491247e-01
3.62540185e-01 -8.50897312e-01 7.76007414e-01 -8.66220772e-01
-1.38313079e+00 2.43857265e-01 -4.20247674e-01 -6.43052859e-03
3.02553207e-01 1.17276225e-03 -4.85027373e-01 2.54863232e-01
2.06570387e-01 6.50889993e-01 8.66552472e-01 -1.29086316e+00
-1.00775278e+00 -3.25242102e-01 9.35830921e-02 6.47593319e-01
-1.89724877e-01 -5.15296571e-02 -4.43479985e-01 -9.77874458e-01
5.47762752e-01 -3.71243238e-01 -2.00381324e-01 2.67016273e-02
-3.14544946e-01 1.06554583e-01 1.06668746e+00 -9.90105093e-01
1.17355299e+00 -2.48419118e+00 2.71518797e-01 -6.39490262e-02
2.11067930e-01 1.39881745e-01 -1.26953572e-01 1.00035392e-01
-2.47961268e-01 -7.59851038e-02 -3.80192369e-01 -3.24177891e-01
-3.46827537e-01 -1.13445491e-01 -2.91495770e-01 3.93684417e-01
5.22388041e-01 1.05770671e+00 -8.91247809e-01 -5.34790814e-01
5.36499500e-01 6.88945949e-01 -3.55069339e-01 1.83276504e-01
1.23001859e-01 7.38115251e-01 -6.91868544e-01 6.93529665e-01
1.15476453e+00 -1.48963153e-01 -4.61364031e-01 -6.63940907e-01
-2.48315006e-01 -9.58401337e-02 -1.22643399e+00 1.87027454e+00
-4.72892463e-01 1.92844555e-01 2.70326853e-01 -9.62045014e-01
9.91668224e-01 3.00997436e-01 4.17904198e-01 -9.92805541e-01
1.15025088e-01 2.20716044e-01 -1.39077395e-01 -3.34137022e-01
4.95540172e-01 -4.35359448e-01 -9.56841931e-03 -9.51234624e-02
-1.32887110e-01 1.15148328e-01 -7.08128363e-02 -1.25349835e-01
6.50088787e-01 1.90423056e-01 6.51100874e-02 -1.08448513e-01
9.01274145e-01 -3.79566759e-01 9.66826499e-01 2.13571370e-01
-2.61567067e-02 7.35107541e-01 1.09136693e-01 -2.89120406e-01
-8.04991901e-01 -1.00494421e+00 1.31440699e-01 5.97778499e-01
7.46923447e-01 -1.35639191e-01 -3.84096503e-01 -3.57210368e-01
-2.29687795e-01 2.79984117e-01 -4.44936514e-01 -4.65209872e-01
-4.39859360e-01 -7.67475843e-01 -7.47397542e-03 4.08431858e-01
1.26933134e+00 -1.11566067e+00 -4.52185959e-01 4.37745035e-01
-5.70361316e-01 -9.95212317e-01 -6.00563467e-01 -7.46200085e-02
-7.77904451e-01 -7.52478302e-01 -1.02531552e+00 -8.36835563e-01
7.37541676e-01 9.56429482e-01 5.47142982e-01 1.07193224e-01
-4.77864072e-02 -3.77812594e-01 -4.24646527e-01 -2.85546184e-01
4.91216406e-02 -3.32358986e-01 -4.35150385e-01 5.69523692e-01
9.88376364e-02 -4.96812940e-01 -9.58827615e-01 8.89653340e-02
-1.18979704e+00 5.80330789e-01 1.17335451e+00 1.19062972e+00
7.15193748e-01 6.65616035e-01 5.79184592e-01 -2.34027222e-01
4.05484468e-01 -3.79180431e-01 -3.25302064e-01 3.30000103e-01
-2.71731406e-01 1.00800231e-01 7.28512168e-01 -5.23258485e-02
-1.67778993e+00 2.09527388e-02 -3.27710919e-02 -4.52596754e-01
7.26250634e-02 5.64810693e-01 -6.04723155e-01 -1.20549746e-01
-2.51101516e-02 7.60215223e-01 1.24385685e-03 -5.79024434e-01
1.03495263e-01 6.43273294e-01 6.25334024e-01 -2.03783408e-01
7.59333611e-01 5.85610926e-01 2.10196548e-03 -4.43728924e-01
-9.12428796e-01 -2.04483718e-01 -1.62811458e-01 -1.25268316e-02
9.90283906e-01 -1.34947276e+00 -4.48421001e-01 7.85100818e-01
-1.02804506e+00 3.79161507e-01 -1.30736470e-01 5.92664719e-01
-2.78652877e-01 5.46058416e-01 -6.75065756e-01 -6.68258667e-01
-5.21099031e-01 -1.41576898e+00 1.29336512e+00 8.90005767e-01
6.17808044e-01 -3.94797385e-01 -6.55928969e-01 3.45273674e-01
6.12239063e-01 3.70979756e-01 6.92730129e-01 1.29135847e-01
-8.01325440e-01 -4.80976924e-02 -6.96663141e-01 4.95390356e-01
3.19616884e-01 -4.12521482e-01 -8.57942820e-01 -3.25615972e-01
1.85207859e-01 -3.00631635e-02 1.13056338e+00 2.71959782e-01
1.27440453e+00 -1.77801684e-01 -1.87673017e-01 7.48645127e-01
1.55622184e+00 1.42506927e-01 8.83097470e-01 2.94507921e-01
7.04228878e-01 3.72907639e-01 7.14569986e-01 2.81328261e-01
3.94405574e-01 5.48845947e-01 3.24739099e-01 -6.01756454e-01
-2.57275164e-01 -3.06160808e-01 2.32387245e-01 5.77291906e-01
-5.90153504e-03 4.61873077e-02 -3.82789135e-01 3.53482336e-01
-1.84880424e+00 -9.55352426e-01 1.35378286e-01 2.16822195e+00
7.85196483e-01 5.64591885e-02 -4.36383516e-01 2.96590298e-01
9.24615443e-01 4.99679208e-01 -3.88796091e-01 1.91723183e-01
-5.62237442e-01 2.53035784e-01 5.23847342e-01 2.30050117e-01
-1.01012242e+00 6.38178825e-01 4.81580305e+00 1.08935845e+00
-1.31728542e+00 9.60966051e-02 6.68190777e-01 1.74595103e-01
-4.19013590e-01 1.00057133e-01 -4.89369631e-01 8.65882933e-01
2.81980425e-01 -2.03607813e-01 5.30551434e-01 2.20146209e-01
2.84988225e-01 -2.55854040e-01 -4.00272936e-01 1.16926837e+00
-5.64166643e-02 -1.12847400e+00 1.01927713e-01 -3.09546143e-02
7.53528416e-01 -2.54339784e-01 6.56512976e-02 1.25116736e-01
4.06700708e-02 -7.98641682e-01 6.86924040e-01 8.07885289e-01
7.25734591e-01 -9.88434076e-01 9.73297179e-01 3.59391332e-01
-1.67150259e+00 -3.37257296e-01 -5.41224360e-01 4.36171144e-02
3.33696961e-01 9.65190828e-01 1.43647820e-01 1.42859614e+00
8.10872436e-01 9.02501345e-01 -4.06649679e-01 1.12599313e+00
-1.27290308e-01 2.43850239e-02 -2.85040349e-01 4.82406348e-01
2.25465655e-01 -3.34200859e-01 3.89183551e-01 9.08564806e-01
6.56107128e-01 2.51911372e-01 2.78152049e-01 7.50076950e-01
-7.28608966e-02 -4.82976958e-02 -4.05379981e-01 2.91318357e-01
3.21087003e-01 1.45858943e+00 -5.65723240e-01 -4.88549411e-01
-6.55686438e-01 1.21546078e+00 7.48826191e-02 4.93132204e-01
-7.45425463e-01 -5.72460592e-01 5.48021138e-01 -1.73168674e-01
4.48182434e-01 -3.86341847e-02 -3.51843208e-01 -1.36170936e+00
4.01025683e-01 -7.26569831e-01 1.36074200e-01 -9.76554930e-01
-1.18313777e+00 5.42175293e-01 -1.44932404e-01 -1.55715644e+00
4.39771861e-01 -1.20765261e-01 -7.91777372e-01 1.39306021e+00
-2.06490397e+00 -1.31654632e+00 -8.05370390e-01 6.68695331e-01
4.08642232e-01 1.91176273e-02 3.96596164e-01 3.81955206e-01
-6.87802553e-01 3.10447693e-01 1.24271601e-01 -1.62047267e-01
5.84338248e-01 -7.57206321e-01 7.97681287e-02 1.00690138e+00
-4.12807703e-01 3.57366323e-01 3.44998628e-01 -7.20921159e-01
-1.31138253e+00 -1.29335594e+00 6.55525446e-01 3.34953099e-01
1.57394230e-01 1.57531851e-03 -1.07798445e+00 2.77052641e-01
2.41858020e-01 4.54280406e-01 1.89364046e-01 -6.24919176e-01
-1.35602430e-01 -3.98692846e-01 -1.21410489e+00 5.85934043e-01
8.64460230e-01 -4.24050182e-01 -5.91338158e-01 3.11654899e-02
7.80162036e-01 -4.07125264e-01 -8.82360935e-01 6.91255927e-01
4.61187094e-01 -1.25351369e+00 1.02922630e+00 1.78523026e-02
7.14739144e-01 -7.14734852e-01 -1.77945986e-01 -1.30973840e+00
-5.49085438e-01 -3.96323085e-01 1.48176014e-01 1.42114353e+00
6.31540343e-02 -8.02568316e-01 3.42187822e-01 2.92424321e-01
-1.74645111e-01 -8.38249445e-01 -9.31600630e-01 -5.09431541e-01
-4.01627958e-01 2.41013572e-01 8.29152226e-01 7.38407671e-01
-3.17948699e-01 3.04738641e-01 -3.99653763e-01 2.30044350e-01
6.72402620e-01 3.19781303e-01 4.28815067e-01 -8.79723907e-01
-5.28333820e-02 -4.05406684e-01 -3.49103749e-01 -1.08395231e+00
-1.15079314e-01 -7.03250825e-01 1.49638474e-01 -1.72970724e+00
2.91670859e-01 -2.69433141e-01 -6.89502180e-01 3.95513147e-01
-8.12342048e-01 3.59391421e-01 2.41207793e-01 2.46607199e-01
-3.74354452e-01 1.04984605e+00 1.71472037e+00 -1.85634136e-01
-9.92708951e-02 -3.13940883e-01 -1.07281017e+00 5.27951002e-01
6.90296233e-01 -1.63861215e-01 -2.82068849e-01 -4.67809737e-01
-1.87955946e-01 3.04096550e-01 6.37547851e-01 -1.14186859e+00
2.64178544e-01 -1.33620724e-01 8.61011446e-01 -5.53689122e-01
3.58751953e-01 -8.85157824e-01 1.34416193e-01 3.90825838e-01
4.16897908e-02 -3.54979970e-02 1.62038222e-01 6.94317997e-01
-7.22540617e-01 2.90356189e-01 6.47528648e-01 -1.05481803e-01
-8.21138382e-01 4.68951702e-01 1.03128754e-01 -4.56205994e-01
1.07027662e+00 -2.92414576e-01 -4.65251356e-01 -1.28676444e-01
-3.75488520e-01 2.53783703e-01 4.95531678e-01 5.18739223e-01
7.84377873e-01 -1.64888752e+00 -9.99850988e-01 4.85011667e-01
1.44554615e-01 2.73288757e-01 8.44648778e-01 1.11127234e+00
-2.89408475e-01 4.13988680e-02 -4.98118639e-01 -3.47982436e-01
-1.02475083e+00 6.60054922e-01 1.36065066e-01 -3.50740820e-01
-8.38302612e-01 6.37670338e-01 5.22879660e-01 8.16692114e-02
-2.30257615e-01 -2.73044378e-01 -2.19440594e-01 -1.15382269e-01
7.00060785e-01 4.37155440e-02 8.32643434e-02 -7.93823004e-01
-2.18538851e-01 6.91742301e-01 -5.54002412e-02 1.42696947e-01
1.29422534e+00 -5.00648558e-01 -2.31504649e-01 4.81210798e-02
1.26453280e+00 4.30286340e-02 -1.42454541e+00 -6.24177992e-01
-5.35964906e-01 -7.79297709e-01 5.02004802e-01 -6.29240394e-01
-1.42812240e+00 8.24214697e-01 6.64095640e-01 -1.34841921e-02
1.76905274e+00 -3.26134026e-01 1.14377630e+00 -6.47168159e-02
2.57543892e-01 -7.67257869e-01 -1.62016839e-01 -2.88084857e-02
1.00835562e+00 -1.35260177e+00 1.12711810e-01 -6.68717384e-01
-6.31417155e-01 8.54825377e-01 5.81165254e-01 -2.42382333e-01
5.68218291e-01 2.23097831e-01 -1.77214563e-01 5.44277951e-02
-5.13911963e-01 -2.47321948e-01 3.25975269e-01 2.93675333e-01
2.45170221e-01 -5.92187308e-02 -3.89993995e-01 7.70242035e-01
2.24173203e-01 2.13390157e-01 1.52514204e-01 1.04620624e+00
-4.59022552e-01 -7.21231520e-01 -4.88731712e-01 5.52774251e-01
-5.53310275e-01 -2.68778205e-01 9.30856764e-02 2.54910201e-01
3.96536738e-01 1.12021434e+00 -8.67016837e-02 -6.03702307e-01
3.55808139e-01 -2.20611811e-01 2.42102683e-01 -2.78607130e-01
-4.96085018e-01 1.91043571e-01 -2.82406956e-01 -5.46364129e-01
-5.37413538e-01 -2.66310543e-01 -1.09804142e+00 -2.46512651e-01
-3.68248433e-01 3.03833727e-02 3.57405990e-01 8.33394229e-01
4.38652843e-01 8.44500601e-01 8.76619101e-01 -9.79826212e-01
-3.94615918e-01 -9.93344307e-01 -6.22600973e-01 6.04042530e-01
5.50373137e-01 -7.18852818e-01 -3.07741106e-01 4.50794660e-02]
|
[10.556697845458984, -1.8569436073303223]
|
8a6b1c24-fa7a-4d5b-81de-8e7a33a1a646
|
synchronized-audio-visual-frames-with
|
2112.14088
| null |
https://arxiv.org/abs/2112.14088v1
|
https://arxiv.org/pdf/2112.14088v1.pdf
|
Synchronized Audio-Visual Frames with Fractional Positional Encoding for Transformers in Video-to-Text Translation
|
Video-to-Text (VTT) is the task of automatically generating descriptions for short audio-visual video clips, which can support visually impaired people to understand scenes of a YouTube video for instance. Transformer architectures have shown great performance in both machine translation and image captioning, lacking a straightforward and reproducible application for VTT. However, there is no comprehensive study on different strategies and advice for video description generation including exploiting the accompanying audio with fully self-attentive networks. Thus, we explore promising approaches from image captioning and video processing and apply them to VTT by developing a straightforward Transformer architecture. Additionally, we present a novel way of synchronizing audio and video features in Transformers which we call Fractional Positional Encoding (FPE). We run multiple experiments on the VATEX dataset to determine a configuration applicable to unseen datasets that helps describe short video clips in natural language and improved the CIDEr and BLEU-4 scores by 37.13 and 12.83 points compared to a vanilla Transformer network and achieve state-of-the-art results on the MSR-VTT and MSVD datasets. Also, FPE helps increase the CIDEr score by a relative factor of 8.6%.
|
['Rainer Lienhart', 'Moritz Einfalt', 'Philipp Harzig']
|
2021-12-28
| null | null | null | null |
['video-description']
|
['computer-vision']
|
[ 5.37825048e-01 -1.24039799e-01 -1.57414138e-01 -3.39692205e-01
-1.12582815e+00 -4.65382427e-01 6.91174388e-01 -2.75924951e-01
-3.38442624e-01 7.21343815e-01 4.94426221e-01 -1.38279557e-01
2.35588104e-01 -4.16031718e-01 -8.94653201e-01 -3.19398135e-01
3.32642184e-03 3.18964571e-01 2.64199048e-01 -3.12816978e-01
-1.50724733e-02 7.70031586e-02 -1.84013271e+00 1.01154363e+00
5.49871445e-01 1.15845704e+00 4.18985456e-01 8.98869157e-01
-1.80641174e-01 1.03599036e+00 -5.82039714e-01 -5.91356337e-01
2.05149963e-01 -4.59368199e-01 -7.35462189e-01 1.28017098e-01
8.98812652e-01 -5.69314539e-01 -7.10363388e-01 7.59161532e-01
7.35395014e-01 -5.37866019e-02 5.95929742e-01 -1.55940378e+00
-8.74516189e-01 7.69125640e-01 -1.87383786e-01 6.45328835e-02
8.54789019e-01 3.90484661e-01 9.20598447e-01 -8.30308139e-01
9.18084204e-01 1.17146766e+00 4.68476683e-01 8.32431674e-01
-9.19222713e-01 -6.90769255e-01 -7.55991563e-02 7.69266784e-01
-1.50310588e+00 -7.71637022e-01 2.44167939e-01 -3.90479326e-01
1.21503425e+00 4.88681674e-01 8.94271433e-01 1.58488047e+00
-1.99830517e-01 9.59191501e-01 7.51195252e-01 -2.11232111e-01
-7.88905695e-02 2.16571495e-01 -4.41902846e-01 3.95354629e-01
-1.80482507e-01 -1.34723008e-01 -8.24675858e-01 3.51383775e-01
5.97833931e-01 -4.52015787e-01 -4.88703072e-01 -1.04263656e-01
-1.47389865e+00 7.06177592e-01 1.83119699e-01 7.04607517e-02
-1.55381739e-01 3.95900309e-01 7.79239416e-01 3.70240480e-01
1.91498116e-01 2.68811017e-01 -9.27206650e-02 -5.46435177e-01
-1.07527304e+00 2.98580766e-01 4.61139321e-01 1.43776286e+00
2.28472561e-01 1.25577584e-01 -9.22265470e-01 7.98958302e-01
-2.74170488e-02 7.10525632e-01 4.33194786e-01 -1.17875218e+00
7.86583483e-01 7.23602697e-02 9.25108269e-02 -6.05622590e-01
3.35093215e-02 -1.02861591e-01 -6.71836495e-01 -1.91901639e-01
2.15877324e-01 2.80565992e-02 -1.16134202e+00 1.64811194e+00
-3.66331846e-01 2.04093978e-01 8.15494433e-02 1.06985152e+00
1.30944967e+00 9.69606578e-01 2.41865125e-02 -1.19130507e-01
1.35656786e+00 -1.08314288e+00 -8.68654668e-01 -1.90494046e-01
4.02310133e-01 -8.43310833e-01 1.21109104e+00 1.78533390e-01
-1.18110430e+00 -5.18531680e-01 -7.91733623e-01 -3.38255435e-01
-2.03305781e-01 2.44859740e-01 2.52997518e-01 4.45825845e-01
-1.59253156e+00 1.74568206e-01 -4.38134491e-01 -5.60372055e-01
4.16242152e-01 2.67805845e-01 -5.39072454e-01 -2.58938432e-01
-1.26662350e+00 7.46176183e-01 4.33396131e-01 -1.59058884e-01
-1.18118238e+00 -6.32108510e-01 -9.38154280e-01 7.31200352e-02
2.03574464e-01 -9.05911148e-01 1.50721061e+00 -1.06203580e+00
-1.54225671e+00 7.57567704e-01 -4.74066436e-01 -9.27497983e-01
8.51022661e-01 -9.69595686e-02 -3.60024780e-01 5.38685322e-01
2.69973487e-01 1.58613098e+00 9.77529109e-01 -9.30402100e-01
-6.30288184e-01 2.60396630e-01 2.45622292e-01 4.38251197e-01
-4.60476428e-01 2.03818813e-01 -8.83817255e-01 -6.11364901e-01
-4.75564510e-01 -9.00484562e-01 3.01280022e-01 2.80373663e-01
-4.50670868e-01 -1.41723957e-02 9.16803062e-01 -9.29207265e-01
1.02643251e+00 -2.12412477e+00 4.82969135e-02 -2.83360749e-01
2.22464904e-01 3.04572135e-01 -5.32055676e-01 4.66038644e-01
-3.28053124e-02 2.20284671e-01 -1.51999923e-03 -3.61568600e-01
6.84535280e-02 2.05598518e-01 -5.17392576e-01 -7.75981024e-02
2.13538915e-01 9.90107298e-01 -8.83049369e-01 -4.72491592e-01
3.21361840e-01 8.88351142e-01 -6.69822454e-01 2.25669742e-01
-3.40418428e-01 2.40356773e-01 -8.87945741e-02 5.74360907e-01
3.03116024e-01 -2.15728343e-01 -1.82371259e-01 -2.81198293e-01
-1.34876892e-01 2.03684688e-01 -6.99717641e-01 1.80227745e+00
-4.04941797e-01 1.36270988e+00 -3.75736982e-01 -5.56594253e-01
6.92439914e-01 7.05208659e-01 4.89144474e-01 -9.39989269e-01
1.15067586e-01 1.43232182e-01 -4.42479342e-01 -8.97773504e-01
7.01039374e-01 2.64996201e-01 6.59427419e-02 2.14990098e-02
1.61571175e-01 8.83020833e-02 4.70122069e-01 3.09544027e-01
1.17004442e+00 5.49729988e-02 1.30748590e-02 3.23941678e-01
4.04277861e-01 4.26927246e-02 2.53607910e-02 7.87232339e-01
-1.56954095e-01 1.14118338e+00 4.39671367e-01 -3.01910967e-01
-1.30331278e+00 -1.05051279e+00 1.87884390e-01 8.59502137e-01
4.70552454e-03 -9.69089150e-01 -8.33508730e-01 -3.03153425e-01
-4.09941882e-01 7.64896214e-01 -4.68143731e-01 -4.81482781e-02
-2.45710656e-01 -1.67586759e-01 7.28154004e-01 4.76551324e-01
9.22643125e-01 -9.23475683e-01 -5.52010298e-01 1.58067308e-02
-9.55895960e-01 -1.80971122e+00 -6.77078784e-01 -3.87925148e-01
-3.19878787e-01 -6.70476198e-01 -1.14252329e+00 -7.99392700e-01
1.84732258e-01 3.71260107e-01 1.01259005e+00 -3.24803799e-01
-1.41905501e-01 5.02780735e-01 -6.95687830e-01 -2.13055208e-01
-4.95906144e-01 4.54594707e-03 -1.53373443e-02 5.62708043e-02
1.79621771e-01 -3.39380473e-01 -5.26485801e-01 2.91371226e-01
-8.28161776e-01 5.33293605e-01 5.69055021e-01 5.91999054e-01
6.16162360e-01 -2.70320058e-01 1.88523293e-01 -1.81822985e-01
5.52251220e-01 -3.28385442e-01 -3.61429930e-01 2.05785841e-01
-3.60420942e-01 -4.23156805e-02 5.05699635e-01 -6.40952706e-01
-7.48212159e-01 1.41957968e-01 -2.55693525e-01 -8.76598001e-01
-8.90957341e-02 1.43850327e-01 -6.02190122e-02 -6.18056729e-02
5.31653821e-01 5.26115596e-01 -7.79446810e-02 -1.27111360e-01
4.47339207e-01 9.25174594e-01 8.08476388e-01 -1.79499656e-01
7.72498071e-01 3.28622937e-01 -3.67420882e-01 -7.90598452e-01
-5.68564773e-01 -3.85366738e-01 -1.47050530e-01 -5.50232589e-01
1.15868211e+00 -1.21537852e+00 -6.63340688e-01 3.87238622e-01
-1.41541767e+00 -3.28047127e-01 -1.62187457e-01 3.68636221e-01
-9.68327641e-01 2.51987189e-01 -4.26172018e-01 -3.87499064e-01
-3.58896285e-01 -1.45763838e+00 1.26016724e+00 1.04953609e-02
-2.19729736e-01 -4.64616120e-01 -2.87907600e-01 6.22437358e-01
6.21154547e-01 1.15412220e-01 5.51194131e-01 -3.59429777e-01
-7.97604144e-01 7.52701014e-02 -5.30813992e-01 3.44605714e-01
-1.68774307e-01 -5.21621592e-02 -1.11402524e+00 -2.67546535e-01
-5.47652364e-01 -3.21681470e-01 9.16485071e-01 4.12166655e-01
1.17987514e+00 -5.20978749e-01 -8.59284699e-02 7.50708520e-01
1.39388764e+00 2.23698750e-01 1.02268505e+00 5.53384483e-01
6.41787350e-01 3.69986504e-01 4.25895691e-01 4.15645152e-01
6.31182909e-01 1.14630771e+00 5.15475035e-01 1.74379032e-02
-6.85161829e-01 -5.04333556e-01 8.18330944e-01 5.93022406e-01
-9.66004729e-02 -5.56390584e-01 -6.98283792e-01 6.61631107e-01
-1.74680614e+00 -1.16634142e+00 -9.52256247e-02 1.97779858e+00
7.25288272e-01 -1.36376009e-03 8.96635950e-02 9.93611589e-02
8.99523556e-01 9.45974067e-02 -2.88872182e-01 -3.98892730e-01
-3.92156422e-01 4.76954831e-03 4.97295648e-01 2.27701798e-01
-9.16634738e-01 1.03121090e+00 6.60708380e+00 9.46879566e-01
-1.27436483e+00 1.96187422e-01 4.17267501e-01 -4.07884717e-01
-1.29852280e-01 -3.20662111e-01 -6.61053240e-01 4.51726198e-01
1.35342336e+00 -2.41626933e-01 6.48689032e-01 5.83783925e-01
5.67672253e-01 2.45193526e-01 -1.17159963e+00 1.58658350e+00
4.17700350e-01 -1.57030737e+00 6.56940043e-01 -1.55274451e-01
5.81893444e-01 2.18335032e-01 2.66851008e-01 3.02841812e-01
-2.78794825e-01 -1.06993818e+00 1.26478624e+00 3.58195812e-01
1.42528522e+00 -3.88478875e-01 7.28947043e-01 -2.19017595e-01
-1.26998925e+00 -1.55279741e-01 -2.17462629e-01 1.37921408e-01
4.45255667e-01 1.05122417e-01 -1.19789839e+00 3.85249168e-01
1.00814915e+00 8.80232275e-01 -7.58995235e-01 1.36596382e+00
-1.99667901e-01 4.59373832e-01 -1.16266727e-01 5.57963066e-02
3.82466465e-01 2.49303818e-01 7.58632779e-01 1.36839628e+00
7.31171668e-01 -2.18352661e-01 -2.44314641e-01 7.04214752e-01
-2.33676359e-01 1.42468974e-01 -7.87074685e-01 -1.42219916e-01
3.98162901e-01 8.74983847e-01 -3.99392158e-01 -3.88534337e-01
-4.28413630e-01 1.19218719e+00 -1.45885289e-01 4.00546223e-01
-1.17027235e+00 -3.09487253e-01 7.36660540e-01 4.00078982e-01
5.40939212e-01 -5.98561279e-02 2.87063271e-01 -1.21967769e+00
3.30797493e-01 -1.02207017e+00 1.65157124e-01 -1.61869657e+00
-9.26278234e-01 1.04424059e+00 2.60003835e-01 -1.67678118e+00
-6.05883121e-01 -5.39154649e-01 -2.36494094e-01 4.81231272e-01
-1.56526327e+00 -1.16533279e+00 -7.42833018e-01 9.94725347e-01
9.22405362e-01 -3.15255910e-01 6.84327245e-01 5.66659689e-01
-2.63139307e-01 7.93188274e-01 -3.08219716e-02 1.70946106e-01
9.49936748e-01 -8.58264208e-01 5.05733430e-01 9.48009610e-01
2.03694254e-01 3.29668857e-02 9.31876779e-01 -4.13069695e-01
-1.31570268e+00 -1.37687242e+00 9.99539793e-01 -3.76067191e-01
5.41478813e-01 -4.51577514e-01 -5.22285163e-01 6.05398297e-01
5.45912206e-01 -1.51857197e-01 3.56470138e-01 -6.69109702e-01
-5.81127524e-01 -1.42471194e-01 -9.62451458e-01 8.37369382e-01
1.33452380e+00 -6.83850944e-01 -2.78382003e-01 5.02867877e-01
1.16654527e+00 -5.71025372e-01 -6.96865320e-01 1.50828019e-01
5.84967792e-01 -9.46819365e-01 9.21393216e-01 -2.23095953e-01
6.18508220e-01 -3.50545526e-01 -4.79605675e-01 -1.10310316e+00
-4.45029251e-02 -9.52833474e-01 7.99001679e-02 1.25808346e+00
4.39974219e-01 -2.64102906e-01 5.30084372e-01 3.29682022e-01
-2.81528771e-01 -3.68160933e-01 -9.93957281e-01 -8.51582646e-01
-4.75113243e-01 -8.33593667e-01 4.58707005e-01 5.57907701e-01
-1.79208145e-01 3.16753924e-01 -8.01682591e-01 -1.40998632e-01
4.61844474e-01 -4.42207307e-01 6.32277071e-01 -7.17737019e-01
-8.01334083e-02 -3.37406188e-01 -8.02165151e-01 -1.06571734e+00
1.15090467e-01 -9.69176590e-01 -1.21656351e-03 -1.82266533e+00
2.63790578e-01 2.09212497e-01 2.23689333e-01 5.62739491e-01
4.06650394e-01 6.56756341e-01 5.25627136e-01 1.46681935e-01
-7.50675797e-01 5.74212492e-01 1.18498170e+00 -5.97492039e-01
-4.20168936e-02 -4.12655532e-01 -5.96126437e-01 1.28316060e-01
6.61026299e-01 -2.73023039e-01 -6.88325942e-01 -9.09645140e-01
1.08267926e-01 2.24211589e-01 5.21999717e-01 -1.20811486e+00
2.18619525e-01 1.37992427e-01 2.21202031e-01 -4.22412544e-01
7.22770095e-01 -6.50598586e-01 3.69447142e-01 1.55245766e-01
-5.86039186e-01 1.83012843e-01 3.79413605e-01 3.87710035e-01
-4.20559675e-01 5.25800362e-02 5.04604161e-01 5.91484644e-02
-1.01699674e+00 3.96097124e-01 -6.50477350e-01 5.34321256e-02
9.43177760e-01 -4.83709246e-01 -5.86616635e-01 -1.02499294e+00
-5.76423943e-01 2.21593857e-01 3.50879312e-01 8.47430944e-01
1.05623150e+00 -1.50314820e+00 -1.04422593e+00 9.19516906e-02
4.95564133e-01 -4.46573496e-01 2.86207557e-01 6.74912930e-01
-5.64128041e-01 7.50752032e-01 -6.14315391e-01 -8.92689288e-01
-1.48449206e+00 3.22154284e-01 2.75118738e-01 1.55964419e-01
-7.98437655e-01 7.16884375e-01 -8.89105052e-02 2.64475286e-01
3.82713109e-01 -3.84388894e-01 -3.48993301e-01 -2.13447353e-03
7.82955170e-01 1.11869305e-01 7.33972192e-02 -8.86863530e-01
-2.07285732e-01 5.52780867e-01 -6.96690008e-02 -3.81610513e-01
1.15776610e+00 -3.16038072e-01 1.89918309e-01 7.34287202e-02
1.34701312e+00 -4.21270400e-01 -1.06852627e+00 2.39686556e-02
-4.69687611e-01 -4.94093180e-01 2.05140170e-02 -8.25698137e-01
-1.10663259e+00 8.17973256e-01 8.29485059e-01 2.42859442e-02
1.33757603e+00 6.56654611e-02 9.72465396e-01 4.80129480e-01
3.00951868e-01 -6.93569303e-01 1.41749501e-01 4.42606896e-01
1.17388773e+00 -1.02627242e+00 -6.01499617e-01 -3.21766287e-01
-9.21855450e-01 1.16557896e+00 4.95045811e-01 4.33550805e-01
-6.11372367e-02 1.39775559e-01 5.24570011e-02 1.95032820e-01
-1.03907084e+00 -2.71223366e-01 4.40305710e-01 9.49761629e-01
1.76137432e-01 -8.59500915e-02 1.31697938e-01 2.97707647e-01
-4.99272704e-01 2.38929704e-01 8.67202997e-01 3.91540200e-01
-2.86531627e-01 -8.72701705e-01 -2.68890262e-01 3.04488093e-01
-3.46930295e-01 -3.09295088e-01 -3.91055107e-01 7.70308197e-01
-1.25946581e-01 1.05806398e+00 9.33907405e-02 -7.64548421e-01
2.95637637e-01 -3.80007066e-02 4.60650653e-01 -3.95511568e-01
-3.44731987e-01 -1.00371800e-02 4.25119400e-01 -8.00902843e-01
-5.75656295e-01 -5.53351343e-01 -9.01062667e-01 -3.04908633e-01
1.91761270e-01 3.49339209e-02 6.73581839e-01 8.11788678e-01
3.89252275e-01 5.24740040e-01 2.64759272e-01 -9.29926157e-01
-1.57636553e-02 -8.95126283e-01 -8.34425017e-02 2.89399743e-01
4.29097086e-01 -4.51660007e-01 -3.17000598e-01 4.08145159e-01]
|
[10.6807222366333, 0.8798953890800476]
|
fde0e387-5f3c-4aea-9250-82e4f8e3db2d
|
image-generation-from-freehand-scene-sketches
|
2003.02683
| null |
https://arxiv.org/abs/2003.02683v5
|
https://arxiv.org/pdf/2003.02683v5.pdf
|
SketchyCOCO: Image Generation from Freehand Scene Sketches
|
We introduce the first method for automatic image generation from scene-level freehand sketches. Our model allows for controllable image generation by specifying the synthesis goal via freehand sketches. The key contribution is an attribute vector bridged Generative Adversarial Network called EdgeGAN, which supports high visual-quality object-level image content generation without using freehand sketches as training data. We have built a large-scale composite dataset called SketchyCOCO to support and evaluate the solution. We validate our approach on the tasks of both object-level and scene-level image generation on SketchyCOCO. Through quantitative, qualitative results, human evaluation and ablation studies, we demonstrate the method's capacity to generate realistic complex scene-level images from various freehand sketches.
|
['Li-Min Wang', 'Jianzhuang Liu', 'Changqing Zou', 'Chengying Gao', 'Qi Xu', 'Qi Liu']
|
2020-03-05
|
sketchycoco-image-generation-from-freehand
|
http://openaccess.thecvf.com/content_CVPR_2020/html/Gao_SketchyCOCO_Image_Generation_From_Freehand_Scene_Sketches_CVPR_2020_paper.html
|
http://openaccess.thecvf.com/content_CVPR_2020/papers/Gao_SketchyCOCO_Image_Generation_From_Freehand_Scene_Sketches_CVPR_2020_paper.pdf
|
cvpr-2020-6
|
['sketch-to-image-translation']
|
['computer-vision']
|
[ 5.35340369e-01 3.53970557e-01 9.07159001e-02 6.91906661e-02
-8.92556906e-01 -8.57423544e-01 1.15261388e+00 -8.39497924e-01
1.81197554e-01 7.59228051e-01 1.46017242e-02 -1.97176725e-01
3.13464373e-01 -1.13047540e+00 -8.54603350e-01 -4.70267266e-01
1.00852162e-01 2.96870440e-01 -1.85424656e-01 -3.34264785e-01
7.78197199e-02 7.91409194e-01 -1.40846956e+00 4.99965131e-01
7.63129711e-01 7.66079128e-01 1.72182500e-01 1.03335011e+00
2.00668108e-02 6.68356836e-01 -9.83151495e-01 -6.38730586e-01
6.22558594e-01 -6.80925369e-01 -2.28310913e-01 2.76466012e-01
9.84935999e-01 -7.27665901e-01 -4.53558475e-01 7.03205764e-01
6.75052226e-01 -2.10300520e-01 9.09394085e-01 -1.87762094e+00
-8.34564209e-01 2.90526122e-01 -1.85157150e-01 -7.63892114e-01
5.11360347e-01 8.59888673e-01 8.00162077e-01 -1.03575420e+00
1.22164321e+00 1.33397901e+00 4.13226873e-01 1.01276028e+00
-1.52838469e+00 -1.09467506e+00 -2.55775899e-01 -5.86056352e-01
-1.43305647e+00 -5.46530843e-01 1.06581533e+00 -4.54179138e-01
5.37830770e-01 3.73587370e-01 8.59886587e-01 1.52535117e+00
1.40104994e-01 9.62038994e-01 1.23247290e+00 -4.52013582e-01
9.78693515e-02 1.44661620e-01 -9.13529813e-01 8.71638298e-01
-9.77921300e-03 8.08553696e-01 -3.90985280e-01 -1.38162434e-01
1.50234604e+00 -4.40685809e-01 -1.26387060e-01 -4.99627441e-01
-1.27754211e+00 7.28485107e-01 4.02774751e-01 -1.11412615e-01
-2.83717960e-01 9.00878310e-01 -9.77771655e-02 1.80930689e-01
1.28208712e-01 9.16091502e-01 1.24342240e-01 2.34719217e-01
-1.08863914e+00 7.97984719e-01 6.40792847e-01 1.61676002e+00
4.23696727e-01 6.20266676e-01 -7.65010655e-01 5.97647011e-01
4.46639322e-02 1.10039723e+00 -3.03066932e-02 -1.23738968e+00
3.54759961e-01 1.77006900e-01 3.25948149e-01 -9.33763802e-01
3.48331332e-01 -3.65115069e-02 -7.47013628e-01 1.07297599e+00
8.60532150e-02 -2.43849427e-01 -1.33423197e+00 1.72304463e+00
-2.20535975e-02 -2.06556283e-02 5.98154999e-02 8.15619230e-01
1.03329122e+00 7.08671629e-01 2.45325714e-01 3.01502585e-01
1.09251308e+00 -8.74149084e-01 -6.44439220e-01 1.14901893e-01
-1.67211205e-01 -7.38542020e-01 1.32868147e+00 3.18792969e-01
-1.46621549e+00 -7.15920806e-01 -1.23978913e+00 7.45701641e-02
-3.67217183e-01 3.29533666e-01 7.35212624e-01 7.73039401e-01
-1.20197165e+00 3.39296043e-01 -2.36978084e-01 1.40978485e-01
6.81937158e-01 5.08926325e-02 -4.74159449e-01 5.02603911e-02
-1.15285897e+00 6.30823374e-01 1.95241019e-01 -3.73634726e-01
-1.40224314e+00 -9.47925150e-01 -8.72649729e-01 1.06402151e-02
1.23149537e-01 -1.07066321e+00 9.38550293e-01 -9.27214086e-01
-1.83987629e+00 9.95649457e-01 4.09271777e-01 -6.22872449e-02
8.44490170e-01 2.70625681e-01 -2.44294494e-01 2.17386216e-01
8.27376768e-02 1.57106173e+00 1.34159899e+00 -1.88613021e+00
-2.63475090e-01 2.70623416e-01 7.47277737e-02 -2.66935602e-02
2.14196190e-01 -2.86996573e-01 -4.93704945e-01 -1.26897216e+00
-3.68854970e-01 -9.27407622e-01 -2.30109990e-01 3.87181222e-01
-8.31300795e-01 3.07633251e-01 9.14146125e-01 -3.87259781e-01
5.02657831e-01 -1.94525540e+00 1.44601926e-01 2.70297498e-01
2.23833010e-01 2.73045421e-01 -7.72726297e-01 7.22379863e-01
-1.94318034e-02 5.39111018e-01 -3.20411086e-01 -3.28945845e-01
1.86535582e-01 -1.00484438e-01 -9.01563168e-01 -1.63578108e-01
4.82181102e-01 1.66792631e+00 -8.22348356e-01 -5.51679790e-01
4.81652111e-01 5.84910214e-01 -6.43795609e-01 6.05561614e-01
-6.06854081e-01 4.48411435e-01 -2.93303192e-01 7.06753194e-01
5.88727355e-01 4.67573032e-02 -1.17606036e-01 -2.31468797e-01
2.56833971e-01 -4.48041052e-01 -9.23054576e-01 1.94691324e+00
-7.01078892e-01 7.88928986e-01 1.26348019e-01 -6.54235575e-03
9.00880396e-01 3.40097040e-01 9.73098353e-02 -6.86461926e-01
-2.15494424e-01 -5.05838823e-03 -4.16934133e-01 -1.42582059e-01
5.48767686e-01 -2.94795156e-01 -5.60279727e-01 4.65719193e-01
2.29911059e-01 -1.30317867e+00 -1.75401766e-03 6.42610550e-01
8.95572782e-01 3.86606067e-01 -3.69635485e-02 -1.23772018e-01
2.36383140e-01 -1.33555934e-01 -1.61197439e-01 9.90050256e-01
2.99926609e-01 1.21792650e+00 4.69602227e-01 -2.54487634e-01
-1.52840042e+00 -1.62731493e+00 1.83817700e-01 4.70169574e-01
1.11028783e-01 -3.77848685e-01 -8.31035495e-01 -7.28467941e-01
1.57944351e-01 8.65425408e-01 -7.28526652e-01 -7.55011588e-02
-3.36062580e-01 1.24674119e-01 9.44739580e-01 3.44792396e-01
4.39434201e-01 -1.68063891e+00 -3.93390000e-01 6.27042353e-02
1.98791698e-01 -1.18064606e+00 -6.35322452e-01 -5.87767243e-01
-2.80516148e-01 -9.29700375e-01 -1.05680156e+00 -4.93842483e-01
8.66845489e-01 -1.40159830e-01 1.41665077e+00 1.41186178e-01
-8.49365115e-01 7.67362356e-01 5.04072290e-03 -4.41026181e-01
-9.36187327e-01 -2.59956658e-01 -3.20318371e-01 -2.62164026e-01
-8.81702363e-01 -7.28890955e-01 -7.67815053e-01 2.15273023e-01
-1.18497717e+00 6.77293241e-01 6.34366453e-01 9.70659137e-01
4.60565239e-01 -2.49794126e-01 6.59496188e-01 -7.86189079e-01
9.72040772e-01 6.86866790e-02 -8.80225241e-01 2.20761731e-01
-3.84985626e-01 5.57776168e-02 7.04526901e-01 -5.58092773e-01
-1.06501603e+00 7.74134919e-02 -6.72576576e-02 -6.91194296e-01
-1.92452058e-01 -2.40208536e-01 -4.64561671e-01 -2.98178852e-01
7.72911787e-01 4.47479248e-01 8.14517587e-03 2.08850265e-01
1.16391683e+00 2.91130602e-01 9.76901591e-01 -9.31604862e-01
1.28496206e+00 4.59105164e-01 -8.73679370e-02 -7.52203524e-01
-6.09817952e-02 6.17766500e-01 -3.32300752e-01 -4.21552390e-01
9.57964838e-01 -8.24556768e-01 -7.99958587e-01 5.29849052e-01
-1.24104464e+00 -8.39112043e-01 -5.88208675e-01 -3.53204221e-01
-1.06534600e+00 -1.42989695e-01 -4.94853437e-01 -6.97532654e-01
-6.60586596e-01 -1.20957971e+00 1.65708566e+00 -5.62262535e-03
-2.19576254e-01 -6.97632253e-01 -5.20669892e-02 6.84284791e-02
5.42660713e-01 9.44927156e-01 7.62366235e-01 2.91342586e-01
-1.33612669e+00 -2.24179804e-01 -2.76371658e-01 3.13964903e-01
2.32873093e-02 2.20742092e-01 -8.35023582e-01 -3.51772964e-01
-7.82757223e-01 -6.28422141e-01 4.83149439e-01 1.05050214e-01
1.22569859e+00 -4.56174046e-01 -2.92125046e-01 9.01681066e-01
1.50521207e+00 1.89043239e-01 1.18188512e+00 -3.63829076e-01
8.90513062e-01 1.97746009e-01 3.82433653e-01 2.17291847e-01
-1.32777274e-01 8.53961408e-01 2.10188061e-01 -5.49570441e-01
-9.14515615e-01 -9.72385466e-01 4.64663319e-02 1.15691170e-01
8.14623684e-02 -6.89919531e-01 -5.07877111e-01 5.46329200e-01
-1.20302224e+00 -1.03212261e+00 2.28873149e-01 1.77319598e+00
8.40055704e-01 -1.86357543e-01 -3.94376516e-02 -2.03882948e-01
5.51445901e-01 1.12447910e-01 -3.89669508e-01 -3.24808180e-01
-8.10190141e-02 5.67303061e-01 3.13336939e-01 5.95015824e-01
-7.99751222e-01 1.38315475e+00 7.33498621e+00 1.07764995e+00
-1.13896203e+00 -2.16037348e-01 5.97367108e-01 -2.25445881e-01
-8.22974443e-01 -1.48649171e-01 -2.95560449e-01 5.15858054e-01
2.18128845e-01 -1.42268389e-01 7.24870503e-01 6.33081198e-01
-1.23624600e-01 2.01166198e-01 -1.03531241e+00 1.06760025e+00
-1.36588304e-03 -1.96191943e+00 5.96516073e-01 1.21947818e-01
1.11097205e+00 -6.96704090e-01 3.93838853e-01 1.97562665e-01
7.67239630e-01 -1.47693527e+00 9.33659136e-01 5.56685805e-01
1.96018231e+00 -8.25205743e-01 -1.57527238e-01 8.96075144e-02
-9.44025636e-01 4.10236597e-01 9.82502252e-02 4.55476552e-01
2.64513880e-01 1.50700092e-01 -8.49632740e-01 2.72324264e-01
1.59782562e-02 1.84272721e-01 -4.17515457e-01 4.20260936e-01
-6.53931916e-01 2.89194554e-01 -9.12893936e-03 2.15046719e-01
1.37087315e-01 -3.57450433e-02 6.24028087e-01 1.11782682e+00
3.59405726e-01 2.97871739e-01 -1.21204695e-02 1.77993774e+00
-2.21142158e-01 -3.30620021e-01 -1.28696394e+00 -3.91634554e-01
5.93075752e-01 1.30596781e+00 -6.09065652e-01 -4.83800322e-01
3.28592032e-01 1.45241165e+00 -1.80794671e-03 5.59875071e-01
-8.89872968e-01 -7.11064994e-01 7.16208875e-01 1.97930157e-01
2.25878328e-01 -1.90394014e-01 -4.17293459e-01 -9.77588117e-01
-1.79380029e-01 -1.07634139e+00 -3.85312974e-01 -1.37265611e+00
-1.13331914e+00 7.42634833e-01 -1.88691299e-02 -1.05055916e+00
-5.92185795e-01 -4.51785952e-01 -7.60561645e-01 1.06734002e+00
-9.12410796e-01 -1.87886584e+00 -5.69186807e-01 6.25187218e-01
4.19606894e-01 -5.73282123e-01 1.01764345e+00 7.13615194e-02
8.70645717e-02 9.94478106e-01 -4.78629380e-01 1.72889203e-01
4.98315990e-01 -1.23129141e+00 1.11947250e+00 6.34010136e-01
2.59691894e-01 3.72587949e-01 6.10140502e-01 -7.64512777e-01
-1.45810080e+00 -1.06264365e+00 2.36203372e-01 -7.31647134e-01
8.60154405e-02 -9.03890610e-01 -2.27035120e-01 4.58903223e-01
3.94860595e-01 2.50369813e-02 6.73587322e-02 -5.93284488e-01
-4.25077766e-01 2.01788932e-01 -1.57181990e+00 1.02294302e+00
1.38549173e+00 -6.78944945e-01 -2.04811096e-02 4.37364839e-02
8.90486360e-01 -4.72633868e-01 -8.99169743e-01 4.05984521e-01
8.02128792e-01 -7.86365867e-01 1.45499682e+00 -3.54674101e-01
9.94006455e-01 -2.73865819e-01 1.15505129e-01 -1.42457318e+00
-2.87654519e-01 -9.85595822e-01 8.86962414e-02 1.20008242e+00
3.46318722e-01 -2.11418644e-01 9.25963938e-01 5.50481796e-01
-4.89123305e-03 -4.77431566e-01 -4.56338853e-01 -8.47779036e-01
1.16273001e-01 -4.56068069e-01 9.32795763e-01 8.21283042e-01
-4.43531543e-01 1.71038687e-01 -7.41679788e-01 -1.95933267e-01
7.33200371e-01 3.34194928e-01 1.41004658e+00 -6.36594296e-01
-1.62530541e-01 -5.02094388e-01 -4.36646789e-01 -6.67916656e-01
3.44771236e-01 -8.78702164e-01 1.56431317e-01 -1.47096479e+00
-3.51600498e-02 -7.52907932e-01 4.65970337e-01 4.12574708e-01
-1.64984018e-01 7.17107177e-01 7.36534655e-01 -2.01658577e-01
-1.37946188e-01 6.38821065e-01 1.92643237e+00 -3.38065058e-01
-6.42319303e-03 -4.60976988e-01 -5.22792161e-01 1.38754666e-01
5.16608357e-01 -2.21413329e-01 -7.12628663e-01 -9.44845825e-02
-1.11264538e-03 3.69850665e-01 8.53510499e-01 -9.76176560e-01
-2.53182888e-01 -4.15918082e-01 6.08799398e-01 -3.53285044e-01
5.68812907e-01 -5.98558307e-01 7.18852878e-01 3.70571792e-01
-5.15974224e-01 -2.52833486e-01 3.55373502e-01 3.09628278e-01
8.05220678e-02 2.90370762e-01 8.90469909e-01 -2.75856674e-01
-6.31406784e-01 5.00875831e-01 -1.11529663e-01 6.71820119e-02
1.06407738e+00 -1.13545693e-01 -2.26030469e-01 -8.47918332e-01
-5.53715169e-01 -2.23046556e-01 8.92859042e-01 6.95574760e-01
1.12650704e+00 -1.97301745e+00 -1.02079594e+00 6.05807543e-01
2.66101390e-01 -3.11295182e-01 2.25843471e-02 -3.34076375e-01
-7.09384084e-01 3.18670928e-01 -5.84599793e-01 -4.32119489e-01
-9.71425831e-01 7.70757079e-01 2.88679928e-01 -8.88642594e-02
-6.81790769e-01 7.32794404e-01 6.44336343e-01 -6.67003095e-01
-1.70519471e-01 4.96953949e-02 5.56253314e-01 -5.55826664e-01
3.14892113e-01 -9.51522402e-03 -5.13298452e-01 -2.10438907e-01
1.16516583e-01 2.89402366e-01 4.70992088e-01 -8.06387782e-01
8.53494406e-01 3.89375746e-01 2.72430718e-01 -1.61057219e-01
1.09265924e+00 2.22669825e-01 -1.60270894e+00 5.09620190e-01
-8.25453639e-01 -7.61890769e-01 -1.60007253e-01 -1.31925416e+00
-1.23260128e+00 6.95927501e-01 5.15305817e-01 -1.96652472e-01
8.58293295e-01 -1.96817905e-01 8.12274277e-01 1.50617152e-01
6.21540189e-01 -6.37904167e-01 5.29754460e-01 -5.94083853e-02
1.65963233e+00 -9.51497436e-01 -1.82954639e-01 -3.68095160e-01
-6.55829668e-01 6.81044817e-01 7.25711167e-01 -3.95360738e-01
1.28787950e-01 7.62047708e-01 1.66021377e-01 -2.67593741e-01
-6.34986401e-01 1.23002633e-01 5.49821258e-01 9.64877367e-01
1.45666093e-01 2.66554058e-01 2.63247620e-02 2.35132173e-01
-4.03183341e-01 5.24847172e-02 4.04335022e-01 5.10232806e-01
2.25003794e-01 -1.30362332e+00 -3.54609996e-01 1.90664500e-01
-1.14397496e-01 -2.75626242e-01 -6.51330352e-01 1.08529747e+00
7.86128119e-02 6.22074902e-01 -3.84596065e-02 -3.22828501e-01
2.28071660e-01 -1.35078326e-01 9.57409441e-01 -4.12144154e-01
-4.63549316e-01 -3.52304906e-01 1.02581836e-01 -7.77578890e-01
1.42354041e-01 -6.17644154e-02 -8.55469942e-01 -1.69286698e-01
3.15808430e-02 -1.56691879e-01 8.12046349e-01 2.27708340e-01
5.55757999e-01 6.23240232e-01 6.37892663e-01 -1.24144971e+00
-1.18423644e-02 -6.05488181e-01 -4.06639546e-01 8.08579862e-01
1.48096681e-01 -3.80834311e-01 -1.06338352e-01 3.14873368e-01]
|
[11.650811195373535, -0.43018606305122375]
|
59d005a9-d5cc-455d-bdfa-1974b92c3d17
|
sparse-group-learning-with-lipschitz-loss
|
1910.08880
| null |
https://arxiv.org/abs/1910.08880v7
|
https://arxiv.org/pdf/1910.08880v7.pdf
|
Improved error rates for sparse (group) learning with Lipschitz loss functions
|
We study a family of sparse estimators defined as minimizers of some empirical Lipschitz loss function -- which include the hinge loss, the logistic loss and the quantile regression loss -- with a convex, sparse or group-sparse regularization. In particular, we consider the L1 norm on the coefficients, its sorted Slope version, and the Group L1-L2 extension. We propose a new theoretical framework that uses common assumptions in the literature to simultaneously derive new high-dimensional L2 estimation upper bounds for all three regularization schemes. %, and to improve over existing results. For L1 and Slope regularizations, our bounds scale as $(k^*/n) \log(p/k^*)$ -- $n\times p$ is the size of the design matrix and $k^*$ the dimension of the theoretical loss minimizer $\B{\beta}^*$ -- and match the optimal minimax rate achieved for the least-squares case. For Group L1-L2 regularization, our bounds scale as $(s^*/n) \log\left( G / s^* \right) + m^* / n$ -- $G$ is the total number of groups and $m^*$ the number of coefficients in the $s^*$ groups which contain $\B{\beta}^*$ -- and improve over the least-squares case. We show that, when the signal is strongly group-sparse, Group L1-L2 is superior to L1 and Slope. In addition, we adapt our approach to the sub-Gaussian linear regression framework and reach the optimal minimax rate for Lasso, and an improved rate for Group-Lasso. Finally, we release an accelerated proximal algorithm that computes the nine main convex estimators of interest when the number of variables is of the order of $100,000s$.
|
['Antoine Dedieu']
|
2019-10-20
| null | null | null | null |
['l2-regularization']
|
['methodology']
|
[ 1.71116471e-01 3.44078302e-01 -3.86937171e-01 -3.49249870e-01
-1.44110107e+00 -2.53134340e-01 -3.87906611e-01 1.41461298e-01
-4.67279166e-01 1.00664890e+00 -8.32345635e-02 -2.55950391e-01
-4.97816443e-01 -5.99640429e-01 -1.10827231e+00 -9.88776624e-01
-8.21249962e-01 1.54403061e-01 -2.88263649e-01 -5.60050905e-02
2.90861934e-01 2.26341948e-01 -1.06494117e+00 -3.95755231e-01
9.91222799e-01 1.47424567e+00 -2.73382425e-01 4.38724756e-01
2.59412944e-01 5.37866533e-01 1.55359134e-02 -2.62875289e-01
7.32532501e-01 -6.42320037e-01 -3.01716089e-01 -7.62566999e-02
6.58845961e-01 1.50943547e-01 -5.36891669e-02 1.39210343e+00
5.29102623e-01 2.85174042e-01 7.69062102e-01 -1.04943800e+00
-1.92065343e-01 5.27462721e-01 -1.20285380e+00 -5.77877238e-02
5.35937250e-02 -2.62182653e-01 9.46916699e-01 -8.76822829e-01
4.07002270e-01 1.11125028e+00 1.13157606e+00 2.38920957e-01
-1.56060231e+00 -9.80779827e-01 1.95605204e-01 -7.17566907e-01
-1.58441043e+00 -4.96277630e-01 4.86133337e-01 -6.75276458e-01
6.49891376e-01 1.90076426e-01 6.28893673e-02 3.44098300e-01
-5.32772988e-02 4.15268213e-01 9.27015960e-01 -5.03438890e-01
2.40750328e-01 9.79134515e-02 2.29905143e-01 1.23750317e+00
5.44495344e-01 4.80502732e-02 -3.98973554e-01 -4.86317307e-01
1.01530337e+00 -1.07613541e-01 -4.30634081e-01 -2.03065917e-01
-8.43520343e-01 1.24696171e+00 1.86778918e-01 1.52366683e-01
-2.59757996e-01 4.94111270e-01 4.48213309e-01 3.89041245e-01
8.59003961e-01 3.98530841e-01 -4.54235584e-01 6.70088232e-02
-1.39456928e+00 2.26803601e-01 8.98339570e-01 1.25062764e+00
9.90543127e-01 2.82747149e-01 1.51139840e-01 9.66995656e-01
1.81104138e-01 8.23768198e-01 -5.32185882e-02 -1.35600185e+00
9.63198423e-01 3.13434213e-01 1.84550092e-01 -1.15574789e+00
-3.16158175e-01 -5.12079418e-01 -1.17769456e+00 8.33037496e-02
6.82729304e-01 -3.64843398e-01 -4.05098379e-01 2.11829734e+00
3.52061354e-03 -2.30247676e-02 -3.53212327e-01 6.56370223e-01
-3.48019367e-03 3.95341009e-01 -6.63643703e-02 -1.03540492e+00
9.97552991e-01 -6.00330889e-01 -4.83235270e-01 -2.47523606e-01
9.36008513e-01 -5.97782850e-01 9.83226001e-01 3.42485547e-01
-1.60128438e+00 -2.99066603e-01 -6.92080498e-01 2.25907639e-01
1.42460778e-01 5.74998371e-02 5.92192590e-01 5.60562611e-01
-9.04216826e-01 7.73571432e-01 -6.67026579e-01 5.07599413e-02
3.70356530e-01 6.06669128e-01 -4.23148394e-01 -1.74358934e-01
-6.76247478e-01 2.38938138e-01 -1.77748442e-01 1.51486278e-01
-6.54558301e-01 -1.02964389e+00 -9.05469596e-01 2.45381221e-02
3.93259168e-01 -4.44261044e-01 5.40528417e-01 -7.58567989e-01
-9.24324334e-01 8.22230101e-01 -3.62590015e-01 -4.19036776e-01
6.33378267e-01 -8.09578449e-02 9.93000418e-02 -3.55398916e-02
4.96054500e-01 1.36391316e-02 9.37764823e-01 -9.94633555e-01
-5.15339732e-01 -6.62619531e-01 -3.79664838e-01 -3.19855586e-02
7.80558912e-03 4.86600213e-02 -2.00887829e-01 -8.11405659e-01
4.65085268e-01 -1.00705934e+00 -7.19730914e-01 5.70649989e-02
-3.24582219e-01 1.42773837e-01 2.38407969e-01 -7.57645905e-01
1.58975995e+00 -2.57496262e+00 1.71505123e-01 7.65563726e-01
2.54665732e-01 -2.29101121e-01 1.37386888e-01 1.28162205e-01
-3.05119753e-01 3.11988115e-01 -6.48549736e-01 -4.79460508e-01
-6.92700073e-02 -3.64712887e-02 -1.10255711e-01 1.01497352e+00
-4.12269950e-01 3.16259623e-01 -6.21935964e-01 -3.61154556e-01
-3.77958179e-01 2.30966046e-01 -7.03005731e-01 5.23764156e-02
8.70925188e-02 9.29294154e-02 -6.22468889e-01 6.18761003e-01
8.96371663e-01 -4.25217211e-01 1.55632868e-01 -1.14524305e-01
-9.74532962e-02 -2.47945547e-01 -1.57873034e+00 1.48175395e+00
-3.98497671e-01 2.76109815e-01 9.72076595e-01 -1.53798306e+00
1.04090750e+00 1.64168373e-01 9.01171327e-01 -2.21354663e-01
-1.37970615e-02 9.14635539e-01 -5.89635015e-01 -3.78521234e-01
-5.24995700e-02 -6.92935944e-01 -2.32957259e-01 2.15664163e-01
-1.68581977e-01 -1.33341672e-02 2.43086711e-01 8.84539187e-02
1.05928910e+00 -2.47037575e-01 3.87662590e-01 -7.55649805e-01
3.67319822e-01 -3.42745870e-01 8.77935827e-01 8.49778652e-01
-1.86509773e-01 4.83687729e-01 1.07028258e+00 -1.76067621e-01
-9.17772889e-01 -8.95781517e-01 -3.14692050e-01 1.19624865e+00
-6.91585690e-02 -1.23371340e-01 -7.43875086e-01 -6.11561418e-01
4.18892145e-01 5.41366756e-01 -7.37522602e-01 1.85162753e-01
-7.59338677e-01 -1.21135235e+00 4.17095959e-01 4.27871555e-01
3.44826400e-01 -5.61517656e-01 6.14297055e-02 2.31723547e-01
1.23648401e-02 -7.82911003e-01 -9.63311493e-01 5.42121351e-01
-1.14157093e+00 -1.11016738e+00 -8.59402716e-01 -9.12154973e-01
1.15080047e+00 -1.77555010e-01 9.35097158e-01 -1.86816737e-01
-2.88597226e-01 3.45656812e-01 -6.40603900e-02 -1.75452128e-01
1.61051199e-01 -1.68568879e-01 2.50298589e-01 5.13720512e-02
-1.48051098e-01 -7.52079546e-01 -7.98719466e-01 3.59109133e-01
-4.88007993e-01 -6.07477784e-01 2.03037366e-01 9.24643219e-01
9.80348766e-01 -1.14942864e-01 5.71519911e-01 -1.18584609e+00
3.73015553e-01 -6.00823820e-01 -8.98196757e-01 7.66235357e-03
-8.79691899e-01 1.01125129e-02 6.74937189e-01 -3.86689633e-01
-4.35049325e-01 -3.36709917e-02 1.71556428e-01 -5.93294740e-01
6.11461103e-01 7.16412306e-01 1.09201610e-01 -4.29644614e-01
9.28872108e-01 -6.45602643e-02 1.20005488e-01 -5.47296286e-01
2.56292135e-01 3.82604361e-01 2.12785333e-01 -8.81808758e-01
5.35123527e-01 5.97047269e-01 3.95102054e-01 -8.75918150e-01
-9.81318831e-01 -3.90966922e-01 -7.54210502e-02 4.06339526e-01
5.52329063e-01 -9.94736314e-01 -8.64975274e-01 -7.67659023e-02
-6.11848593e-01 -3.95091921e-01 -7.66778529e-01 8.12097251e-01
-1.12371397e+00 3.66239190e-01 -5.01027226e-01 -1.14601994e+00
-3.56806517e-01 -1.12827778e+00 9.14134145e-01 -1.94836915e-01
1.07096672e-01 -1.04478264e+00 -7.24131316e-02 1.95206329e-01
4.30351883e-01 5.34709394e-01 9.39435542e-01 -4.36975718e-01
-2.63517022e-01 -3.75132710e-01 -4.56139207e-01 8.22431624e-01
-2.36933500e-01 -5.22192836e-01 -2.40705997e-01 -6.78758442e-01
3.38354737e-01 -2.47652307e-01 8.08718026e-01 1.12099695e+00
1.20052075e+00 -8.12045753e-01 -4.34650570e-01 1.09780717e+00
1.73746955e+00 -5.53098805e-02 4.55662161e-01 -1.21832520e-01
4.26971674e-01 3.62964392e-01 4.47574317e-01 7.45112658e-01
-7.71253631e-02 4.62259501e-01 1.56097010e-01 -1.52701303e-01
3.51537913e-01 -8.27057511e-02 4.26813900e-01 9.13590789e-01
-2.22112209e-01 2.14364290e-01 -5.90033948e-01 6.38976455e-01
-1.84203386e+00 -8.29976320e-01 -2.14090899e-01 2.74239087e+00
9.65436995e-01 2.62779705e-02 2.69508660e-01 -1.04756810e-01
6.12052977e-01 1.26589715e-01 -3.00064355e-01 -4.69609499e-01
-2.07148552e-01 6.04791284e-01 1.12972414e+00 9.36548054e-01
-1.00185966e+00 5.10233819e-01 6.27052164e+00 1.27419937e+00
-8.02209795e-01 1.37901027e-02 7.89030790e-01 -3.95981342e-01
-1.67559922e-01 1.38008177e-01 -9.39707637e-01 6.43837750e-01
9.46125746e-01 -1.66899934e-01 5.45296729e-01 1.13535821e+00
4.82870847e-01 -2.60721177e-01 -1.13457251e+00 1.11601448e+00
5.62071763e-02 -1.06364155e+00 -5.34111977e-01 4.65304911e-01
9.31309462e-01 -6.79603517e-02 1.37137175e-01 1.97824463e-01
2.57577389e-01 -1.18624890e+00 3.63242179e-01 3.31731290e-01
1.37283075e+00 -6.47934496e-01 4.35111791e-01 3.55988711e-01
-1.33324754e+00 -1.99125886e-01 -5.10999501e-01 1.74481690e-01
3.08165699e-01 1.09356523e+00 -2.79607117e-01 3.23236436e-01
7.43261516e-01 6.43704653e-01 1.28326237e-01 7.86393166e-01
1.88924327e-01 7.31107891e-01 -7.70353496e-01 2.66052783e-01
2.06699297e-01 -6.47914708e-01 6.68613732e-01 1.36540580e+00
6.66002631e-01 4.84482825e-01 4.83238697e-01 5.17518103e-01
-2.40956232e-01 4.55109835e-01 -5.62537134e-01 1.08487539e-01
3.01493853e-01 7.15159118e-01 -3.75153244e-01 -1.89459354e-01
-3.51728797e-01 5.13248384e-01 3.40319276e-01 5.58570087e-01
-6.90056145e-01 -7.18513906e-01 6.56565428e-01 5.61465979e-01
3.67850870e-01 -3.32204849e-01 -4.57865387e-01 -1.12693560e+00
1.03807136e-01 -9.01393890e-01 5.70705593e-01 -8.88587013e-02
-1.43295848e+00 1.39691845e-01 4.65709642e-02 -8.87508392e-01
-6.84079379e-02 -6.01075947e-01 -1.12589046e-01 1.08666289e+00
-1.12158763e+00 -5.91586649e-01 3.11384559e-01 6.83666646e-01
1.11042097e-01 -2.41600543e-01 6.11588120e-01 4.90355074e-01
-5.17749488e-01 8.91646087e-01 6.07634425e-01 1.95015110e-02
5.28095543e-01 -1.09680641e+00 -4.64649409e-01 5.78320146e-01
-4.60108906e-01 6.88659370e-01 7.41182268e-01 -5.54784238e-01
-1.02359867e+00 -9.27894771e-01 9.18726146e-01 1.31856993e-01
6.55041575e-01 -2.52541989e-01 -8.58297408e-01 8.97123992e-01
-5.26945591e-01 5.23004413e-01 8.69880736e-01 2.18903407e-01
-3.80565882e-01 -4.73566085e-01 -1.46441996e+00 2.35202298e-01
1.11763787e+00 -3.97063106e-01 2.07642958e-01 6.02874935e-01
3.56322527e-01 -2.75098711e-01 -1.31795144e+00 4.90153164e-01
4.74234879e-01 -9.03907955e-01 9.29183185e-01 -5.46120107e-01
9.39380080e-02 -3.97265628e-02 -8.20441484e-01 -7.12026179e-01
-2.62563407e-01 -1.02296531e+00 1.32695943e-01 7.87484825e-01
6.01417959e-01 -6.46352887e-01 1.01143456e+00 5.78330576e-01
-5.89919873e-02 -8.61528039e-01 -1.34408689e+00 -8.20429683e-01
3.50972533e-01 -2.24997148e-01 -1.43823653e-01 9.35181141e-01
2.54821647e-02 -1.44839451e-01 -5.17387748e-01 -3.37196402e-02
1.12349641e+00 -1.19322658e-01 4.91223902e-01 -9.14474726e-01
-6.62594557e-01 -4.14312452e-01 -1.73955292e-01 -1.02509856e+00
1.36032686e-01 -9.87089634e-01 -6.41374812e-02 -8.16188872e-01
2.16553420e-01 -9.76073205e-01 -2.34336600e-01 3.79020691e-01
-3.05654816e-02 5.88403568e-02 1.32609740e-01 1.87728196e-01
-2.91750103e-01 3.16742867e-01 1.03861833e+00 4.02305871e-02
-4.01218712e-01 1.66134179e-01 -7.53524482e-01 8.59016597e-01
2.98685074e-01 -6.28239453e-01 -1.73967347e-01 -2.76134849e-01
5.44076383e-01 6.09078765e-01 1.14385836e-01 -7.22855091e-01
-7.16770515e-02 -2.88837016e-01 -6.81481138e-02 2.67736576e-02
1.76795602e-01 -6.70832157e-01 9.19415429e-02 2.95147687e-01
-5.05434513e-01 -1.45777330e-01 -1.28547460e-01 6.63966358e-01
-1.79430440e-01 -3.79032284e-01 1.20780325e+00 -2.48742342e-01
1.94503978e-01 4.09719676e-01 3.08712363e-01 4.66473162e-01
8.89806449e-01 -1.04048304e-01 -2.76984926e-02 -5.35443842e-01
-9.28361416e-01 1.92273021e-01 1.99822381e-01 -5.32287121e-01
2.57268310e-01 -1.26898491e+00 -8.81837189e-01 2.82334596e-01
-3.33944827e-01 2.39832833e-01 2.11417019e-01 1.40117490e+00
-5.46743691e-01 1.61104165e-02 4.84356850e-01 -5.07992506e-01
-7.65325367e-01 3.55029434e-01 4.57150429e-01 -4.21142310e-01
-2.63135970e-01 1.20795262e+00 2.93632120e-01 -7.53992200e-02
3.44665766e-01 -1.24037147e-01 5.34302771e-01 1.05583826e-02
2.49978244e-01 8.17429781e-01 -1.53419256e-01 -4.87897992e-01
-3.53481442e-01 7.99002588e-01 2.35177442e-01 -5.79917468e-02
1.51178730e+00 -1.47752970e-01 -5.69417536e-01 2.16220796e-01
1.85657978e+00 4.11133200e-01 -1.32058597e+00 -1.49678588e-01
-2.17770442e-01 -3.65731150e-01 -1.68285266e-01 -3.60254943e-01
-1.20148432e+00 6.46217108e-01 3.74474049e-01 1.47611454e-01
1.05191195e+00 6.11494621e-03 2.85537153e-01 2.67288201e-02
4.57058877e-01 -1.23644543e+00 -2.29272157e-01 3.52853268e-01
9.22725916e-01 -1.18251836e+00 3.50126743e-01 -5.45121431e-01
-2.43716031e-01 7.60896802e-01 1.70861304e-01 -6.26747847e-01
1.07536471e+00 3.13653767e-01 -3.30143094e-01 -4.94097099e-02
-1.62389696e-01 1.99464351e-01 1.74950808e-01 2.54235923e-01
7.05908537e-01 8.45067650e-02 -7.84833074e-01 8.99691820e-01
-1.02136575e-01 2.76797377e-02 1.45422384e-01 7.94825554e-01
-5.80368340e-01 -9.88676727e-01 -4.02823806e-01 8.53257298e-01
-5.59549749e-01 -7.51270801e-02 4.07989264e-01 7.41116524e-01
1.66312784e-01 7.69112587e-01 -1.91888228e-01 1.11911833e-01
4.38352793e-01 5.22041805e-02 4.06122625e-01 -3.69107157e-01
-4.70135450e-01 5.83075166e-01 -3.87484133e-02 -7.41950035e-01
-3.52800339e-01 -7.87839055e-01 -1.09644222e+00 -5.64689100e-01
-4.15665209e-01 4.63189155e-01 6.37625456e-01 5.65423906e-01
9.80955735e-02 -1.98109686e-01 8.60249579e-01 -6.69988513e-01
-1.00754893e+00 -7.34476924e-01 -1.41258204e+00 4.17877674e-01
4.05199677e-01 -4.73917425e-01 -9.99451756e-01 4.62116711e-02]
|
[6.7392897605896, 4.554749965667725]
|
2e37513c-aaf0-40fd-908a-ed2b1fc81f6d
|
mot20-a-benchmark-for-multi-object-tracking
|
2003.09003
| null |
https://arxiv.org/abs/2003.09003v1
|
https://arxiv.org/pdf/2003.09003v1.pdf
|
MOT20: A benchmark for multi object tracking in crowded scenes
|
Standardized benchmarks are crucial for the majority of computer vision applications. Although leaderboards and ranking tables should not be over-claimed, benchmarks often provide the most objective measure of performance and are therefore important guides for research. The benchmark for Multiple Object Tracking, MOTChallenge, was launched with the goal to establish a standardized evaluation of multiple object tracking methods. The challenge focuses on multiple people tracking, since pedestrians are well studied in the tracking community, and precise tracking and detection has high practical relevance. Since the first release, MOT15, MOT16, and MOT17 have tremendously contributed to the community by introducing a clean dataset and precise framework to benchmark multi-object trackers. In this paper, we present our MOT20benchmark, consisting of 8 new sequences depicting very crowded challenging scenes. The benchmark was presented first at the 4thBMTT MOT Challenge Workshop at the Computer Vision and Pattern Recognition Conference (CVPR) 2019, and gives to chance to evaluate state-of-the-art methods for multiple object tracking when handling extremely crowded scenarios.
|
['Laura Leal-Taixé', 'Stefan Roth', 'Anton Milan', 'Hamid Rezatofighi', 'Patrick Dendorfer', 'Javen Shi', 'Konrad Schindler', 'Daniel Cremers', 'Ian Reid']
|
2020-03-19
| null | null | null | null |
['multiple-people-tracking', 'multiple-object-tracking-with-transformer']
|
['computer-vision', 'computer-vision']
|
[-2.17684925e-01 -5.09739339e-01 3.32468003e-02 -1.85556579e-02
-5.43287098e-01 -4.67548013e-01 7.49770641e-01 1.59709468e-01
-7.82714188e-01 9.26237166e-01 -5.79741858e-02 2.03658432e-01
1.29837334e-01 -1.90765977e-01 -6.90368593e-01 -5.63666224e-01
-1.50523067e-01 8.00246477e-01 9.47660446e-01 -1.62781421e-02
-3.38190347e-02 3.76585752e-01 -2.09507442e+00 1.88894019e-01
3.66034299e-01 6.82436883e-01 2.35426486e-01 9.41498876e-01
1.77061200e-01 7.31981218e-01 -9.81672466e-01 -7.31501758e-01
4.77399468e-01 -1.88857511e-01 -4.47520018e-01 -1.55132860e-01
1.39627004e+00 -8.52411613e-02 -6.97577372e-02 1.08534563e+00
7.22833574e-01 8.14633593e-02 3.57345015e-01 -1.78192782e+00
-6.26665130e-02 3.56625319e-01 -6.93690419e-01 7.79522359e-01
2.90737599e-01 4.36858982e-01 9.76920009e-01 -5.59015155e-01
7.83925116e-01 1.37140906e+00 9.50496137e-01 6.74896002e-01
-1.09129000e+00 -6.81033254e-01 2.30200037e-01 5.44704795e-01
-1.20544720e+00 -5.96524000e-01 2.96696164e-02 -9.05688167e-01
5.83406150e-01 6.50778472e-01 8.64916384e-01 1.18808937e+00
-1.83954202e-02 1.09331858e+00 9.88179862e-01 -2.15246648e-01
-6.64658844e-02 3.23063850e-01 3.46958876e-01 5.44858396e-01
9.00226653e-01 3.69381249e-01 -4.14680749e-01 -1.41581342e-01
2.69642711e-01 -1.78675458e-01 -1.97987095e-01 -6.51673675e-01
-1.56519747e+00 6.94859743e-01 3.30993086e-01 3.26703250e-01
-1.51984066e-01 3.34963888e-01 6.01708472e-01 9.76771489e-02
7.73853362e-02 1.63155347e-01 -5.95183037e-02 -2.18949020e-01
-8.69793177e-01 8.97455871e-01 5.54111838e-01 9.07879531e-01
4.49450910e-02 -8.28460902e-02 -7.42764652e-01 5.46490133e-01
4.78045970e-01 9.17842746e-01 9.49421227e-02 -8.81780088e-01
3.76278788e-01 3.04814011e-01 6.43497467e-01 -8.54598761e-01
-4.43661839e-01 -2.72837520e-01 -5.36871910e-01 6.76000655e-01
9.15248632e-01 -2.51556724e-01 -5.20435512e-01 1.51129830e+00
4.00558531e-01 4.27927315e-01 -2.50616550e-01 9.60442483e-01
1.17243040e+00 2.30580315e-01 1.86018527e-01 -2.38570854e-01
1.48425913e+00 -1.07458341e+00 -4.96717602e-01 -1.53751805e-01
3.60571504e-01 -8.70088100e-01 5.57063103e-01 2.76934415e-01
-1.05022085e+00 -6.86074734e-01 -8.92065823e-01 5.12939215e-01
-2.15604052e-01 -1.26568884e-01 1.59488097e-01 8.41254234e-01
-1.05204713e+00 3.34110022e-01 -6.31781638e-01 -8.49910498e-01
3.96908313e-01 1.37715742e-01 -2.14339122e-01 -4.20051701e-02
-7.89594948e-01 1.31370759e+00 2.47676179e-01 -1.09553196e-01
-1.02609110e+00 -6.87251925e-01 -4.11778033e-01 -4.37148541e-01
4.88585383e-01 -6.54425979e-01 1.22086442e+00 -3.16721559e-01
-8.88732910e-01 1.11596608e+00 3.36114387e-03 -8.37249637e-01
1.15229583e+00 -3.22155505e-01 -6.31009161e-01 -2.41347522e-01
4.49305594e-01 7.07195520e-01 7.77801991e-01 -1.30483127e+00
-1.19234085e+00 -9.78865102e-02 -6.58929870e-02 -1.23998746e-01
3.82914320e-02 5.64034939e-01 -5.42958617e-01 -3.03054005e-01
-7.99384475e-01 -9.77017581e-01 -3.50865036e-01 7.75048360e-02
-3.11410576e-01 -1.88855872e-01 1.02118623e+00 -2.16200054e-01
9.22892511e-01 -1.95731819e+00 1.00444853e-01 -3.73292834e-01
5.50853789e-01 7.25763559e-01 -8.87189358e-02 2.92090662e-02
2.56757081e-01 -2.68305212e-01 2.89757341e-01 -6.71671271e-01
2.10978255e-01 8.49319175e-02 -6.00654408e-02 7.73511052e-01
-1.16424844e-01 8.52412283e-01 -1.15846407e+00 -8.35017622e-01
6.04022682e-01 2.26793766e-01 -1.45914853e-01 5.44148274e-02
-3.00475568e-01 6.08939052e-01 -1.99202150e-01 4.28091168e-01
6.25108063e-01 -3.25495154e-01 -2.89064050e-01 -7.40819499e-02
-4.59080577e-01 -5.35931468e-01 -1.60625517e+00 1.04476047e+00
1.92488998e-01 1.04166579e+00 8.08012262e-02 -4.27738160e-01
6.00606620e-01 2.50942588e-01 7.31400669e-01 -5.26299596e-01
1.65585563e-01 4.07461561e-02 1.18974671e-01 -2.64721036e-01
7.77066708e-01 2.37783983e-01 1.34621173e-01 -1.20405443e-01
-1.60065934e-01 1.29172117e-01 1.01622224e+00 1.31372809e-01
1.12606704e+00 7.65306428e-02 3.91923487e-01 -3.51086795e-01
7.44569123e-01 2.38692835e-01 5.68776131e-01 1.06664693e+00
-8.93622756e-01 6.09627306e-01 9.72201757e-04 -7.55739391e-01
-8.50149572e-01 -1.07228851e+00 -1.63313314e-01 1.18147922e+00
3.32280666e-01 -5.73908567e-01 -5.12336433e-01 -5.73921442e-01
1.61738455e-01 2.33583227e-01 -7.69165814e-01 4.78990138e-01
-7.25808263e-01 -7.75626600e-01 6.56272829e-01 3.84815484e-01
4.47042882e-01 -1.09823632e+00 -1.22262466e+00 1.78903252e-01
-1.89720228e-01 -1.41326761e+00 -4.78536785e-01 -2.12306753e-01
-3.89040112e-01 -1.60176206e+00 -1.13256049e+00 -5.16662836e-01
1.81647092e-01 5.96999407e-01 1.40963519e+00 3.65826577e-01
-6.38700008e-01 6.03284061e-01 -1.54774517e-01 -8.68032157e-01
-2.97386199e-01 -1.59986466e-01 2.98203856e-01 5.01232632e-02
2.84339070e-01 5.76578751e-02 -3.92273158e-01 7.75093138e-01
-3.31909448e-01 -2.80829012e-01 3.69977593e-01 4.73231584e-01
4.33321118e-01 -5.23711681e-01 8.65382794e-03 -4.87828553e-01
2.98890352e-01 2.36449372e-02 -1.29774833e+00 3.90142977e-01
-3.05632707e-02 -1.95525318e-01 2.29511037e-01 -3.49482417e-01
-6.75584376e-01 1.12064399e-01 7.04882890e-02 -4.11938995e-01
-3.29766065e-01 -4.84328240e-01 4.87892888e-02 -3.82178605e-01
8.66332114e-01 -5.59940338e-02 -4.39617246e-01 -4.56934154e-01
2.57562220e-01 2.06255719e-01 7.09467530e-01 -4.05417234e-01
1.04183340e+00 6.33965135e-01 2.55381405e-01 -1.08353603e+00
-8.26155186e-01 -9.79550421e-01 -6.19085610e-01 -7.03189015e-01
1.13150561e+00 -1.04900360e+00 -1.11298490e+00 4.83661354e-01
-1.38133824e+00 -1.64626360e-01 -3.34933907e-01 4.06550318e-01
-2.27321744e-01 3.13096911e-01 -6.19942136e-02 -8.97813082e-01
-1.01903409e-01 -1.15445888e+00 1.15120697e+00 3.51228416e-01
-1.67855263e-01 -9.67869639e-01 4.60203737e-01 3.26028734e-01
4.99364763e-01 6.34207785e-01 -8.34651887e-02 -5.54197550e-01
-8.43557477e-01 -2.18914315e-01 -3.18093300e-01 7.14934245e-02
-2.97306776e-01 1.39146134e-01 -1.05383956e+00 -5.41744888e-01
-6.56805336e-01 -1.48252636e-01 9.96682882e-01 6.14640534e-01
5.67956686e-01 3.41036052e-01 -8.28392625e-01 2.30046213e-01
1.29124880e+00 -1.29274711e-01 3.53816956e-01 6.40234292e-01
7.09487200e-01 3.08708340e-01 4.94555593e-01 4.40393925e-01
4.47472572e-01 1.22674060e+00 5.40080607e-01 2.17813477e-01
-5.76179266e-01 3.76795471e-01 4.91311044e-01 3.57587636e-01
-4.82409507e-01 -3.76322150e-01 -9.95856285e-01 4.70801085e-01
-1.92702639e+00 -1.34771860e+00 -7.88778424e-01 2.40052629e+00
1.01076342e-01 4.08003658e-01 9.02243733e-01 -9.65318903e-02
9.28100884e-01 2.27728009e-01 -9.74250287e-02 4.30233359e-01
-1.49099097e-01 -4.19956535e-01 5.13666689e-01 4.70000178e-01
-1.45813477e+00 6.76427186e-01 6.79370975e+00 6.61144197e-01
-6.53872848e-01 4.03129309e-01 -1.09370731e-01 -2.93090820e-01
6.04259908e-01 -4.67057198e-01 -1.65587366e+00 6.30927980e-01
9.55874264e-01 -3.17436606e-01 -2.69223470e-02 7.58571267e-01
-1.17524967e-01 -2.73947477e-01 -1.04389060e+00 1.15662229e+00
4.67446595e-02 -1.25533032e+00 -3.17096889e-01 1.46107189e-02
6.79319322e-01 5.82430065e-01 6.53956309e-02 3.93594056e-01
5.71660578e-01 -8.20432246e-01 1.04090917e+00 4.72343415e-01
2.76052177e-01 -2.60504574e-01 8.86921942e-01 1.84550837e-01
-1.62423205e+00 3.20191719e-02 -5.06390572e-01 1.31092101e-01
3.89386237e-01 2.58774370e-01 -6.25315309e-01 4.94435370e-01
1.09383619e+00 7.79969692e-01 -9.31493461e-01 2.03082061e+00
2.13122666e-01 1.76951513e-01 -2.85228550e-01 -2.90661901e-01
1.27835408e-01 1.76061064e-01 1.04247248e+00 1.37621987e+00
-6.52444363e-02 -4.11045313e-01 5.50420046e-01 2.00854316e-01
1.06955640e-01 -1.31458744e-01 -4.99146044e-01 2.28140026e-01
1.03781201e-01 1.34562778e+00 -9.06966925e-01 -4.41841930e-01
-5.35764754e-01 4.64852154e-01 2.08693936e-01 -7.85590485e-02
-1.14087486e+00 1.67458847e-01 9.87994790e-01 6.43233284e-02
4.97122556e-01 -1.14440791e-01 9.61360857e-02 -9.36331034e-01
-6.53810576e-02 -7.17289388e-01 5.20197392e-01 -4.32595402e-01
-1.16189373e+00 7.62334108e-01 3.11650366e-01 -1.44503224e+00
-5.15954532e-02 -7.52720416e-01 -4.54909742e-01 5.03158510e-01
-1.32187891e+00 -1.16417384e+00 -7.51355827e-01 4.21234906e-01
4.92340505e-01 -3.83702517e-01 5.05456388e-01 5.99394739e-01
-5.28919756e-01 4.93893504e-01 2.04717159e-01 1.28734320e-01
7.97544539e-01 -1.31619048e+00 5.74113667e-01 1.01496840e+00
4.36713964e-01 1.23009332e-01 1.17812777e+00 -5.88671625e-01
-1.10818744e+00 -1.27714574e+00 6.12728715e-01 -1.40367270e+00
6.25918925e-01 -3.89936179e-01 -4.89099562e-01 7.06350327e-01
2.92685330e-01 3.92464727e-01 3.23906332e-01 -9.18514580e-02
-7.70366192e-02 -8.81692469e-02 -8.84371161e-01 4.62446302e-01
1.08739436e+00 2.53364444e-01 -5.33796370e-01 6.55856669e-01
4.35678065e-01 -7.00816095e-01 -5.98015487e-01 2.04213053e-01
5.38912177e-01 -1.21410573e+00 1.23246157e+00 -4.20643032e-01
-2.82272637e-01 -6.97620988e-01 -2.51641810e-01 -1.12171125e+00
-3.20252895e-01 -5.06434262e-01 -2.80353010e-01 1.07815182e+00
1.59784555e-01 -2.76380062e-01 7.95966983e-01 4.99209128e-02
6.34306576e-04 -1.13626622e-01 -1.07967842e+00 -1.37214553e+00
-2.74712861e-01 -2.70630389e-01 3.99948269e-01 3.38423043e-01
-7.66526699e-01 2.45383143e-01 -7.87669539e-01 7.19928890e-02
1.42353499e+00 -9.10560489e-02 1.38748264e+00 -1.73392248e+00
-2.71695495e-01 -8.23125720e-01 -9.66038227e-01 -6.87671065e-01
-3.67589861e-01 -5.94695628e-01 1.87375933e-01 -1.57768917e+00
4.38414961e-01 -6.50209635e-02 -2.21610650e-01 1.02860928e-01
-4.55548555e-01 6.07478142e-01 8.09110284e-01 1.89113334e-01
-1.54669249e+00 1.35709435e-01 1.20320094e+00 -3.37895393e-01
2.73628563e-01 4.18590456e-01 -2.90697128e-01 5.24410486e-01
3.49117428e-01 -6.05870306e-01 1.49737895e-01 -4.26461130e-01
-2.00461432e-01 -3.98702890e-01 8.65889728e-01 -1.73102021e+00
4.89607722e-01 -2.33917162e-02 3.52920264e-01 -8.23728263e-01
6.59414291e-01 -8.41200054e-01 3.40107799e-01 9.53737855e-01
1.03071660e-01 2.75649071e-01 4.15836215e-01 5.72644770e-01
-4.61614877e-02 -1.12333046e-02 1.06639111e+00 -1.93032697e-01
-1.19240761e+00 5.27026057e-01 -3.59886028e-02 4.76727098e-01
1.46206915e+00 -2.02511489e-01 -6.12054586e-01 9.07808095e-02
-6.54775739e-01 5.32063365e-01 3.87321323e-01 6.96201742e-01
2.61219054e-01 -1.33965993e+00 -1.15175617e+00 -1.59107953e-01
4.17902023e-01 -5.34861386e-01 -2.64469534e-02 9.44982290e-01
-3.44941974e-01 6.30472660e-01 -5.41749895e-01 -1.11774099e+00
-1.97439289e+00 5.28855503e-01 2.74677813e-01 -3.70293975e-01
-8.14343452e-01 8.55960786e-01 -1.11174650e-01 -3.19834165e-02
5.44257224e-01 -2.14085370e-01 -3.40781629e-01 3.55723798e-02
1.07797384e+00 7.91012406e-01 -8.97958353e-02 -1.03130519e+00
-7.52507210e-01 6.02706969e-01 3.07147950e-03 5.18739447e-02
1.11369658e+00 1.14564516e-01 2.64007002e-01 5.62247097e-01
6.26406908e-01 -1.32045895e-01 -1.25082505e+00 -1.40410423e-01
4.56691563e-01 -7.82226622e-01 -3.42565060e-01 -6.27454638e-01
-8.27023566e-01 6.94109261e-01 1.01085746e+00 4.04467911e-01
3.37883621e-01 1.29321043e-03 4.85234201e-01 3.29607010e-01
6.99180543e-01 -6.58736408e-01 1.26318708e-01 5.64588428e-01
8.23798954e-01 -1.56185496e+00 5.75665496e-02 -5.85568771e-02
-5.78361869e-01 8.55289459e-01 9.57717180e-01 1.56961501e-01
3.05511832e-01 4.44023192e-01 2.06813931e-01 -1.99288815e-01
-4.76891518e-01 -6.95872486e-01 4.93521929e-01 9.52804208e-01
2.83658177e-01 1.38826305e-02 1.09349461e-02 2.53865998e-02
-7.68915191e-03 -1.13490179e-01 1.74618244e-01 7.56694019e-01
-7.67998695e-01 -1.05263615e+00 -9.84489918e-01 4.07339483e-01
-2.79774517e-01 3.82932007e-01 -3.32329512e-01 1.05255914e+00
3.97170991e-01 1.02264583e+00 -1.32959127e-01 -2.15395600e-01
6.88955128e-01 -3.91652673e-01 7.82964349e-01 -3.55547220e-01
-7.69347310e-01 -4.36826676e-01 1.83925897e-01 -6.57589138e-01
-6.66546524e-01 -1.05752909e+00 -6.23563409e-01 -5.24583757e-01
-2.56414622e-01 3.35340679e-01 5.72502375e-01 1.00571787e+00
5.41586541e-02 6.70906901e-01 -1.48168191e-01 -1.14130354e+00
-2.56641418e-01 -9.39507604e-01 -3.81240427e-01 5.54962695e-01
4.76593047e-01 -1.21492934e+00 2.32915636e-02 -8.84245560e-02]
|
[6.373203277587891, -1.9927502870559692]
|
59594ede-f831-438a-bec2-411847357983
|
unleashing-the-power-of-neural-discourse
|
2011.03203
| null |
https://arxiv.org/abs/2011.03203v1
|
https://arxiv.org/pdf/2011.03203v1.pdf
|
Unleashing the Power of Neural Discourse Parsers -- A Context and Structure Aware Approach Using Large Scale Pretraining
|
RST-based discourse parsing is an important NLP task with numerous downstream applications, such as summarization, machine translation and opinion mining. In this paper, we demonstrate a simple, yet highly accurate discourse parser, incorporating recent contextual language models. Our parser establishes the new state-of-the-art (SOTA) performance for predicting structure and nuclearity on two key RST datasets, RST-DT and Instr-DT. We further demonstrate that pretraining our parser on the recently available large-scale "silver-standard" discourse treebank MEGA-DT provides even larger performance benefits, suggesting a novel and promising research direction in the field of discourse analysis.
|
['Giuseppe Carenini', 'Patrick Huber', 'Grigorii Guz']
|
2020-11-06
| null | null | null | null |
['discourse-parsing']
|
['natural-language-processing']
|
[ 2.70251393e-01 9.33490753e-01 -5.64687550e-01 -3.77474666e-01
-1.31047618e+00 -6.94771349e-01 9.34022725e-01 5.80530584e-01
-3.69750977e-01 1.11468446e+00 1.14840555e+00 -8.44943464e-01
3.45095575e-01 -6.52730823e-01 -5.04965067e-01 -3.41403484e-01
-2.40411267e-01 5.64307332e-01 5.04150331e-01 -6.28919899e-01
4.79058981e-01 -2.08637953e-01 -1.01482880e+00 9.55004811e-01
8.24224472e-01 4.48307008e-01 -7.04351068e-02 9.84554887e-01
-4.10845190e-01 1.22995710e+00 -1.11481190e+00 -6.68735027e-01
-4.88859355e-01 -5.51882148e-01 -1.44631243e+00 -2.37378582e-01
6.98329657e-02 -2.06006303e-01 -4.70839739e-02 5.97643852e-01
4.08016801e-01 5.02998829e-02 4.17078763e-01 -4.13710505e-01
-5.80969572e-01 1.35828245e+00 -7.33165860e-01 5.31085968e-01
8.57583821e-01 -3.04200470e-01 1.44762015e+00 -4.08413321e-01
1.23997879e+00 1.57492626e+00 4.46323723e-01 5.57227015e-01
-8.63011420e-01 -8.28591213e-02 5.05828321e-01 3.18515629e-01
-3.47501874e-01 -4.28419113e-01 8.23893189e-01 -2.92865217e-01
1.44942987e+00 5.07807970e-01 4.86706853e-01 1.35902190e+00
2.24954471e-01 1.25802410e+00 1.10762906e+00 -8.28653634e-01
1.28746286e-01 -3.27393919e-01 5.63560784e-01 4.33180332e-01
-8.21055397e-02 -6.39050245e-01 -7.28384852e-01 -3.28487724e-01
1.53491750e-01 -1.14101350e+00 -2.61299312e-01 4.87466097e-01
-1.26869249e+00 9.76668417e-01 -1.06310673e-01 7.22706020e-01
-1.40818372e-01 -7.49440938e-02 9.57586169e-01 4.82680470e-01
8.11652422e-01 5.65591335e-01 -3.58476460e-01 -7.49394417e-01
-4.19143409e-01 2.70208716e-01 1.30930066e+00 6.85957193e-01
1.01798162e-01 -1.97806597e-01 -2.83340335e-01 5.62537074e-01
1.83266029e-01 7.80372471e-02 5.60899079e-01 -1.15858567e+00
1.04494393e+00 4.81223673e-01 -5.31802438e-02 -7.41104960e-01
-4.50160086e-01 -8.02989770e-03 -5.37591338e-01 -4.65035737e-01
4.78295803e-01 -5.01010299e-01 -3.23301077e-01 1.64591575e+00
5.18971264e-01 -1.94559291e-01 3.81555766e-01 4.89886105e-01
1.18469536e+00 9.51189756e-01 2.16598913e-01 -8.47868502e-01
1.61694098e+00 -8.83365810e-01 -8.99999142e-01 -2.68175513e-01
1.12542582e+00 -7.10740447e-01 8.85645688e-01 3.30520093e-01
-1.34297287e+00 4.28357460e-02 -9.45424378e-01 -4.79315668e-01
1.29854098e-01 3.66914235e-02 7.87279844e-01 5.46008408e-01
-7.63356268e-01 4.96104866e-01 -8.98860037e-01 -1.95692778e-01
3.63571882e-01 -1.36409812e-02 -2.01984778e-01 2.83382416e-01
-1.26963019e+00 1.07190168e+00 5.39149880e-01 -6.39522523e-02
-3.16192627e-01 -4.87264305e-01 -9.49120104e-01 -1.61273092e-01
6.11022234e-01 -4.24427807e-01 1.77123225e+00 -5.78761339e-01
-2.04407191e+00 1.09979737e+00 -7.29298770e-01 -8.64231408e-01
3.75637949e-01 -6.88542426e-01 -1.90024063e-01 2.23727942e-01
1.21856257e-01 4.80064861e-02 2.24199861e-01 -8.59751403e-01
-5.28286099e-01 -3.08005512e-01 3.39451492e-01 3.46866250e-01
-2.08538711e-01 6.79341733e-01 5.64307347e-02 -5.56988001e-01
-8.60692039e-02 -4.41459358e-01 -8.22676197e-02 -7.67189562e-01
-6.62575901e-01 -9.03149128e-01 5.31172156e-01 -9.87547219e-01
1.71328700e+00 -1.64074314e+00 3.93514216e-01 -3.22509140e-01
-1.96437724e-02 3.33498746e-01 2.39745323e-02 7.42922366e-01
1.71731472e-01 3.52272958e-01 -2.59106696e-01 -2.58008778e-01
2.26362981e-02 2.06143394e-01 -4.02500123e-01 1.04284510e-01
3.79579008e-01 1.15470624e+00 -1.00865662e+00 -6.41143024e-01
6.06633676e-03 -1.60418406e-01 -3.90317291e-01 3.11746269e-01
-8.75399530e-01 5.78548312e-01 -6.88272953e-01 3.21324110e-01
-1.31104700e-02 -2.16045246e-01 1.02467191e+00 2.07302794e-01
-3.30296725e-01 1.12559819e+00 -5.47253370e-01 1.73865199e+00
-4.26659100e-02 1.11240113e+00 7.27021396e-02 -1.18354166e+00
6.58802748e-01 4.83350605e-01 6.57735839e-02 -6.65018141e-01
3.35863739e-01 1.02487519e-01 2.58605510e-01 -8.62597823e-01
9.49402750e-01 -6.34793937e-02 -5.87260246e-01 7.07385421e-01
-1.11996040e-01 9.49581265e-02 6.23483777e-01 2.67094940e-01
1.18092668e+00 5.59021048e-02 8.43867898e-01 -3.55549157e-01
6.66527867e-01 3.73623401e-01 4.88382339e-01 3.12714368e-01
-5.60550913e-02 2.83073723e-01 1.04182458e+00 -8.20355490e-02
-9.83242393e-01 -4.89861965e-01 -8.32300708e-02 1.53168356e+00
-3.64087135e-01 -7.53081501e-01 -8.37609649e-01 -7.68226504e-01
-3.95304501e-01 9.59376991e-01 -5.72326422e-01 6.82303190e-01
-1.76146340e+00 -8.62861037e-01 7.31323063e-01 5.65428495e-01
3.83704692e-01 -1.04613221e+00 -3.75571281e-01 4.85565186e-01
-5.25293589e-01 -1.29664958e+00 9.10583511e-02 -1.63932353e-01
-7.62162983e-01 -1.28466368e+00 -4.15348381e-01 -9.77597117e-01
4.13763523e-02 -6.71638250e-02 1.26112270e+00 2.30863504e-02
3.57610941e-01 -1.58940241e-01 -5.23444653e-01 -4.77767825e-01
-1.00101852e+00 4.43031877e-01 -4.72660631e-01 -6.81105554e-01
2.87594318e-01 -3.96793157e-01 -1.43916816e-01 -3.55887383e-01
-3.76637042e-01 4.98199463e-02 8.09125453e-02 8.03606510e-01
2.67102540e-01 -4.85321671e-01 8.98255229e-01 -1.48527884e+00
1.17160451e+00 -4.78502214e-01 -2.12338045e-01 5.19918323e-01
1.37571260e-01 1.18882552e-01 5.13334453e-01 -1.08412378e-01
-1.73359919e+00 -8.76873434e-01 -7.78841197e-01 6.90378010e-01
6.19753846e-05 1.07514656e+00 -1.12588339e-01 7.26418495e-01
6.24852419e-01 -2.39129946e-01 -3.43187779e-01 -4.89849478e-01
6.17300391e-01 6.39391720e-01 7.52098143e-01 -7.95672059e-01
2.88732499e-01 2.81785756e-01 -2.36226544e-01 -1.24585342e+00
-1.15965736e+00 -3.89576346e-01 -7.20264077e-01 6.82478026e-02
1.03238118e+00 -8.11834812e-01 -7.07477212e-01 2.50231981e-01
-1.69026005e+00 -4.34286892e-01 -2.20531255e-01 1.44095257e-01
-3.93021852e-01 8.21083844e-01 -1.10106087e+00 -9.39310312e-01
-6.65764451e-01 -5.65391243e-01 9.58696127e-01 2.04193026e-01
-7.27948844e-01 -1.19485366e+00 4.18592125e-01 5.34192204e-01
-1.15412042e-01 6.60889685e-01 1.22294068e+00 -8.89096260e-01
-2.87167490e-01 4.26463395e-01 -1.97664415e-03 -1.00156637e-02
-2.57186890e-01 2.11460710e-01 -7.28379607e-01 -4.18441966e-02
5.66920713e-02 -6.19973302e-01 8.62903118e-01 3.39011908e-01
6.53309643e-01 -4.78445083e-01 -2.34557003e-01 -4.64192731e-03
9.25012231e-01 5.35282753e-02 6.16712213e-01 6.61606133e-01
3.50548595e-01 8.00393105e-01 8.54329050e-01 2.89096147e-01
8.14438879e-01 1.88737139e-01 -2.08350182e-01 4.66858774e-01
-1.54383004e-01 -1.71999767e-01 5.69291174e-01 1.49380672e+00
-2.62638062e-01 -4.86360490e-01 -1.02426445e+00 7.72736967e-01
-2.12746334e+00 -1.14563835e+00 -5.09239972e-01 1.37805867e+00
1.33925319e+00 4.59915102e-01 3.52301709e-02 6.10789545e-02
5.09844542e-01 7.61407793e-01 -8.87592584e-02 -8.42241526e-01
-4.42194730e-01 4.32470322e-01 -2.16584072e-01 6.37169957e-01
-1.16130781e+00 1.03993154e+00 6.83969831e+00 6.65904939e-01
-7.24523723e-01 2.36606881e-01 6.43997610e-01 1.31673723e-01
-3.57069284e-01 -5.61355315e-02 -8.29436660e-01 1.38493747e-01
1.41418910e+00 -5.83497584e-01 -3.45922291e-01 7.50270963e-01
1.95024356e-01 -5.09316266e-01 -1.08362889e+00 1.68177962e-01
7.61959478e-02 -1.99163973e+00 -7.50084147e-02 -1.28772959e-01
6.76111639e-01 9.30821970e-02 -4.82157648e-01 4.07294035e-01
5.54837823e-01 -8.06800663e-01 6.38318062e-01 -9.72178429e-02
6.82244539e-01 -5.69668770e-01 7.84722686e-01 5.60987532e-01
-9.92242336e-01 8.85897502e-02 -1.38146237e-01 -3.59872043e-01
7.35619724e-01 3.27602744e-01 -8.59546423e-01 8.49260271e-01
1.61901399e-01 6.53688252e-01 -1.62375972e-01 4.06667084e-01
-7.34702528e-01 1.06672454e+00 -1.39868483e-01 -3.40296239e-01
2.72062659e-01 1.61122426e-01 8.53905737e-01 1.57883215e+00
-1.49558485e-01 7.58219421e-01 2.45832235e-01 2.10390657e-01
-4.21794742e-01 1.28250375e-01 -2.30739087e-01 -1.01906084e-01
6.40239716e-01 7.42644668e-01 -4.34723884e-01 -5.66203058e-01
-2.40452781e-01 3.99000883e-01 7.12584496e-01 -3.31355892e-02
-4.78389978e-01 6.75134780e-03 3.32744449e-01 -2.26881549e-01
3.64714324e-01 -4.60201412e-01 -3.39248508e-01 -1.26069903e+00
1.08004555e-01 -1.06135666e+00 5.53605616e-01 -1.42962351e-01
-1.22223175e+00 6.59499228e-01 2.96388745e-01 -6.77420676e-01
-7.64223635e-01 -6.03828728e-01 -1.04496419e+00 4.23150897e-01
-1.78678489e+00 -1.13682127e+00 2.37226754e-01 -3.77701013e-03
1.00120115e+00 -2.81726345e-02 1.00721633e+00 -1.73025697e-01
-7.74610102e-01 3.12244445e-01 2.13604599e-01 4.25410599e-01
6.16948307e-01 -1.44012725e+00 6.99518144e-01 7.23719060e-01
-7.53824562e-02 4.60416019e-01 8.65400910e-01 -6.25252426e-01
-1.29869163e+00 -7.06118584e-01 1.38725805e+00 -5.24467647e-01
1.20703268e+00 -2.19085455e-01 -1.17417204e+00 8.54877949e-01
8.05417955e-01 -9.45532739e-01 8.25069249e-01 6.20311737e-01
-6.35128468e-02 6.44787669e-01 -7.59175360e-01 4.59096462e-01
8.92985642e-01 -5.18875003e-01 -1.55263484e+00 4.35263902e-01
1.06075358e+00 -9.75018740e-01 -1.24245667e+00 2.22975165e-01
3.95387352e-01 -8.88535321e-01 6.59631371e-01 -8.53218019e-01
9.65379059e-01 3.78913701e-01 -4.22666296e-02 -1.02031124e+00
7.99538344e-02 -1.04321134e+00 -6.62291467e-01 1.52542138e+00
6.36495233e-01 -4.37791348e-01 9.91553068e-01 5.79003751e-01
-3.71209919e-01 -8.89210820e-01 -1.15382934e+00 -3.69114012e-01
6.49385989e-01 -4.30800587e-01 4.26669538e-01 1.02577543e+00
9.58069026e-01 9.66115296e-01 2.48049181e-02 -3.67799580e-01
2.46051982e-01 3.07924598e-01 6.74565136e-01 -1.13228917e+00
-2.83144236e-01 -4.54938799e-01 2.87748873e-01 -1.45686150e+00
5.36934078e-01 -6.88396275e-01 -1.02636397e-01 -1.72945845e+00
9.91749540e-02 -1.44148722e-01 4.61161911e-01 -7.20149428e-02
-3.63853276e-01 -3.59621078e-01 1.55344144e-01 3.90585288e-02
-8.84354889e-01 6.63801849e-01 1.18453145e+00 -6.92103431e-02
-3.80118877e-01 -4.06733453e-02 -7.21493363e-01 1.01115572e+00
7.30433524e-01 -1.91194743e-01 -6.10541850e-02 -6.35674179e-01
1.62331522e-01 4.86225098e-01 -3.48089635e-01 -1.69886544e-01
2.20709264e-01 -2.45790720e-01 -1.49993882e-01 -9.60515857e-01
2.53509045e-01 2.44201884e-01 -6.06186152e-01 1.60112113e-01
-6.44772172e-01 1.90494969e-01 2.39444956e-01 4.53466833e-01
-3.91009152e-01 -5.05415022e-01 1.62304789e-01 -3.31004143e-01
-7.31986701e-01 -5.26602745e-01 -5.89876354e-01 6.80616438e-01
8.55619848e-01 6.23443276e-02 -1.14092493e+00 -2.18916506e-01
-5.47114611e-01 3.77410591e-01 7.75018558e-02 1.78183228e-01
3.47786456e-01 -6.98637187e-01 -1.17674839e+00 -5.32176971e-01
-3.28762710e-01 3.48039716e-01 -1.12412954e-02 7.03499496e-01
-3.65715474e-01 7.35105455e-01 1.95378512e-01 -5.57297289e-01
-1.69004953e+00 -1.94886029e-02 -2.64500797e-01 -1.03385472e+00
-7.48577595e-01 9.83328462e-01 -1.18567735e-01 -9.22305211e-02
1.23299837e-01 -5.00997305e-01 -6.06237888e-01 3.17317456e-01
7.19031215e-01 3.43316644e-01 -7.91387483e-02 -6.25195086e-01
-2.39971682e-01 1.08693317e-01 -1.84745982e-01 -2.51479536e-01
1.52983272e+00 -2.31878981e-01 -5.37864625e-01 6.96173191e-01
8.17853868e-01 1.69625252e-01 -6.58398151e-01 -3.11536223e-01
8.61816168e-01 -7.44518936e-02 -3.43032211e-01 -5.33913732e-01
3.99257466e-02 7.48951852e-01 -5.94342768e-01 5.82873881e-01
6.86774373e-01 3.13262939e-01 9.81285930e-01 8.05396259e-01
2.89034128e-01 -1.37225521e+00 1.44992888e-01 1.09261787e+00
1.09982896e+00 -1.11459386e+00 4.05671924e-01 -6.12916410e-01
-8.60856473e-01 1.02082634e+00 4.43377554e-01 -1.10619143e-01
2.80012220e-01 3.57595593e-01 -8.96924436e-02 -3.50989819e-01
-1.28402770e+00 1.26089947e-02 1.36185899e-01 5.67075372e-01
1.09855962e+00 9.35320780e-02 -7.61047721e-01 6.87205732e-01
-6.60824180e-01 -4.18955684e-01 7.81655490e-01 1.09549308e+00
-7.18775153e-01 -1.45563865e+00 -2.22526163e-01 3.22323054e-01
-8.46045911e-01 -6.99877366e-02 -7.43242860e-01 1.02664518e+00
-5.71315765e-01 1.08164763e+00 -1.20284222e-02 1.10952839e-01
2.75367469e-01 1.20743029e-01 7.01737463e-01 -9.16853964e-01
-9.72771645e-01 9.75717530e-02 1.49928296e+00 -1.41690150e-01
-1.01646113e+00 -9.65301514e-01 -1.54418075e+00 -6.22922301e-01
-4.09920901e-01 3.85358453e-01 3.32738996e-01 1.29537809e+00
1.40014052e-01 6.16333604e-01 2.37807423e-01 -7.31063008e-01
-4.95227605e-01 -1.26182401e+00 4.04141247e-02 1.41346931e-01
2.39988625e-01 -1.84209272e-01 2.31133014e-01 4.25119847e-02]
|
[10.810503959655762, 9.457355499267578]
|
7f49623d-58ea-4ba4-b951-a87ecc0f217e
|
aim-2020-challenge-on-rendering-realistic
|
2011.04988
| null |
https://arxiv.org/abs/2011.04988v1
|
https://arxiv.org/pdf/2011.04988v1.pdf
|
AIM 2020 Challenge on Rendering Realistic Bokeh
|
This paper reviews the second AIM realistic bokeh effect rendering challenge and provides the description of the proposed solutions and results. The participating teams were solving a real-world bokeh simulation problem, where the goal was to learn a realistic shallow focus technique using a large-scale EBB! bokeh dataset consisting of 5K shallow / wide depth-of-field image pairs captured using the Canon 7D DSLR camera. The participants had to render bokeh effect based on only one single frame without any additional data from other cameras or sensors. The target metric used in this challenge combined the runtime and the perceptual quality of the solutions measured in the user study. To ensure the efficiency of the submitted models, we measured their runtime on standard desktop CPUs as well as were running the models on smartphone GPUs. The proposed solutions significantly improved the baseline results, defining the state-of-the-art for practical bokeh effect rendering problem.
|
['Jay Zou', 'Hulk Wong', 'Max Zheng', 'Tengyao Wang', 'Xueqin Chen', 'Ge Wu', 'Praseeda S', 'Sanjana A R', 'Minnu A L', 'Saagara M B', 'A. N. Rajagopalan', 'Maitreya Suin', 'Praveen Kandula', 'Kuldeep Purohit', 'Nisarg A. Shah', 'Sourya Dipta Das', 'Saikat Dutta', 'Melvin Kuriakose', 'Hrishikesh P S', 'Jiji C V', 'Densen Puthussery', 'Zhiguo Cao', 'Zijin Wu', 'Ke Xian', 'Xianrui Luo', 'Juewen Peng', 'Jian Cheng', 'Cong Leng', 'Chenghua Li', 'Zhenyu Guo', 'Jiamin Lin', 'Congyu Qiao', 'Ming Qian', 'Radu Timofte', 'Andrey Ignatov']
|
2020-11-10
| null | null | null | null |
['bokeh-effect-rendering']
|
['computer-vision']
|
[ 1.75210699e-01 -2.82696158e-01 6.89999938e-01 -4.47614700e-01
-1.04176342e+00 -3.45937908e-01 4.30707008e-01 -3.46161723e-01
-5.30175209e-01 3.87798429e-01 1.10376358e-01 -8.36971849e-02
-4.48011570e-02 -3.26213270e-01 -8.48594725e-01 -6.12793863e-01
-2.22067133e-01 1.42276272e-01 3.47471893e-01 -1.41796902e-01
3.89300942e-01 2.56594360e-01 -1.90543473e+00 5.40920675e-01
4.06440943e-01 9.83424187e-01 8.83603096e-01 1.74511588e+00
6.63387656e-01 9.84893978e-01 -7.98561335e-01 -2.60509014e-01
4.50001657e-01 -1.32527962e-01 -3.78288507e-01 -2.32122317e-01
1.36973047e+00 -9.21235681e-01 -3.70569915e-01 6.41271770e-01
1.36859918e+00 4.57605541e-01 8.54034796e-02 -1.18273568e+00
-1.37539223e-01 -3.83289605e-01 -6.07959807e-01 4.65363383e-01
7.33959079e-01 3.69773507e-01 6.37781918e-01 -9.59401369e-01
6.23165190e-01 1.36637866e+00 5.45743465e-01 5.05764604e-01
-9.05700386e-01 -7.50891626e-01 -1.66196719e-01 3.52808774e-01
-1.24147785e+00 -6.28660858e-01 2.59799123e-01 -3.11567456e-01
1.24344695e+00 6.51467085e-01 7.49653637e-01 8.83685768e-01
9.13736224e-03 5.94789028e-01 1.36739016e+00 -5.10525942e-01
9.09141153e-02 1.64556921e-01 2.64793914e-02 5.70228636e-01
-4.01121855e-01 4.14525598e-01 -9.98020530e-01 -8.96356106e-02
7.29849100e-01 -5.57118058e-01 -6.21407628e-01 1.00267373e-01
-1.16635871e+00 3.54286313e-01 3.16473603e-01 -2.35814631e-01
2.39583533e-02 4.76401091e-01 3.13680798e-01 3.12746346e-01
6.32154405e-01 1.36102200e-01 -4.42356080e-01 -6.44793093e-01
-8.39731574e-01 5.45392394e-01 3.65168780e-01 7.39721239e-01
4.79047805e-01 -4.52881008e-01 -2.90682852e-01 8.79363596e-01
1.64229989e-01 8.70100021e-01 -9.95050278e-03 -1.48736727e+00
6.88804090e-01 -3.54202509e-01 6.18837893e-01 -1.20820189e+00
-2.34299153e-01 -2.72831079e-02 -4.58764374e-01 5.07880569e-01
5.54376900e-01 -2.25092426e-01 -5.51473379e-01 1.40093422e+00
6.64929271e-01 7.66691923e-01 -2.90154308e-01 1.46076071e+00
9.55617189e-01 1.03601897e+00 -4.33806837e-01 -1.07756272e-01
1.08353972e+00 -1.23498726e+00 -8.55789721e-01 -8.64603817e-02
6.58026218e-01 -1.22691107e+00 1.59995258e+00 9.19934928e-01
-1.40478134e+00 -8.88064444e-01 -9.58961904e-01 -4.73688573e-01
2.07022622e-01 1.60459459e-01 6.18589580e-01 9.41859663e-01
-1.41943145e+00 3.70961487e-01 -5.82516730e-01 -4.63426709e-02
2.47856379e-01 8.05743262e-02 -1.36649758e-01 -6.10554576e-01
-8.67525637e-01 6.60905719e-01 -1.56894267e-01 1.84427619e-01
-1.06190801e+00 -1.40280056e+00 -3.41994047e-01 -2.22721696e-03
2.89087266e-01 -6.32672966e-01 1.61155939e+00 -7.21356571e-01
-1.73389745e+00 1.01573658e+00 -2.02885419e-01 -1.55601308e-01
7.39176929e-01 -8.54658306e-01 -2.35932693e-01 1.15306303e-01
-4.39126313e-01 5.88840783e-01 8.11652601e-01 -1.21063244e+00
-7.78627753e-01 -2.45154619e-01 3.54537010e-01 7.57893384e-01
1.73794165e-01 3.24698627e-01 -6.73443437e-01 -2.14844421e-01
-5.29441297e-01 -9.83266592e-01 -2.82748304e-02 -4.91627085e-04
-1.13386378e-01 1.49383068e-01 8.39968324e-01 -8.89317334e-01
9.55820799e-01 -2.41689730e+00 -9.85041559e-02 -2.14808360e-01
2.88545758e-01 3.17512721e-01 -1.37571573e-01 1.85399085e-01
-8.52020234e-02 -3.81686449e-01 3.61383259e-01 -9.33458805e-01
-7.61664063e-02 -3.02496761e-01 -4.24656749e-01 5.14571369e-01
-8.17237914e-01 5.31152844e-01 -8.89834821e-01 -2.32341260e-01
8.17561984e-01 7.22019970e-01 -6.26392066e-01 8.15433919e-01
5.00039570e-02 5.01598120e-01 2.65496969e-01 6.82817027e-02
1.20291126e+00 1.81552887e-01 -3.02396119e-01 -2.22054087e-02
-1.44921809e-01 3.11576873e-01 -1.45198059e+00 1.96082532e+00
-1.09362340e+00 1.28908908e+00 2.96402246e-01 -3.07718754e-01
4.07143027e-01 1.00278169e-01 3.18180844e-02 -1.15959096e+00
-1.55953869e-01 2.80273408e-01 -2.05311492e-01 -4.22795981e-01
6.37778103e-01 2.29214147e-01 2.27803767e-01 3.79656792e-01
-3.86788607e-01 -6.86880350e-01 -2.19834521e-02 4.24387574e-01
1.11307669e+00 -6.09545223e-02 -6.91138953e-02 -2.07467273e-01
1.63762346e-01 -3.16373974e-01 -2.71656923e-02 8.64461064e-01
8.67851302e-02 1.04214168e+00 1.04038335e-01 -5.05757451e-01
-1.00269973e+00 -1.10786283e+00 1.17221422e-01 1.25636482e+00
2.21466228e-01 -7.46614754e-01 -1.18949842e+00 -1.95161104e-01
-6.13375247e-01 6.04552686e-01 -4.10359263e-01 3.18261445e-01
-4.31777447e-01 -5.87119997e-01 2.83728689e-01 4.90459763e-02
8.68957579e-01 -1.04389620e+00 -1.28918588e+00 -1.95251033e-01
-4.21485990e-01 -1.27511609e+00 -5.87650836e-01 -6.59635782e-01
-2.65930325e-01 -1.12327385e+00 -8.90771389e-01 -2.40107521e-01
1.70113325e-01 8.36014748e-01 1.46540654e+00 1.50931388e-01
-6.58261180e-01 7.40007639e-01 -1.80927709e-01 -3.82258475e-01
-2.96341907e-02 -4.20874596e-01 -2.57367253e-01 -3.87591422e-02
-2.82467335e-01 -4.72128958e-01 -1.22519207e+00 4.51464981e-01
-6.70759082e-01 4.95813936e-01 -8.55551474e-03 3.30598354e-01
3.41039121e-01 1.56713113e-01 -1.92321956e-01 -4.50814873e-01
7.31110573e-01 2.01032564e-01 -8.44343066e-01 1.42978087e-01
-1.37043744e-02 -7.00580537e-01 4.49233800e-01 -4.14756238e-01
-1.44498229e+00 -4.35678005e-01 -1.18738785e-01 -3.64336699e-01
-1.34680690e-02 -2.83541411e-01 -2.54667431e-01 -1.17750131e-01
7.51190841e-01 -5.15989475e-02 -6.64726198e-01 -3.16411704e-01
3.27555865e-01 9.61002529e-01 5.32622576e-01 -4.75359350e-01
1.07728466e-01 5.59436977e-01 1.20334722e-01 -1.10594225e+00
-7.84192741e-01 -5.81487894e-01 -4.26904932e-02 -5.74678540e-01
6.32607698e-01 -1.05622423e+00 -1.12671232e+00 1.10103023e+00
-1.32702231e+00 -1.01531947e+00 -5.27979843e-02 4.01419640e-01
-7.11024463e-01 1.28212512e-01 -4.14831251e-01 -9.47168469e-01
-3.17700386e-01 -1.20587897e+00 1.40947759e+00 2.10739523e-01
1.63445681e-01 -6.06704473e-01 5.25153399e-01 5.82456470e-01
4.35677528e-01 1.07441396e-02 6.34288862e-02 5.39592385e-01
-7.61850119e-01 5.71981305e-03 -5.45122504e-01 4.61908549e-01
-2.80959576e-01 -1.53844446e-01 -1.50032282e+00 -5.11467576e-01
1.35325402e-01 -4.69234973e-01 4.39984679e-01 8.00656736e-01
1.36266398e+00 7.94799179e-02 5.88398166e-02 9.87133682e-01
1.44730186e+00 1.34165436e-01 9.52021778e-01 5.00541367e-02
9.29542303e-01 5.03679335e-01 9.39350545e-01 6.21949017e-01
2.88915902e-01 1.14067388e+00 3.48242283e-01 -2.90473938e-01
-4.74120557e-01 -8.77503678e-02 3.00348938e-01 3.20791990e-01
-2.20861703e-01 -6.12656295e-01 -8.16870332e-01 3.84347230e-01
-1.69085050e+00 -7.01561689e-01 -4.62254912e-01 2.43005824e+00
3.58784854e-01 -3.06600720e-01 -9.24867168e-02 7.79592395e-02
5.03661036e-01 3.36304098e-01 -1.72537819e-01 -7.40837336e-01
1.29859531e-02 3.30864668e-01 4.55647886e-01 9.55582738e-01
-5.77610672e-01 7.27457404e-01 6.12327099e+00 1.06224656e+00
-1.13515198e+00 4.57407922e-01 8.59412909e-01 -7.43943512e-01
1.89129472e-01 -3.15370589e-01 -8.11982214e-01 5.68295896e-01
1.26720345e+00 4.04280633e-01 9.46197808e-01 6.05396390e-01
6.29304647e-01 -8.07856441e-01 -1.00773668e+00 1.63557255e+00
1.87414393e-01 -1.22649968e+00 -7.69128084e-01 7.09511638e-02
7.64450073e-01 2.17694610e-01 3.03871572e-01 1.13733402e-02
2.05658898e-01 -1.16075265e+00 4.88454282e-01 5.13347745e-01
1.21073806e+00 -6.96047664e-01 2.82038569e-01 5.43930471e-01
-9.05568242e-01 1.59643009e-01 -3.63550961e-01 -3.39501381e-01
3.91876966e-01 6.74789429e-01 -7.40761697e-01 2.39293724e-01
1.39666295e+00 4.29596961e-01 -4.51059133e-01 9.74158525e-01
-2.34301221e-02 4.83767271e-01 -7.26367891e-01 3.40591431e-01
-6.65151030e-02 2.53651924e-02 4.33319807e-01 1.40454638e+00
5.24591804e-01 3.25161070e-01 -2.55267471e-01 1.49749905e-01
1.12835996e-01 -6.43957555e-02 -3.90949458e-01 6.21992230e-01
1.75971046e-01 1.20241034e+00 -3.44348252e-01 -3.14089894e-01
-2.73167312e-01 1.17745543e+00 1.76680416e-01 3.75525415e-01
-1.18500066e+00 -1.17484316e-01 7.32310653e-01 4.45191324e-01
6.48029819e-02 -2.46105418e-01 -1.57945864e-02 -9.95641589e-01
1.05575204e-01 -8.89358222e-01 8.77183378e-02 -1.59302735e+00
-6.11170948e-01 5.22923529e-01 2.18386516e-01 -9.11973357e-01
-1.40555324e-02 -5.11066616e-01 -2.49749482e-01 1.15663779e+00
-1.43923843e+00 -8.48321080e-01 -1.03328705e+00 8.19684684e-01
6.43669128e-01 2.73832619e-01 7.30910778e-01 6.16710663e-01
-1.75166592e-01 4.71700609e-01 1.69283479e-01 -5.98826826e-01
1.01290858e+00 -1.16403949e+00 5.84670365e-01 8.32785666e-01
-3.09689194e-02 2.13194191e-01 8.72001588e-01 -2.79915392e-01
-1.08921993e+00 -4.11974549e-01 7.36044168e-01 -5.96565723e-01
3.99309918e-02 -8.71520042e-01 -4.57975954e-01 5.37602186e-01
2.79585719e-01 5.76355219e-01 2.23285794e-01 -4.38858196e-02
6.72522932e-02 -2.32953802e-01 -1.10597801e+00 5.24576306e-01
1.35220027e+00 -3.67206454e-01 9.51669365e-02 2.73245782e-01
6.19986176e-01 -1.13938630e+00 -2.75455475e-01 -8.06001425e-02
6.88150764e-01 -1.84040010e+00 1.31210876e+00 4.96203676e-02
4.39086527e-01 -2.07795545e-01 -4.52645928e-01 -1.51100838e+00
1.26138732e-01 -8.54388654e-01 -1.52238086e-01 9.32982206e-01
-3.31737071e-01 -3.32961500e-01 5.22926748e-01 7.71142244e-01
4.26178724e-02 -4.26963091e-01 -9.82945919e-01 -1.99466541e-01
-5.88771641e-01 -9.36548769e-01 4.30900216e-01 3.67646992e-01
-6.68508172e-01 3.31581503e-01 -9.36497629e-01 2.26320699e-01
7.15363801e-01 -1.93717957e-01 1.51729822e+00 -6.41333222e-01
-7.06130207e-01 1.53668568e-01 -6.65855929e-02 -1.42825675e+00
-1.88286185e-01 -1.36872306e-01 -6.83355108e-02 -1.20248795e+00
1.66972563e-01 -4.32776541e-01 2.07092673e-01 -3.53996664e-01
-2.55463511e-01 4.24442679e-01 7.41066217e-01 -5.32292128e-02
-6.78784609e-01 1.58641160e-01 1.41791105e+00 4.20217812e-01
-1.98665515e-01 4.07081470e-02 -2.40528330e-01 5.90370476e-01
2.16269538e-01 -1.63317710e-01 -1.00122797e+00 -8.63065958e-01
6.58331931e-01 5.57315171e-01 7.99780488e-01 -1.10892749e+00
1.12714224e-05 7.51082152e-02 1.62753373e-01 -4.82845515e-01
9.45476592e-01 -8.36496949e-01 2.71950603e-01 -3.57576832e-02
-2.11517230e-01 -7.16096088e-02 3.23404521e-01 1.97809458e-01
1.89694807e-01 7.17039928e-02 9.35295403e-01 4.28787507e-02
-7.42349267e-01 -5.95683418e-03 1.49607271e-01 1.13812223e-01
8.96037281e-01 -9.14965048e-02 -4.53660637e-01 -8.59696925e-01
-4.39895719e-01 -1.91450223e-01 3.75510931e-01 1.97573468e-01
8.72830629e-01 -9.28795934e-01 -6.82114899e-01 1.74962193e-01
-3.12138081e-01 2.00250521e-01 1.02352285e+00 5.89905739e-01
-9.62885261e-01 2.49859229e-01 -9.42142308e-02 -7.35594690e-01
-1.84708738e+00 3.21581364e-01 6.68505788e-01 -3.84087712e-01
-8.39613199e-01 1.22674596e+00 8.18791986e-01 -3.28887939e-01
3.31611395e-01 -3.26871186e-01 1.87112629e-01 -3.38270128e-01
1.13163221e+00 8.06625664e-01 3.57022315e-01 -4.56836313e-01
-2.98628122e-01 6.44371510e-01 2.72908509e-01 -5.25481462e-01
1.16651022e+00 -6.15504920e-01 3.43244255e-01 2.84551293e-01
1.31720781e+00 1.01987183e-01 -1.61111546e+00 1.47742778e-01
-9.38704252e-01 -1.36965728e+00 5.99729776e-01 -1.10107803e+00
-9.09607172e-01 1.10129821e+00 1.16062474e+00 -3.18208307e-01
1.50661278e+00 -3.27485859e-01 9.29326952e-01 1.19158953e-01
4.44854081e-01 -8.84359479e-01 9.74259302e-02 1.81817517e-01
8.66554916e-01 -1.25652730e+00 -9.43484753e-02 -5.53505242e-01
-6.45547330e-01 4.92675781e-01 5.78832507e-01 -3.19649875e-01
3.96865487e-01 5.35702646e-01 2.13354483e-01 -1.10780351e-01
-8.19229186e-01 2.32761636e-01 3.03125829e-01 7.16370583e-01
2.13928491e-01 -1.89466197e-02 6.82628900e-02 4.36396040e-02
-5.20568013e-01 1.75892323e-01 5.96036136e-01 6.09608948e-01
9.95108709e-02 -5.34781277e-01 -5.88146210e-01 -4.76559848e-02
-4.94423389e-01 -2.24786043e-01 3.51068415e-02 7.48880744e-01
1.59631625e-01 1.15806472e+00 2.07965851e-01 -1.18863463e-01
6.63840532e-01 -7.44334579e-01 9.88561630e-01 -2.53676236e-01
-6.85473442e-01 7.27084428e-02 1.20486744e-01 -1.11155403e+00
-3.64490062e-01 -5.01856089e-01 -7.72011518e-01 -6.80605292e-01
1.74126066e-02 -1.56899348e-01 8.31239045e-01 5.97177982e-01
4.45885897e-01 5.45249581e-01 6.45384848e-01 -1.42666495e+00
4.84380275e-02 -1.14639962e+00 -6.19602561e-01 4.70715523e-01
5.07899940e-01 -4.60076958e-01 -3.29018950e-01 1.57289863e-01]
|
[10.537376403808594, -2.3173508644104004]
|
347d841b-c583-488e-954d-82ec6ea3cebd
|
colored-transparent-object-matting-from-a
|
1910.02222
| null |
https://arxiv.org/abs/1910.02222v1
|
https://arxiv.org/pdf/1910.02222v1.pdf
|
Colored Transparent Object Matting from a Single Image Using Deep Learning
|
This paper proposes a deep learning based method for colored transparent object matting from a single image. Existing approaches for transparent object matting often require multiple images and long processing times, which greatly hinder their applications on real-world transparent objects. The recently proposed TOM-Net can produce a matte for a colorless transparent object from a single image in a single fast feed-forward pass. In this paper, we extend TOM-Net to handle colored transparent object by modeling the intrinsic color of a transparent object with a color filter. We formulate the problem of colored transparent object matting as simultaneously estimating an object mask, a color filter, and a refractive flow field from a single image, and present a deep learning framework for learning this task. We create a large-scale synthetic dataset for training our network. We also capture a real dataset for evaluation. Experiments on both synthetic and real datasets show promising results, which demonstrate the effectiveness of our method.
|
['Kwan-Yee Kenneth Wong', 'Jamal Ahmed Rahim']
|
2019-10-05
| null | null | null | null |
['transparent-objects']
|
['computer-vision']
|
[ 2.59270877e-01 -1.84882641e-01 5.37440360e-01 -3.12251896e-01
-3.60594422e-01 -3.83714706e-01 4.29770127e-02 -8.62211347e-01
-1.49005294e-01 5.88432074e-01 -3.53539228e-01 -2.47136563e-01
5.65665662e-01 -7.57050693e-01 -1.01720583e+00 -7.94100165e-01
3.48804623e-01 3.49738359e-01 4.21089768e-01 3.41681838e-01
-8.42614323e-02 2.26315945e-01 -1.08677924e+00 6.50382757e-01
1.15018213e+00 1.12098515e+00 5.83298504e-01 8.19568992e-01
-3.71455818e-01 1.07861304e+00 -6.04077995e-01 -5.17312586e-01
9.07677770e-01 -2.37548247e-01 -5.35550594e-01 6.36355698e-01
9.86402154e-01 -8.37746859e-01 -4.48863626e-01 9.11746562e-01
1.43136621e-01 -5.32484017e-02 6.41591787e-01 -1.03496218e+00
-1.08547866e+00 4.25544143e-01 -7.28659034e-01 -2.69059509e-01
-9.67541561e-02 5.25447428e-01 4.82604891e-01 -9.28539395e-01
3.18845958e-01 1.28487611e+00 5.56659639e-01 7.50097811e-01
-1.16109300e+00 -5.82105994e-01 3.30834180e-01 -2.86270380e-01
-8.48600864e-01 -3.34161758e-01 8.48922074e-01 -5.51084995e-01
2.37513974e-01 1.26447752e-01 1.14890087e+00 6.42860770e-01
1.37037545e-01 9.14760411e-01 1.62308288e+00 -1.58004567e-01
1.26201302e-01 7.59392828e-02 -2.50216424e-01 1.11683440e+00
5.47744930e-01 2.85493326e-03 -4.95508611e-01 6.93006665e-02
1.17796373e+00 1.48656100e-01 -3.08157295e-01 -3.68000746e-01
-1.19399166e+00 4.40754026e-01 7.51718998e-01 -4.30063337e-01
-7.99960122e-02 6.82190418e-01 -3.53051275e-01 3.46553773e-01
9.04135585e-01 1.33416250e-01 -1.40619516e-01 1.84481204e-01
-8.74243557e-01 -2.87543740e-02 8.93535018e-01 8.77244473e-01
9.12274718e-01 5.44007719e-01 -1.49555549e-01 6.01555288e-01
6.34167075e-01 1.07696748e+00 -1.61656037e-01 -1.12312579e+00
5.86918652e-01 4.66554105e-01 5.75236738e-01 -3.49408925e-01
-8.05341452e-02 -8.23007375e-02 -7.05763161e-01 9.50225353e-01
5.19474804e-01 -4.11946654e-01 -1.43536341e+00 1.19708741e+00
4.89392251e-01 3.07549804e-01 -3.01775709e-02 1.17685497e+00
9.40690935e-01 9.13290083e-01 -6.42786264e-01 3.33845541e-02
8.56628180e-01 -1.54434931e+00 -6.20444059e-01 -4.03320789e-01
4.28739488e-02 -1.04345644e+00 1.26279712e+00 6.67695224e-01
-1.52052689e+00 -3.99832368e-01 -1.07316339e+00 -2.69779205e-01
-6.31776825e-02 4.46945429e-01 1.08000422e+00 6.76866412e-01
-1.01684761e+00 4.79856312e-01 -1.01186097e+00 2.85283208e-01
6.72973514e-01 3.71893078e-01 1.51400017e-02 -2.46229231e-01
-4.56509888e-01 6.05122507e-01 -1.15829967e-02 7.39627898e-01
-1.26631272e+00 -7.55529463e-01 -6.32219076e-01 -1.23539582e-01
2.15057254e-01 -8.86673748e-01 1.14316630e+00 -1.20890963e+00
-2.01395917e+00 7.29027510e-01 -1.44595057e-01 -1.81957752e-01
8.81054759e-01 -6.81825399e-01 -1.32355332e-01 3.31835479e-01
-2.30974421e-01 4.76313382e-01 1.34926200e+00 -1.86121166e+00
-1.47021532e-01 1.03571221e-01 2.06967488e-01 1.09763578e-01
-2.84345210e-01 -7.61234015e-02 -5.05657971e-01 -6.12098157e-01
1.37031138e-01 -9.11322773e-01 -1.68528110e-01 7.91753888e-01
-7.53285050e-01 4.77149874e-01 9.27053034e-01 -7.02347755e-01
4.55118001e-01 -1.90932429e+00 4.96944785e-02 -7.01541528e-02
6.95750535e-01 3.72756720e-01 -3.72180581e-01 -4.36728746e-02
1.99130490e-01 -2.51547039e-01 -4.93604630e-01 -6.35966063e-01
-3.05099696e-01 -6.07302152e-02 -4.37586099e-01 6.55527592e-01
6.16471022e-02 1.07972121e+00 -6.41505301e-01 -3.78254622e-01
1.75098702e-01 4.99820918e-01 -3.67019981e-01 5.22038817e-01
-5.93417108e-01 3.95140350e-01 -2.96011418e-01 7.26555586e-01
1.30314362e+00 -3.11476290e-01 -2.47068793e-01 -3.40886891e-01
-1.56724274e-01 9.87782255e-02 -1.00565100e+00 1.53453398e+00
-6.27993286e-01 8.00946712e-01 3.94820005e-01 -3.10989588e-01
8.55152488e-01 6.43559843e-02 5.87044537e-01 -4.83185470e-01
9.33861658e-02 2.21393123e-01 -1.56740262e-03 -5.34193635e-01
3.19913566e-01 -7.35488608e-02 6.85843110e-01 9.38767254e-01
-4.01953280e-01 -7.78281093e-01 -4.19506058e-02 2.12536559e-01
7.22831011e-01 4.18438166e-01 -7.39124835e-01 -1.73289180e-02
-4.77084816e-02 -2.83962518e-01 3.90128702e-01 6.53821528e-01
1.78510159e-01 1.00200903e+00 -4.41984423e-02 -8.00019383e-01
-1.00718701e+00 -1.45036030e+00 2.75860220e-01 5.16601980e-01
7.30839372e-01 2.87961476e-02 -7.26971745e-01 -4.60673243e-01
8.81955028e-02 6.72368482e-02 -6.09163761e-01 -4.77339812e-02
-7.49021530e-01 -7.99163818e-01 -1.42943710e-01 3.75012130e-01
9.67592418e-01 -1.02240813e+00 -6.70916617e-01 -1.06136225e-01
-9.53319520e-02 -1.30355155e+00 -6.59557283e-01 7.09806941e-03
-1.04835725e+00 -1.04621911e+00 -1.01805067e+00 -8.29819202e-01
8.79489541e-01 9.45295095e-01 1.22102368e+00 3.68837178e-01
-4.20660406e-01 1.57072365e-01 1.04326129e-01 -5.69659591e-01
-3.68069053e-01 -3.57471108e-01 -1.11235920e-02 3.40402067e-01
-5.61141431e-01 -2.99688041e-01 -8.25315118e-01 1.84623316e-01
-9.92605746e-01 7.59601176e-01 6.51358485e-01 5.39092720e-01
3.58730525e-01 -3.86394531e-01 1.20047014e-02 -9.31786299e-01
1.77223191e-01 8.25048536e-02 -1.07025564e+00 3.14473063e-01
-1.43723354e-01 -7.73838535e-02 3.98076177e-01 -6.22728705e-01
-1.52995956e+00 2.83288926e-01 3.37321907e-01 -5.38952172e-01
3.84807467e-01 3.58778238e-02 -6.58618659e-02 -5.96085012e-01
4.42100137e-01 7.78900757e-02 -1.04145169e-01 -4.72785801e-01
4.59314942e-01 3.76973361e-01 3.49329084e-01 -5.71547389e-01
1.63535571e+00 1.15417993e+00 -1.60304949e-01 -5.96624911e-01
-1.28594387e+00 -1.25189885e-01 -5.67957520e-01 -3.93706560e-01
8.54413271e-01 -1.09296203e+00 -9.15318072e-01 9.56202149e-01
-1.18219221e+00 -1.14748788e+00 -2.06155255e-01 4.84764695e-01
-5.53785324e-01 2.60344446e-01 -9.39664662e-01 -7.14912474e-01
-4.97249484e-01 -1.07837141e+00 1.10564923e+00 4.70320523e-01
5.47233462e-01 -9.07291949e-01 -1.10071942e-01 5.22438109e-01
6.01113498e-01 -6.70605302e-02 6.24772787e-01 5.92486084e-01
-1.53558564e+00 1.99925452e-01 -7.00182974e-01 4.20619279e-01
4.53148693e-01 1.86477065e-01 -1.32024670e+00 -1.90680727e-01
2.13818904e-02 -5.80040455e-01 1.38002992e+00 4.65688169e-01
1.31010282e+00 -2.94569522e-01 -9.52187832e-03 1.16020417e+00
1.46590292e+00 -2.03091875e-02 4.96085346e-01 6.23534545e-02
1.36148429e+00 1.81183547e-01 4.10948992e-01 1.57569557e-01
8.60448703e-02 4.67522949e-01 7.86471665e-01 -7.91319191e-01
-6.40281618e-01 1.93335950e-01 5.60560584e-01 7.10741520e-01
-3.18226635e-01 -2.58731961e-01 -3.67470354e-01 1.81302845e-01
-1.55658376e+00 -6.82960987e-01 -3.00012887e-01 2.10836315e+00
1.04044688e+00 6.20165169e-02 5.86246559e-03 -3.44536781e-01
3.75263602e-01 -4.98092845e-02 -7.33364522e-01 -1.09323531e-01
-1.93015605e-01 9.43361595e-02 4.44369555e-01 6.44978464e-01
-1.05326939e+00 1.24782526e+00 6.54671621e+00 1.81903243e-01
-1.52968204e+00 -2.18238421e-02 6.86957181e-01 -1.21747255e-01
-6.05392635e-01 -1.71015128e-01 -3.67755353e-01 5.13564587e-01
3.58925223e-01 1.48468032e-01 7.69705534e-01 4.71516609e-01
2.20175594e-01 -3.30417544e-01 -9.31957841e-01 1.07613301e+00
5.84994815e-02 -1.46140671e+00 2.00741842e-01 1.79172959e-02
1.13310909e+00 3.09195697e-01 4.98752058e-01 -4.94919270e-01
6.27342641e-01 -8.04247916e-01 1.06421959e+00 7.11751699e-01
7.39373863e-01 -5.35308272e-02 1.13977231e-01 -1.92159027e-01
-1.11901081e+00 1.12233356e-01 -5.43312013e-01 9.79499742e-02
5.89495152e-02 8.51279795e-01 -5.60050189e-01 2.12799057e-01
6.90645874e-01 7.88994968e-01 -6.17684364e-01 1.66042721e+00
-2.26851791e-01 7.62006760e-01 -2.89844483e-01 1.20737731e-01
5.59465867e-03 -6.18506372e-01 2.54078746e-01 8.07331264e-01
1.01009332e-01 -2.81831771e-01 2.71666288e-01 1.53352487e+00
-3.58112931e-01 -4.85213190e-01 -2.59128898e-01 -9.55438688e-02
-1.00476675e-01 1.50922692e+00 -9.58846092e-01 -4.47543174e-01
-4.79477316e-01 1.32099700e+00 2.82204717e-01 9.22649026e-01
-9.30329680e-01 -2.40911096e-01 5.50557196e-01 -4.83319722e-02
1.88029304e-01 -4.63774502e-01 -5.15530169e-01 -1.58672416e+00
3.50881636e-01 -4.74586934e-01 -3.45850974e-01 -1.33243716e+00
-1.22321975e+00 4.78047222e-01 -4.39047396e-01 -1.55214858e+00
6.53325558e-01 -9.66046214e-01 -7.67646015e-01 9.74718332e-01
-1.72214997e+00 -1.41344213e+00 -7.89671898e-01 6.45602584e-01
4.13489729e-01 1.38024688e-01 4.17193443e-01 1.45067051e-01
-3.62929016e-01 3.87005620e-02 1.38628989e-01 2.41594732e-01
8.50830853e-01 -1.30555904e+00 6.94276512e-01 1.08616257e+00
2.03407109e-01 4.32787210e-01 4.60346669e-01 -6.61368668e-01
-1.83290195e+00 -1.23041356e+00 -1.29627541e-01 -3.92432034e-01
4.78258550e-01 -7.76205182e-01 -8.41943443e-01 6.88983500e-01
4.21161979e-01 2.87225664e-01 5.06871380e-02 -4.98627096e-01
-4.69159096e-01 -3.31252337e-01 -1.01131380e+00 7.31211185e-01
9.35010016e-01 -2.86544204e-01 2.55435109e-02 6.71637356e-01
9.45912242e-01 -7.57450938e-01 -3.44432384e-01 -3.30063887e-02
6.09536767e-01 -9.17164207e-01 9.89039063e-01 -2.00819135e-01
6.63935781e-01 -5.42389452e-01 4.38772291e-01 -1.53216588e+00
-1.02707250e-02 -9.70545948e-01 -3.50658029e-01 7.84323573e-01
4.62084413e-01 -5.65632701e-01 1.02939868e+00 8.08785915e-01
-4.48893100e-01 -5.24712384e-01 -3.62890631e-01 -6.05620563e-01
-1.18125625e-01 -1.59144372e-01 4.00182784e-01 5.39562285e-01
-7.91108966e-01 -3.99625972e-02 -7.77120650e-01 1.85999662e-01
1.11522079e+00 9.19905186e-01 9.81867552e-01 -1.16310608e+00
-3.67512077e-01 -8.68214443e-02 2.69409623e-02 -1.12293744e+00
-6.39291629e-02 -5.14707148e-01 4.84035581e-01 -1.93825042e+00
6.59818888e-01 -6.37645304e-01 2.17993125e-01 3.57028484e-01
-5.27544737e-01 6.47654235e-01 2.48168662e-01 1.37140810e-01
-5.53227246e-01 7.00537264e-01 2.01500750e+00 -6.75094366e-01
-2.30035648e-01 2.23178625e-01 -4.04041171e-01 7.30060935e-01
8.18054914e-01 -3.14503074e-01 -3.47525299e-01 -1.28385746e+00
3.17263454e-01 -2.13074148e-01 3.77771199e-01 -1.04709566e+00
-6.33389056e-02 -5.74866235e-01 8.57068658e-01 -3.44245255e-01
8.50719571e-01 -8.97761285e-01 -1.91705570e-01 5.02931714e-01
-8.36174712e-02 -3.34625036e-01 2.92995661e-01 5.04326344e-01
1.97499052e-01 7.97088891e-02 9.20950651e-01 -2.42896408e-01
-3.53636920e-01 8.45425904e-01 -1.37097374e-01 -9.29977819e-02
7.31552899e-01 -2.72315741e-01 -7.01715112e-01 -3.70925933e-01
-5.33433318e-01 -3.72254588e-02 8.04237902e-01 2.22359523e-01
9.52176690e-01 -1.03848159e+00 -5.80622792e-01 1.94899514e-01
-1.39205024e-01 2.80355811e-01 1.38378605e-01 5.14074147e-01
-1.11307585e+00 -2.56003380e-01 -1.78592175e-01 -4.52422321e-01
-1.20419443e+00 3.45944762e-01 7.97711670e-01 4.07845914e-01
-1.03960848e+00 1.15344453e+00 6.97541952e-01 -1.91995353e-01
1.34092286e-01 -7.32511580e-01 6.29315317e-01 -7.91281939e-01
5.54576278e-01 8.13786387e-02 -7.78245255e-02 -2.24742621e-01
1.84259593e-01 7.00879693e-01 -7.29936035e-03 -2.63837546e-01
1.31491828e+00 -1.52328327e-01 -4.47895736e-01 6.23660803e-01
8.15430462e-01 1.45909891e-01 -1.89005113e+00 -2.19631523e-01
-6.69908464e-01 -7.50005126e-01 -5.65648973e-02 -6.74266875e-01
-1.89623499e+00 1.14827371e+00 5.20894885e-01 -8.71283785e-02
8.21926236e-01 -1.33226663e-01 1.09227836e+00 6.37624621e-01
1.60864383e-01 -8.75642657e-01 6.76653624e-01 2.41516918e-01
7.44811714e-01 -1.35394144e+00 1.54084995e-01 -8.73389065e-01
-3.98324579e-01 1.22818649e+00 1.09033763e+00 -1.03264771e-01
5.18663585e-01 5.81472516e-01 7.36655533e-01 -1.24474749e-01
-5.18135607e-01 4.39766701e-03 2.73229659e-01 5.21352232e-01
2.80373961e-01 7.90026411e-02 5.91850936e-01 -1.78947389e-01
5.77935725e-02 6.52315021e-02 9.55636740e-01 9.13292408e-01
-5.51377416e-01 -8.54837537e-01 -4.06941116e-01 4.75927025e-01
-3.04817289e-01 -2.89451957e-01 -5.28288603e-01 2.20502734e-01
-1.36100650e-01 8.12976301e-01 5.59967831e-02 -5.83684770e-03
-3.87269437e-01 -4.91101384e-01 9.03754056e-01 -7.62394726e-01
-2.63110638e-01 8.30573738e-02 -2.47981906e-01 -5.84214151e-01
-6.43899798e-01 -7.83323348e-02 -1.12383235e+00 -1.14406981e-01
-3.04072738e-01 -3.27777714e-02 7.16622353e-01 6.23059154e-01
-2.59497799e-02 5.15766025e-01 7.48190522e-01 -1.18550777e+00
-4.05991636e-02 -8.44225407e-01 -7.48020828e-01 2.60528535e-01
6.80300891e-01 -4.45992649e-01 -3.02471370e-01 4.68568176e-01]
|
[10.514371871948242, -1.038562536239624]
|
28b6a075-21bb-4c6e-92e8-f6f57c9f4d7b
|
dibimt-a-novel-benchmark-for-measuring-word
| null | null |
https://aclanthology.org/2022.acl-long.298
|
https://aclanthology.org/2022.acl-long.298.pdf
|
DiBiMT: A Novel Benchmark for Measuring Word Sense Disambiguation Biases in Machine Translation
|
Lexical ambiguity poses one of the greatest challenges in the field of Machine Translation. Over the last few decades, multiple efforts have been undertaken to investigate incorrect translations caused by the polysemous nature of words. Within this body of research, some studies have posited that models pick up semantic biases existing in the training data, thus producing translation errors. In this paper, we present DiBiMT, the first entirely manually-curated evaluation benchmark which enables an extensive study of semantic biases in Machine Translation of nominal and verbal words in five different language combinations, namely, English and one or other of the following languages: Chinese, German, Italian, Russian and Spanish. Furthermore, we test state-of-the-art Machine Translation systems, both commercial and non-commercial ones, against our new test bed and provide a thorough statistical and linguistic analysis of the results. We release DiBiMT at https://nlp.uniroma1.it/dibimt as a closed benchmark with a public leaderboard.
|
['Roberto Navigli', 'Francesco Saina', 'Federico Martelli', 'Niccolò Campolungo']
| null | null | null | null |
acl-2022-5
|
['word-sense-disambiguation']
|
['natural-language-processing']
|
[ 1.63104460e-01 -1.01742469e-01 -3.80031675e-01 -4.11098212e-01
-8.63245964e-01 -8.75649869e-01 9.95756924e-01 1.81151867e-01
-5.63350260e-01 1.20037973e+00 3.26721191e-01 -6.74401700e-01
3.43525499e-01 -4.20704603e-01 -5.79164743e-01 -1.99778900e-01
6.23935997e-01 1.01716971e+00 -1.06435210e-01 -7.93478191e-01
4.42253411e-01 2.10449815e-01 -1.09393418e+00 4.51258540e-01
1.12244785e+00 4.59214211e-01 8.13274309e-02 1.83804452e-01
-2.28356302e-01 2.15915367e-01 -7.12736726e-01 -1.08180058e+00
4.81516942e-02 -5.16811848e-01 -1.05656993e+00 -3.37570190e-01
3.79742533e-01 3.60442400e-01 1.50648817e-01 1.28209209e+00
7.02362001e-01 -2.87625194e-01 4.86711591e-01 -1.13058138e+00
-9.94591236e-01 8.43028069e-01 -1.50774911e-01 4.77147251e-01
6.23690963e-01 1.27139956e-01 1.14145339e+00 -1.37934065e+00
1.06169283e+00 1.41363811e+00 4.58902061e-01 7.40704000e-01
-1.26699817e+00 -6.46729350e-01 -1.45240813e-01 2.76411653e-01
-1.33789742e+00 -6.40405357e-01 5.58004975e-01 -2.61213571e-01
1.21054149e+00 4.53850120e-01 2.00298458e-01 1.80287838e+00
4.15568054e-01 5.26774883e-01 1.81706071e+00 -8.37673306e-01
6.61271811e-02 5.37912667e-01 -1.44037034e-03 2.09396839e-01
4.16868120e-01 7.25867972e-02 -7.92384684e-01 -1.60543561e-01
2.30602473e-01 -6.55051053e-01 -2.51299590e-01 4.25378233e-02
-1.53508663e+00 7.47784436e-01 9.68652125e-03 7.79547215e-01
-1.81456447e-01 -2.84247488e-01 5.81575572e-01 7.47308135e-01
8.08346033e-01 7.95879543e-01 -7.21246064e-01 -3.18423808e-01
-7.02904105e-01 2.44449824e-01 8.95553529e-01 9.39248145e-01
6.47409260e-01 -2.90887296e-01 5.28567582e-02 9.61188495e-01
5.29277697e-02 8.10544193e-01 9.34657872e-01 -4.32031780e-01
9.56363440e-01 4.08085436e-01 2.42200613e-01 -9.06157613e-01
-1.10174939e-01 -6.00670099e-01 -4.92373109e-01 -3.08693409e-01
2.97436744e-01 -1.24564417e-01 -5.75494587e-01 1.81924331e+00
3.49808298e-02 -5.88843644e-01 3.22650224e-01 1.04232597e+00
6.49538636e-01 2.87646979e-01 1.00067593e-02 -1.54792279e-01
1.27216721e+00 -1.13564801e+00 -8.39476287e-01 -7.06269205e-01
8.05864573e-01 -1.46274173e+00 1.41939640e+00 2.09077641e-01
-1.13013947e+00 -2.85684049e-01 -9.63799655e-01 -9.54555050e-02
-6.54548824e-01 1.48918197e-01 3.48362446e-01 4.82199013e-01
-9.43414092e-01 5.55747569e-01 -5.38049340e-01 -7.64611244e-01
-5.62521070e-02 7.43763074e-02 -3.36761981e-01 -1.42680660e-01
-1.70593178e+00 1.61553931e+00 5.00294343e-02 -4.82529663e-02
-3.79368067e-01 -2.87380338e-01 -6.24635398e-01 -4.31235045e-01
2.98046898e-02 -6.02469981e-01 1.33600056e+00 -1.52813590e+00
-1.31560314e+00 1.53725588e+00 -5.68871498e-01 -2.15420842e-01
7.71386445e-01 -3.95225048e-01 -7.01326072e-01 -3.91272098e-01
5.64273119e-01 4.89729434e-01 4.33303326e-01 -9.43813324e-01
-3.01204205e-01 -4.62992072e-01 -1.26879886e-01 3.46023411e-01
-1.30766124e-01 6.42770112e-01 1.38482347e-01 -8.76142859e-01
-1.53749049e-01 -1.09702265e+00 9.05781686e-02 -6.85345411e-01
-4.29112196e-01 -4.67261106e-01 7.24767074e-02 -7.67989814e-01
1.22464132e+00 -2.00320935e+00 6.05498970e-01 -2.21324973e-02
-2.83457011e-01 1.87215000e-01 -1.81170419e-01 8.18663538e-01
2.93480437e-02 4.14751679e-01 -2.57301211e-01 -5.92870593e-01
5.89258410e-02 3.48844200e-01 -4.63064104e-01 2.96507686e-01
2.62717158e-01 1.09363747e+00 -1.03734159e+00 -3.57102692e-01
-9.06594470e-02 1.67951643e-01 -4.64321561e-02 -1.36875540e-01
-2.15852052e-01 5.25285006e-01 -2.06176370e-01 6.35396659e-01
4.07379299e-01 2.66014002e-02 2.58748025e-01 2.93859035e-01
-2.64227241e-01 1.15978885e+00 -5.66228747e-01 1.85686183e+00
-5.99324226e-01 6.43374264e-01 -2.30395064e-01 -5.41158319e-01
7.86638975e-01 3.91502947e-01 -1.67727068e-01 -9.60687935e-01
2.06699684e-01 1.05575943e+00 1.67258322e-01 -1.28032461e-01
6.87616289e-01 -3.07538897e-01 -1.30710721e-01 3.94573033e-01
1.14416987e-01 -1.46522805e-01 4.12608594e-01 -1.84672490e-01
7.14559734e-01 5.69341891e-02 4.37096804e-01 -5.52190661e-01
6.28658175e-01 4.37116086e-01 4.81172353e-01 3.40848595e-01
-2.92721421e-01 6.60153925e-01 1.54841781e-01 -3.65951985e-01
-1.02078629e+00 -9.54306602e-01 -1.77459940e-01 1.14080453e+00
1.56388059e-01 -4.83670861e-01 -7.57181287e-01 -6.71615005e-01
-2.30219305e-01 1.14388561e+00 -4.98654783e-01 -9.70685259e-02
-7.08970129e-01 -8.87695670e-01 7.80920863e-01 1.22351222e-01
2.88060009e-01 -1.11183393e+00 -1.59151405e-01 2.75044054e-01
-7.95630634e-01 -1.27212453e+00 -3.89823794e-01 -7.17903674e-03
-7.64099836e-01 -7.15527117e-01 -3.24452430e-01 -8.31084669e-01
3.56662661e-01 1.61610633e-01 1.75708807e+00 1.62625879e-01
5.47341704e-02 -1.87584579e-01 -6.61340654e-01 -6.19105995e-01
-8.41991007e-01 5.90530097e-01 1.16797715e-01 -2.91580021e-01
9.52395856e-01 -3.01463068e-01 -2.72437274e-01 5.64981878e-01
-6.42231107e-01 8.44899341e-02 5.22799194e-01 6.86580956e-01
2.96427011e-01 -7.30770886e-01 5.04630625e-01 -8.07721257e-01
9.60817218e-01 -6.26230538e-01 -2.76193768e-01 2.95840710e-01
-6.77879214e-01 -4.62398790e-02 5.64061224e-01 -2.45203122e-01
-8.34120095e-01 -5.04953682e-01 -2.48513341e-01 2.11737335e-01
-1.31455779e-01 6.05246365e-01 -1.00443633e-02 1.45125687e-01
9.97837961e-01 3.04186553e-01 -1.99754104e-01 -5.62044919e-01
2.59028047e-01 9.89030957e-01 1.52186543e-01 -6.50814891e-01
6.92878842e-01 1.04731016e-01 -3.96041989e-01 -5.67538619e-01
-8.11202645e-01 -3.04376990e-01 -6.83944941e-01 1.17202803e-01
5.86600482e-01 -8.16859484e-01 2.58893639e-01 3.62359345e-01
-1.51644480e+00 -2.27474973e-01 6.82879016e-02 4.38409120e-01
-3.39442492e-01 1.32365108e-01 -6.13479733e-01 -1.30863190e-01
-5.46497107e-01 -1.28216815e+00 1.18368506e+00 -2.49003470e-01
-1.02533007e+00 -1.21725035e+00 3.52407068e-01 5.20832717e-01
6.24116123e-01 -1.41715124e-01 9.83515024e-01 -9.43234205e-01
-8.85885581e-02 1.07365064e-01 -2.58571748e-02 3.32343638e-01
1.51456431e-01 -1.47609234e-01 -6.92539096e-01 -2.18137994e-01
2.50200219e-02 -4.16316897e-01 5.12367487e-01 -2.62990147e-01
1.84025317e-01 -3.09086472e-01 -1.69459820e-01 2.19244242e-01
1.29966152e+00 -2.49528348e-01 4.72140580e-01 7.85651922e-01
2.97196448e-01 7.12235510e-01 8.15090954e-01 -1.58091456e-01
4.57810104e-01 8.72059047e-01 1.20860338e-01 6.57952577e-02
-2.45384008e-01 -2.05602467e-01 6.73429668e-01 1.23333263e+00
4.11591865e-02 -3.12810063e-01 -1.09429681e+00 7.70177186e-01
-1.62004101e+00 -5.31620085e-01 -3.08790475e-01 2.06558728e+00
1.32965279e+00 -1.75162777e-02 -2.31261134e-01 -1.02800749e-01
7.98729718e-01 6.40097121e-03 6.49089366e-02 -9.29031134e-01
-5.34519196e-01 4.46485728e-01 3.77453089e-01 7.49522388e-01
-8.18809569e-01 1.51521409e+00 5.94232225e+00 7.91905761e-01
-1.45486271e+00 3.90296996e-01 4.47831064e-01 1.51550919e-01
-5.26545644e-01 -3.38269607e-03 -7.95764446e-01 5.44677734e-01
1.15190518e+00 -3.89497548e-01 6.66018605e-01 6.40541196e-01
1.49049208e-01 1.51770309e-01 -1.28466785e+00 6.34172976e-01
2.89053857e-01 -8.91395509e-01 1.70181602e-01 -1.29034773e-01
7.88039207e-01 6.01611018e-01 -2.09297836e-02 1.83467850e-01
2.59072572e-01 -1.12254953e+00 9.09675241e-01 -5.79922087e-02
7.92374790e-01 -4.54062343e-01 1.02829003e+00 3.72301251e-01
-2.45612189e-01 3.44625175e-01 -4.22165930e-01 -3.74726713e-01
3.50714736e-02 4.70342308e-01 -7.64082015e-01 6.29204750e-01
6.01370633e-01 5.81979215e-01 -7.93723285e-01 5.04531324e-01
-6.65549755e-01 7.91306794e-01 -3.72120477e-02 -1.70139983e-01
3.70813876e-01 -2.84179181e-01 8.17809463e-01 1.44351149e+00
3.30953389e-01 -4.80832964e-01 -2.36488059e-02 7.28775978e-01
-3.46789390e-01 6.21783733e-01 -7.04022646e-01 -1.78571895e-01
5.08203268e-01 1.07991433e+00 -4.48187858e-01 -4.21620101e-01
-3.20568472e-01 1.25999606e+00 5.37191331e-01 2.07653657e-01
-6.17106676e-01 -1.27936840e-01 9.56085742e-01 -8.62412080e-02
-2.07924739e-01 -3.21909606e-01 -6.04409039e-01 -1.47973132e+00
2.64911860e-01 -1.40532660e+00 4.59861681e-02 -7.40856946e-01
-1.48009551e+00 9.16637123e-01 -2.71998435e-01 -9.67721462e-01
-4.33492273e-01 -7.47960985e-01 -2.31664121e-01 1.34141707e+00
-1.59910965e+00 -9.57599044e-01 8.41806903e-02 1.84354350e-01
9.19372618e-01 -2.52709359e-01 1.12553918e+00 4.80108619e-01
-4.47188824e-01 5.47492146e-01 1.44350871e-01 1.43489361e-01
1.29973686e+00 -1.09172964e+00 9.34945762e-01 7.67414689e-01
2.66499430e-01 8.49067450e-01 1.01708925e+00 -6.58213854e-01
-1.23220420e+00 -9.51981425e-01 2.03639030e+00 -9.95859444e-01
9.99905169e-01 -5.03593445e-01 -7.27431417e-01 8.31459761e-01
6.84507787e-01 -4.33117896e-01 6.88406169e-01 6.94518089e-02
-4.05858189e-01 3.61092091e-01 -9.73717213e-01 8.33128452e-01
1.22951615e+00 -5.01474917e-01 -1.11015463e+00 5.99594593e-01
7.21219778e-01 -4.54701424e-01 -6.07070565e-01 3.41210216e-01
4.08748657e-01 -7.60734081e-01 7.63776600e-01 -9.43621397e-01
8.07344139e-01 -9.86279175e-02 -3.25284332e-01 -2.02490330e+00
-1.89300239e-01 -5.91802955e-01 5.37351847e-01 1.17847002e+00
8.26855898e-01 -9.47694361e-01 1.67955309e-01 1.02863222e-01
-1.81481019e-01 -6.50496781e-01 -1.12035215e+00 -8.93784106e-01
6.78479671e-01 -3.65171045e-01 4.90752518e-01 1.16757154e+00
1.24621183e-01 8.14123750e-01 -7.68003687e-02 -2.78927922e-01
3.25987816e-01 1.30667672e-01 5.90183079e-01 -1.01252401e+00
-4.24523950e-02 -5.65169573e-01 -2.05842912e-01 -6.54205263e-01
3.76636893e-01 -1.24735212e+00 -1.29323751e-01 -1.34851897e+00
-3.16939503e-02 -3.80176306e-01 -1.27160072e-01 3.26474190e-01
-3.80451024e-01 6.26265287e-01 2.84566134e-02 4.70200181e-01
-2.46974021e-01 4.34624374e-01 1.12426889e+00 2.85651040e-04
1.74858749e-01 -2.98105687e-01 -7.76015878e-01 5.15540004e-01
1.18728936e+00 -7.43252516e-01 -3.05428188e-02 -1.14095640e+00
6.84301257e-01 -5.85707307e-01 1.74436897e-01 -4.73798841e-01
-2.38263652e-01 -2.29849920e-01 7.14224353e-02 -9.95237306e-02
8.85897875e-02 -5.06814718e-01 -9.30576324e-02 4.15103972e-01
-3.74950767e-01 8.98221195e-01 5.92685901e-02 -6.08678460e-02
-3.95796090e-01 -2.36886099e-01 5.12098610e-01 -3.37777555e-01
-3.87738734e-01 -2.99495935e-01 -3.18347305e-01 5.44565499e-01
6.57923341e-01 1.52279109e-01 -5.31040430e-01 -5.68020232e-02
-2.85965592e-01 2.74684746e-02 7.76856422e-01 9.86095905e-01
2.30485231e-01 -1.27409339e+00 -1.14169943e+00 -8.18286315e-02
4.11424816e-01 -7.00801432e-01 -4.09704417e-01 1.18825233e+00
-5.88035345e-01 6.82514966e-01 -3.13743860e-01 -3.08598220e-01
-1.26812184e+00 3.67827326e-01 1.68330684e-01 -2.73658544e-01
-1.01887472e-01 6.04028344e-01 -2.86873370e-01 -8.44426274e-01
-3.59775960e-01 -1.08033754e-01 9.24837887e-02 -5.36788926e-02
2.78508842e-01 3.67755860e-01 5.09987354e-01 -1.03023529e+00
-6.03932679e-01 3.35067779e-01 3.61213498e-02 -4.61121708e-01
8.27890217e-01 -3.23235452e-01 -6.12982333e-01 6.80555999e-01
9.66304600e-01 2.98353493e-01 -9.81624145e-03 -3.65706146e-01
3.99369925e-01 -4.70891953e-01 -3.62344235e-01 -1.36042404e+00
-3.33490282e-01 9.26817954e-01 9.93936434e-02 3.24115492e-02
6.59566164e-01 -1.41226679e-01 8.32169116e-01 2.73975283e-01
7.07003534e-01 -1.24791002e+00 -3.95709783e-01 1.04661655e+00
1.15822601e+00 -1.35970223e+00 -4.09218282e-01 -5.44670522e-01
-6.71438277e-01 8.60854685e-01 3.71048748e-01 6.95913732e-02
2.75735892e-02 -1.19687960e-01 7.13714898e-01 5.99198937e-02
-8.84442091e-01 -3.18944678e-02 2.49232173e-01 3.18093210e-01
1.03531551e+00 3.04379165e-01 -1.22538352e+00 5.14744818e-01
-7.53705561e-01 -1.72052592e-01 4.52081442e-01 8.18464279e-01
-4.05437835e-02 -1.58298254e+00 -4.04924780e-01 -4.45280224e-02
-7.99053788e-01 -6.51927948e-01 -1.24769270e+00 7.46299922e-01
-2.30088010e-02 1.16007864e+00 -2.31704488e-01 -2.63349175e-01
3.16922039e-01 5.20125091e-01 5.24052083e-01 -7.02787519e-01
-9.78043735e-01 -4.02331918e-01 6.60358608e-01 -4.33382422e-01
-2.70297617e-01 -7.76720881e-01 -1.00937521e+00 -4.49602187e-01
-1.43488958e-01 5.14829159e-01 1.04605544e+00 1.07062864e+00
5.45900285e-01 9.37445015e-02 3.38304818e-01 -5.22296429e-01
-8.12280953e-01 -1.43431544e+00 1.57746464e-01 7.11792052e-01
-2.56678201e-02 -5.43757677e-01 -5.31504393e-01 -2.40391433e-01]
|
[11.423468589782715, 10.308006286621094]
|
83d2ffb6-d519-45c7-92cf-337fbc9cf599
|
evaluation-of-deep-neural-networks-for
| null | null |
https://www.sciencedirect.com/science/article/abs/pii/S092523121830924X
|
https://www.sciencedirect.com/science/article/abs/pii/S092523121830924X
|
Evaluation of deep neural networks for traffic sign detection systems
|
Traffic sign detection systems constitute a key component in trending real-world applications, such as autonomous driving, and driver safety and assistance. This paper analyses the state-of-the-art of several object-detection systems (Faster R-CNN, R-FCN, SSD, and YOLO V2) combined with various feature extractors (Resnet V1 50, Resnet V1 101, Inception V2, Inception Resnet V2, Mobilenet V1, and Darknet-19) previously developed by their corresponding authors. We aim to explore the properties of these object-detection models which are modified and specifically adapted to the traffic sign detection problem domain by means of transfer learning. In particular, various publicly available object-detection models that were pre-trained on the Microsoft COCO dataset are fine-tuned on the German Traffic Sign Detection Benchmark dataset. The evaluation and comparison of these models include key metrics, such as the mean average precision (mAP), memory allocation, running time, number of floating point operations, number of parameters of the model, and the effect of traffic sign image sizes. Our findings show that Faster R-CNN Inception Resnet V2 obtains the best mAP, while R-FCN Resnet 101 strikes the best trade-off between accuracy and execution time. YOLO V2 and SSD Mobilenet merit a special mention, in that the former achieves competitive accuracy results and is the second fastest detector, while the latter, is the fastest and the lightest model in terms of memory consumption, making it an optimal choice for deployment in mobile and embedded devices.
|
['Luis M. Soria-Morillo', 'Juan Antonio Álvarez-García', 'Álvaro Arcos-García']
|
2018-11-17
| null | null | null |
neurocomputing-2018-11
|
['traffic-sign-detection']
|
['computer-vision']
|
[-1.28640398e-01 -4.73534971e-01 -3.13238651e-01 -8.14067200e-02
-3.12313020e-01 -1.85805321e-01 6.28782153e-01 -4.96074617e-01
-8.85465205e-01 2.03260601e-01 -3.17632049e-01 -7.40619481e-01
-1.87887073e-01 -5.33374369e-01 -4.18570310e-01 -5.89508712e-01
8.73804837e-02 2.12849155e-01 9.94750619e-01 -3.95842165e-01
5.66112578e-01 7.83669412e-01 -2.11199331e+00 3.71328406e-02
6.11715436e-01 1.36318099e+00 2.13515535e-02 9.47817504e-01
1.30408341e-02 8.33345711e-01 -5.98968387e-01 -6.38665020e-01
4.85136539e-01 6.70368876e-03 -1.04387283e-01 -6.12515748e-01
9.36758339e-01 -3.70546341e-01 -6.00014925e-01 8.51379395e-01
8.05962324e-01 -6.59513101e-02 4.93992478e-01 -1.64485765e+00
-3.15744698e-01 1.78420037e-01 -5.35836995e-01 7.02489972e-01
-3.20542246e-01 7.67362356e-01 6.82634950e-01 -7.91028500e-01
6.40635073e-01 9.67390418e-01 9.42915738e-01 7.44730413e-01
-4.60648507e-01 -1.01408696e+00 -2.66998202e-01 6.91367567e-01
-1.27794719e+00 -3.85283262e-01 3.56732875e-01 -4.04844493e-01
1.15669060e+00 3.20695907e-01 5.22819519e-01 6.60478830e-01
3.61041069e-01 9.17088926e-01 1.18180048e+00 -3.24875355e-01
1.09726541e-01 2.36194447e-01 6.98200643e-01 8.94808114e-01
6.54780924e-01 4.93638366e-01 -5.34680545e-01 6.56504035e-02
5.01493275e-01 -2.97638714e-01 1.09442025e-01 -1.85734451e-01
-8.96519899e-01 5.67746818e-01 5.62986970e-01 3.27951610e-01
-2.88381308e-01 5.41439891e-01 6.99599326e-01 4.50987667e-01
-4.98708338e-02 5.63856773e-02 -4.89126444e-01 -4.98157948e-01
-7.75924861e-01 2.76156068e-01 5.84454000e-01 9.90883589e-01
3.76320720e-01 2.10247293e-01 -4.20243979e-01 5.78648210e-01
2.27292538e-01 8.98653448e-01 6.40569448e-01 -5.04949510e-01
5.79884768e-01 5.96909523e-01 -5.81883751e-02 -5.76384962e-01
-6.28099501e-01 -3.22975844e-01 -3.97900194e-01 7.18568504e-01
6.82970583e-01 -8.14975128e-02 -1.33622992e+00 1.06462121e+00
-9.33224242e-03 1.23014241e-01 6.27392530e-02 8.65392685e-01
1.23090756e+00 2.37767071e-01 3.28710556e-01 4.94448900e-01
1.60213184e+00 -1.06823635e+00 -2.07444534e-01 -3.25859785e-01
7.40890622e-01 -6.96113050e-01 7.11360335e-01 2.15949759e-01
-8.13519835e-01 -9.91664827e-01 -1.22880292e+00 -1.15185268e-01
-8.69180918e-01 7.38738000e-01 5.98025858e-01 1.24352634e+00
-1.05162704e+00 2.18554601e-01 -7.05412388e-01 -6.26309872e-01
4.84653175e-01 4.94301170e-01 -1.53077811e-01 3.78541388e-02
-7.91283250e-01 1.32652545e+00 2.02701494e-01 2.02848837e-01
-4.71344829e-01 -6.11333847e-01 -3.48491371e-01 -8.92091617e-02
2.58515805e-01 -4.60384667e-01 1.22684360e+00 -5.84647894e-01
-1.41143274e+00 1.25342274e+00 4.13141698e-02 -8.36353898e-01
9.96357501e-01 -8.66880417e-02 -9.16853368e-01 -8.78832862e-02
-2.72233129e-01 8.21164608e-01 9.56338584e-01 -7.44713366e-01
-1.14136207e+00 -1.48225054e-01 -1.85291857e-01 -3.69452566e-01
2.38323994e-02 2.91392624e-01 -3.88542861e-01 1.79980472e-02
-2.20424578e-01 -1.04083717e+00 1.40606090e-01 2.44543687e-01
-1.22481562e-01 -5.18570125e-01 1.31448328e+00 -4.09137994e-01
1.13365555e+00 -2.41862369e+00 -7.04154074e-01 3.21248621e-01
2.23115399e-01 1.27506506e+00 -3.14857543e-01 -1.07311765e-02
2.02935394e-02 -1.26994073e-01 3.83876413e-01 1.22695461e-01
1.44960657e-01 9.01970789e-02 -2.22394988e-02 4.61931437e-01
5.78757077e-02 1.28209984e+00 -5.76542199e-01 -5.77443004e-01
5.13882935e-01 3.49970996e-01 -1.69395119e-01 -2.70424038e-01
2.69768238e-01 -1.51379108e-01 -3.97004575e-01 7.82481253e-01
7.60569453e-01 6.85885847e-02 -5.12075543e-01 -3.17550629e-01
-5.40354848e-01 1.16339840e-01 -1.15764964e+00 6.92125559e-01
-3.78788203e-01 1.20320523e+00 -1.21952854e-01 -5.86615443e-01
1.14748764e+00 1.13323517e-02 2.12394208e-01 -1.12695634e+00
4.37186092e-01 6.59912646e-01 3.20046574e-01 -6.32440567e-01
7.65056431e-01 2.66370982e-01 2.57560939e-01 2.84490231e-02
-2.04591230e-01 1.17041320e-01 6.00570500e-01 1.62792411e-02
1.18495321e+00 -2.82067627e-01 2.25556761e-01 2.87735444e-02
6.88137233e-01 -5.01873605e-02 4.72286530e-02 9.41141605e-01
-7.55502582e-01 3.86395216e-01 5.02133727e-01 -4.82097924e-01
-9.35823560e-01 -9.00258183e-01 -2.54399419e-01 1.16363800e+00
2.51134992e-01 1.13386093e-02 -5.46688616e-01 -7.14177907e-01
3.60787809e-01 6.95696235e-01 -6.68003798e-01 -2.92098165e-01
-9.62111413e-01 -5.77533066e-01 1.03860879e+00 9.09821689e-01
9.60534453e-01 -1.32967675e+00 -1.38076806e+00 -8.03216025e-02
3.65298003e-01 -1.28022623e+00 -3.46073359e-01 6.18666671e-02
-6.75365031e-01 -1.38236845e+00 -8.39120030e-01 -6.82622135e-01
3.40037316e-01 4.51759994e-01 8.43782365e-01 2.48355150e-01
-5.95287621e-01 2.95280099e-01 -3.36996704e-01 -7.06516683e-01
-3.69101912e-01 3.01915277e-02 -1.94540128e-01 -3.71690746e-03
7.81718612e-01 2.17990905e-01 -6.55887723e-01 7.31130242e-01
-5.92793763e-01 -4.12394106e-01 1.00115788e+00 6.11786723e-01
-1.03059812e-02 -4.87240553e-01 2.16684833e-01 -5.62274873e-01
3.56596410e-01 1.55889019e-01 -9.30431545e-01 2.68331528e-01
-8.53838921e-01 -7.44925067e-02 1.14448838e-01 -3.39719921e-01
-8.06487858e-01 -4.12672348e-02 -2.33585149e-01 -2.86200613e-01
-1.53851910e-02 -5.15197590e-02 2.06365973e-01 -6.84017301e-01
9.32150543e-01 2.19594598e-01 6.30164072e-02 -3.23702067e-01
3.99645776e-01 9.06216502e-01 7.36075044e-01 1.72077164e-01
7.15810835e-01 4.55092818e-01 8.94289836e-02 -1.09849632e+00
-2.11550236e-01 -8.57269943e-01 -5.73986053e-01 -3.72842073e-01
7.47418106e-01 -5.90878904e-01 -9.22121346e-01 9.83024061e-01
-9.30754304e-01 -1.59027979e-01 -3.22677344e-01 4.34754670e-01
-3.36680651e-01 1.55045107e-01 -2.50725895e-01 -6.55933380e-01
-5.23003757e-01 -9.45280135e-01 1.11066484e+00 4.43489343e-01
1.17371693e-01 -3.92224491e-01 -8.67216289e-02 4.58297998e-01
9.00380254e-01 -4.74478826e-02 5.83653331e-01 -6.11331642e-01
-6.53090477e-01 -6.62205338e-01 -9.78819251e-01 5.17655134e-01
-3.50794405e-01 2.50016540e-01 -9.39565778e-01 -5.61543591e-02
-8.11997473e-01 3.86989908e-03 1.38341296e+00 5.54013968e-01
6.37642682e-01 1.99433342e-01 -5.82088768e-01 5.56783199e-01
1.22325814e+00 2.99798697e-01 9.79023874e-01 5.32737494e-01
5.04847229e-01 1.15019701e-01 5.35892308e-01 5.10776825e-02
2.97791034e-01 7.42958724e-01 3.91556889e-01 -9.85899270e-02
-8.30354452e-01 -1.25574365e-01 3.88360888e-01 3.06690603e-01
-5.14302373e-01 -1.48920208e-01 -8.35058033e-01 3.51968944e-01
-1.59465218e+00 -1.07528985e+00 -5.36842942e-01 2.22044063e+00
-8.13895017e-02 5.89008331e-01 6.52002811e-01 4.59683686e-01
5.75822115e-01 7.50286952e-02 -4.33936656e-01 -5.91613114e-01
-6.00829199e-02 3.33962381e-01 1.19455838e+00 -6.76673725e-02
-1.07282507e+00 1.00547826e+00 6.29413271e+00 8.32234859e-01
-1.60893536e+00 2.75080770e-01 9.09735560e-02 -9.63937584e-03
7.65419960e-01 -3.42947632e-01 -1.42061687e+00 4.87931311e-01
1.12554765e+00 1.50724053e-01 -1.46167781e-02 1.21839690e+00
1.96473822e-02 -1.62292227e-01 -7.36443400e-01 9.68315184e-01
9.98356789e-02 -1.23403239e+00 -2.53990978e-01 3.29811350e-02
4.12163287e-01 6.94615424e-01 1.54867396e-01 6.40950680e-01
-5.06747663e-02 -6.83755100e-01 9.38034356e-01 3.60648543e-01
9.77128088e-01 -4.22112018e-01 1.07469666e+00 -1.17337294e-01
-1.29268837e+00 -4.63418871e-01 -2.70172596e-01 2.64178753e-01
2.50270128e-01 1.96716309e-01 -6.14985347e-01 9.11657959e-02
7.60716736e-01 4.62771297e-01 -9.88044322e-01 1.54702723e+00
-1.94806740e-01 6.51716828e-01 -3.88301045e-01 -7.51165330e-01
4.06871498e-01 2.54068285e-01 6.47755265e-01 1.62830532e+00
2.23439053e-01 -2.64122516e-01 -2.59126663e-01 5.60512781e-01
1.11395426e-01 -4.72430214e-02 -3.58819515e-01 2.54292816e-01
2.13436753e-01 1.29160869e+00 -8.49059105e-01 -7.15334654e-01
-4.20655876e-01 3.37786913e-01 -2.39326060e-01 1.90392077e-01
-1.21397078e+00 -7.06372023e-01 6.54736221e-01 4.55345094e-01
7.67092884e-01 -1.35441497e-01 -3.34939539e-01 -6.25404298e-01
6.15202338e-02 -5.70817530e-01 3.42276365e-01 -7.75268078e-01
-8.49278092e-01 5.08093715e-01 -8.48799869e-02 -1.51707578e+00
3.32144499e-02 -1.43789947e+00 -6.86431050e-01 4.18253303e-01
-1.74828625e+00 -1.15166163e+00 -7.46124566e-01 3.71127516e-01
2.64113873e-01 -4.49496835e-01 3.72748729e-03 7.38172710e-01
-5.46100259e-01 1.08909047e+00 8.93038511e-03 2.89848596e-01
4.16510016e-01 -8.37032735e-01 7.97153175e-01 7.66266465e-01
-7.36527294e-02 9.31196958e-02 3.41029137e-01 -4.11616743e-01
-1.28626168e+00 -1.06034052e+00 9.17216718e-01 -4.11210686e-01
6.58407152e-01 -2.09662486e-02 -5.79059005e-01 3.46280605e-01
-4.18898135e-01 4.51268882e-01 -1.51651561e-01 -3.92199606e-01
-6.97246730e-01 -4.75784451e-01 -1.14386272e+00 4.42242444e-01
1.07503164e+00 -3.36334318e-01 -3.22113693e-01 2.16782570e-01
1.32176830e-02 -5.77882409e-01 -2.50973940e-01 3.54772389e-01
1.12756681e+00 -1.14742112e+00 1.02042055e+00 -4.15655136e-01
5.74388914e-03 -4.17692274e-01 1.13942727e-01 -6.55178547e-01
-2.24826142e-01 -4.36465234e-01 -5.02391346e-02 8.28044057e-01
5.53802609e-01 -1.20636022e+00 9.68761444e-01 5.37387669e-01
-2.14320585e-01 -5.57712376e-01 -1.24254489e+00 -1.33578730e+00
-2.62303442e-01 -6.59387827e-01 3.77454460e-01 1.27349362e-01
-3.81357998e-01 -7.18763769e-02 -1.70237854e-01 -1.94603547e-01
4.04697120e-01 -4.80761826e-02 1.22595954e+00 -1.14485347e+00
8.29203874e-02 -1.07484210e+00 -1.20130062e+00 -1.13738453e+00
-3.32501858e-01 -7.26583838e-01 1.00454144e-01 -1.22145307e+00
-9.03523639e-02 -4.75408345e-01 -3.56039017e-01 5.52022934e-01
-5.43937683e-02 5.76321661e-01 4.24347937e-01 8.01833272e-02
-6.11069679e-01 -1.84244830e-02 1.07645178e+00 -1.35571107e-01
-3.72263044e-01 3.95093888e-01 -1.37069419e-01 5.81649065e-01
5.51976323e-01 -3.10577154e-01 2.78356075e-01 -4.26142663e-02
-7.14711398e-02 -5.14857590e-01 7.62653708e-01 -1.39980364e+00
6.51199996e-01 3.45798641e-01 9.05174017e-02 -8.38980019e-01
2.78440684e-01 -5.74058115e-01 -2.25886375e-01 1.04573512e+00
-2.42761839e-02 8.95463973e-02 3.80379111e-01 1.54272422e-01
-8.25687051e-02 -3.58276844e-01 1.12388647e+00 3.47058386e-01
-1.47808826e+00 1.06527500e-01 -4.40944195e-01 -2.56090946e-02
1.01403940e+00 -8.32252264e-01 -8.07701826e-01 -7.66496807e-02
-1.07842833e-01 4.99668196e-02 -7.77599439e-02 8.35405409e-01
5.56485653e-01 -1.06392694e+00 -8.55762482e-01 3.97048116e-01
4.77356374e-01 -8.09673131e-01 8.91550779e-02 1.12272680e+00
-8.42891097e-01 8.57480705e-01 -4.71046954e-01 -7.06206143e-01
-1.73764980e+00 2.97894299e-01 3.61599684e-01 -1.33977339e-01
-6.86644673e-01 7.40496337e-01 -3.88553232e-01 -1.92827329e-01
3.22483629e-01 -8.71332884e-01 -1.08003698e-01 -6.90964013e-02
6.47125304e-01 1.22795212e+00 4.89727288e-01 -8.57420862e-01
-5.41715026e-01 1.10534334e+00 -9.80487093e-03 3.18309575e-01
7.30890989e-01 3.30122352e-01 3.94016057e-01 -2.77782589e-01
1.06151104e+00 -3.69558960e-01 -9.11236644e-01 7.44268894e-02
1.54267818e-01 -3.53016198e-01 1.65982142e-01 -7.98445344e-01
-1.30545807e+00 6.90551460e-01 1.23661697e+00 -4.16316018e-02
8.93541634e-01 -1.12364605e-01 9.30905223e-01 5.70500374e-01
2.56402999e-01 -1.12326527e+00 -1.06070817e-01 5.30510008e-01
6.13407373e-01 -1.24540269e+00 -7.66354501e-02 -1.87992811e-01
-5.13058364e-01 1.04574192e+00 7.23194540e-01 -2.36137107e-01
8.05485189e-01 2.83487737e-01 2.95692712e-01 -3.61325353e-01
-4.35076356e-01 -7.81193614e-01 6.01052582e-01 6.00608587e-01
8.87617543e-02 1.19429432e-01 -5.16335785e-01 1.06717171e-02
-2.87488699e-01 3.90000820e-01 2.23676652e-01 8.77454638e-01
-6.08046353e-01 -5.64689219e-01 -2.89651543e-01 7.11188912e-01
4.28021466e-03 1.06324494e-01 -3.43307644e-01 1.29923499e+00
4.81260717e-01 7.73720086e-01 -4.00784388e-02 -4.70102429e-01
9.42254543e-01 2.06926260e-02 4.09886777e-01 2.62675136e-02
-8.46569836e-01 -6.79928243e-01 1.51065186e-01 -5.25599718e-01
-1.17462218e-01 -5.95541298e-01 -1.05368078e+00 -3.53980422e-01
-7.24430919e-01 -3.56279075e-01 1.09948504e+00 8.78178239e-01
3.68484110e-01 4.19473559e-01 5.19840829e-02 -7.73400128e-01
-7.11640358e-01 -9.56904531e-01 -4.40195531e-01 1.20489523e-01
2.84043401e-01 -7.45216668e-01 -3.61240745e-01 -2.36448720e-01]
|
[7.9945902824401855, -0.8087912201881409]
|
02017271-cba1-4529-92a2-348ea052a69a
|
ngep-a-graph-based-event-planning-framework
|
2210.10602
| null |
https://arxiv.org/abs/2210.10602v1
|
https://arxiv.org/pdf/2210.10602v1.pdf
|
NGEP: A Graph-based Event Planning Framework for Story Generation
|
To improve the performance of long text generation, recent studies have leveraged automatically planned event structures (i.e. storylines) to guide story generation. Such prior works mostly employ end-to-end neural generation models to predict event sequences for a story. However, such generation models struggle to guarantee the narrative coherence of separate events due to the hallucination problem, and additionally the generated event sequences are often hard to control due to the end-to-end nature of the models. To address these challenges, we propose NGEP, an novel event planning framework which generates an event sequence by performing inference on an automatically constructed event graph and enhances generalisation ability through a neural event advisor. We conduct a range of experiments on multiple criteria, and the results demonstrate that our graph-based neural framework outperforms the state-of-the-art (SOTA) event planning approaches, considering both the performance of event sequence generation and the effectiveness on the downstream task of story generation.
|
['Frank Guerin', 'Chenghua Lin', 'Tyler Loakman', 'Zhihao Zhang', 'Chen Tang']
|
2022-10-19
| null | null | null | null |
['story-generation']
|
['natural-language-processing']
|
[ 4.10356462e-01 7.85665929e-01 -8.45564082e-02 -1.79225609e-01
-8.32708597e-01 -5.15506208e-01 1.29601312e+00 1.64591536e-01
2.05405876e-01 8.90633345e-01 1.23395705e+00 -1.09778620e-01
5.43458536e-02 -1.18908346e+00 -6.90167904e-01 1.22516416e-01
-6.55447543e-02 6.86674893e-01 8.11722651e-02 -3.43071252e-01
2.76719779e-01 -7.06380829e-02 -1.40549433e+00 6.19944334e-01
7.95898914e-01 4.29799765e-01 -2.53826100e-03 7.28718221e-01
-1.13472007e-01 1.45119166e+00 -7.33245134e-01 -5.59508204e-01
-7.97903985e-02 -1.07143784e+00 -8.45844269e-01 1.03173889e-01
-2.53504068e-01 -5.39835334e-01 -7.20959723e-01 3.75500083e-01
6.52215004e-01 6.19336963e-01 9.26333249e-01 -1.36851716e+00
-7.20254660e-01 1.43279135e+00 -1.04121678e-01 -9.25524347e-03
9.61914301e-01 2.99649596e-01 1.37072527e+00 -7.38862157e-01
1.20450139e+00 1.13452995e+00 5.70925593e-01 3.99426520e-01
-1.10139549e+00 -4.38340694e-01 2.53936231e-01 2.58978069e-01
-1.20886540e+00 -5.22888124e-01 8.65406454e-01 -3.26299757e-01
1.38005877e+00 6.77250773e-02 9.36173201e-01 1.64192533e+00
3.65336508e-01 9.91818607e-01 3.91029775e-01 -2.71319002e-01
3.82853210e-01 -6.04332745e-01 -3.61317217e-01 5.66661060e-01
-1.83724701e-01 3.34833205e-01 -9.61982846e-01 1.82023346e-02
7.35909104e-01 -4.23234284e-01 -1.08017251e-01 3.33536863e-01
-1.41681290e+00 1.02758121e+00 3.25075865e-01 -2.10130345e-02
-7.62365222e-01 2.08027676e-01 5.36868274e-01 -1.44693032e-01
4.77783322e-01 8.35692108e-01 1.57201752e-01 -3.32207382e-01
-1.16296446e+00 9.83337402e-01 1.07308555e+00 1.12824643e+00
1.05848603e-01 1.96287721e-01 -1.06807947e+00 4.77254719e-01
9.53202248e-02 -1.71862513e-01 3.48841041e-01 -7.04982102e-01
7.77966738e-01 5.09692252e-01 5.58007024e-02 -9.80662107e-01
-4.22049910e-01 -3.22049946e-01 -5.83678901e-01 -1.79711282e-01
3.51659745e-01 -4.81048971e-01 -9.02657628e-01 1.85691845e+00
1.54351056e-01 4.68193412e-01 1.25343159e-01 6.41236484e-01
1.04228020e+00 1.12038565e+00 2.82084048e-01 -2.56310016e-01
1.02424431e+00 -1.09081852e+00 -8.41920674e-01 -5.13865829e-01
5.58234513e-01 -4.50895578e-01 9.07476485e-01 1.38670668e-01
-1.50520754e+00 -2.95469075e-01 -1.01059461e+00 -4.52506468e-02
-9.89763513e-02 -3.33729148e-01 7.86389709e-01 -1.68990381e-02
-7.41343617e-01 6.34010136e-01 -8.34827662e-01 -2.53560811e-01
3.81401569e-01 -1.81980401e-01 -6.86648488e-02 7.10166339e-03
-1.47023416e+00 9.06169057e-01 1.02371478e+00 -1.50232464e-01
-1.16761088e+00 -8.51352990e-01 -1.10169649e+00 2.33909160e-01
5.16030252e-01 -1.02294242e+00 1.55483866e+00 -3.56150657e-01
-1.53512037e+00 1.91237360e-01 5.32845110e-02 -5.64481616e-01
5.42087078e-01 -1.73866197e-01 -3.47962439e-01 6.47161007e-02
2.84423143e-01 9.21598732e-01 3.42583060e-01 -9.49520051e-01
-6.89754486e-01 2.73236662e-01 4.10499461e-02 5.57332158e-01
1.00840904e-01 3.78657095e-02 -1.29740179e-01 -9.36774373e-01
-3.65335017e-01 -7.02344716e-01 -5.24262846e-01 -7.30769455e-01
-9.32707965e-01 -3.80295277e-01 5.44460952e-01 -7.00623155e-01
1.60605121e+00 -1.69283426e+00 3.44283581e-01 -9.70615074e-02
-1.25794664e-01 -3.34275365e-01 -2.57559329e-01 9.80423570e-01
-1.83814496e-01 1.31706029e-01 -1.92212075e-01 -4.31173444e-01
3.45675856e-01 -1.24748908e-01 -6.22384369e-01 -1.96529299e-01
5.02084553e-01 1.21281493e+00 -1.23896825e+00 -6.09823942e-01
8.82059988e-03 3.59816343e-01 -6.29991114e-01 6.62092328e-01
-9.00788605e-01 3.72179061e-01 -3.73593062e-01 2.35389799e-01
-2.87772328e-01 -2.70477772e-01 1.15988724e-01 1.36196539e-01
4.53642681e-02 8.36959660e-01 -1.03638506e+00 2.03850985e+00
-1.68980762e-01 5.35239160e-01 -9.62259650e-01 -3.01633775e-01
6.70152843e-01 7.66438723e-01 2.74974793e-01 -4.91151750e-01
-2.28129327e-02 -2.25874558e-01 -1.56856552e-01 -3.54763210e-01
1.07850480e+00 -2.16315567e-01 -5.98314703e-01 1.02847254e+00
9.72162411e-02 -4.32600111e-01 5.60234249e-01 5.67243993e-01
1.34530318e+00 5.54160714e-01 5.03418386e-01 4.87104654e-01
-2.60241985e-01 3.22636485e-01 3.64735782e-01 8.69305849e-01
3.29663515e-01 1.00064766e+00 7.25309968e-01 -1.85588196e-01
-1.22451115e+00 -1.01074326e+00 4.89550292e-01 9.81301844e-01
-2.16593459e-01 -8.72936130e-01 -8.27885449e-01 -7.45311916e-01
-4.59826171e-01 1.77074206e+00 -5.41573286e-01 -3.66244435e-01
-7.89017856e-01 -6.87084138e-01 8.96206081e-01 7.82418013e-01
1.90727040e-01 -1.67785549e+00 -6.06669188e-01 8.51871550e-01
-6.01177692e-01 -1.15817034e+00 -5.53462982e-01 -1.86057180e-01
-4.03309613e-01 -7.08480239e-01 -3.05883229e-01 -4.56788361e-01
2.54745871e-01 -3.32110256e-01 1.46346748e+00 -3.03609580e-01
1.37423128e-01 2.59682126e-02 -5.70935190e-01 -5.12561381e-01
-7.73382962e-01 4.33252007e-01 -4.71673965e-01 -2.71842778e-01
-3.81110460e-02 -7.89044559e-01 -4.37394202e-01 -2.08284985e-02
-1.06124401e+00 7.42752612e-01 3.20555955e-01 6.38613880e-01
4.14110988e-01 2.28882983e-01 1.01945674e+00 -8.03849280e-01
1.05448627e+00 -8.81408513e-01 -2.55158931e-01 1.58188894e-01
-4.38140184e-01 2.94162389e-02 6.74294233e-01 -3.43573779e-01
-1.45649803e+00 -1.54601380e-01 3.47097665e-02 7.02230185e-02
-1.13329448e-01 9.86553967e-01 -5.95316812e-02 9.78258014e-01
8.21484029e-01 1.64705008e-01 -5.96503258e-01 2.47180849e-01
8.08252335e-01 1.52034104e-01 8.82889032e-01 -5.56529403e-01
6.37290120e-01 6.48379177e-02 -6.11660369e-02 -1.60711557e-01
-1.09424639e+00 2.41898045e-01 -1.71351418e-01 -4.70011145e-01
9.41131532e-01 -9.64895189e-01 -2.01856658e-01 2.08180442e-01
-1.57213879e+00 -6.26871228e-01 -5.45717657e-01 3.09812099e-01
-9.90112782e-01 -6.62159398e-02 -8.75899136e-01 -7.01681376e-01
-5.14315784e-01 -7.99086809e-01 9.91856158e-01 1.97940379e-01
-1.00680196e+00 -1.01555824e+00 1.39464319e-01 1.30474437e-02
4.91211340e-02 9.76454854e-01 1.02467406e+00 -7.03563035e-01
-7.02110469e-01 -1.95493281e-01 1.64347254e-02 -6.25032663e-01
-1.59390956e-01 -3.92243713e-02 -6.35325491e-01 1.74194604e-01
-4.48888361e-01 -3.89732629e-01 5.69877446e-01 3.16132277e-01
7.73329854e-01 -6.23775244e-01 -3.65956575e-01 3.34731519e-01
1.01787674e+00 2.20510632e-01 9.34660912e-01 2.68842757e-01
7.42185593e-01 6.33616805e-01 5.90536535e-01 8.42633426e-01
7.11636245e-01 5.74240148e-01 3.23170453e-01 2.14878485e-01
-4.35517222e-01 -1.08873069e+00 5.17937839e-01 4.62908447e-01
-7.66819268e-02 -1.04406548e+00 -8.67859483e-01 8.44021976e-01
-2.14999366e+00 -1.47838831e+00 -1.10780127e-01 1.68352652e+00
1.05520487e+00 3.93430710e-01 1.76828101e-01 1.40019760e-01
6.07862115e-01 6.04744613e-01 -3.17904472e-01 -2.54712999e-01
-1.21109553e-01 1.66341990e-01 -5.40553257e-02 3.30523580e-01
-8.54712844e-01 1.21772158e+00 6.74571753e+00 6.93087339e-01
-5.69010317e-01 -9.74746346e-02 4.73088533e-01 -4.43497241e-01
-5.73058903e-01 1.73225924e-01 -6.76113307e-01 3.67051989e-01
1.10728574e+00 -5.58155477e-01 4.22593206e-01 5.41250229e-01
4.73824471e-01 3.86248305e-02 -1.59449518e+00 4.41814601e-01
1.14669532e-01 -1.80167103e+00 4.51753795e-01 -9.12072062e-02
9.39369380e-01 -3.62594754e-01 -3.19815278e-01 5.39299726e-01
8.60598803e-01 -1.27688694e+00 1.21173370e+00 5.57313025e-01
6.74988747e-01 -9.73082483e-01 3.21376711e-01 3.51331264e-01
-1.16724133e+00 8.63738731e-02 2.52698362e-01 -3.70666116e-01
1.10681403e+00 4.37505573e-01 -1.44088733e+00 7.04533160e-01
-1.09763965e-01 7.88996518e-01 -4.12065387e-01 8.46431077e-01
-7.70995378e-01 6.72595501e-01 -2.07132697e-02 8.08028728e-02
3.99425834e-01 1.86522543e-01 7.90953636e-01 1.34674692e+00
5.43300152e-01 3.19152862e-01 3.17419380e-01 1.34042716e+00
-2.43495300e-01 -2.72116482e-01 -8.43118966e-01 -4.17512387e-01
5.27217805e-01 1.02797663e+00 -7.61622488e-01 -3.22962910e-01
-5.32826595e-02 9.67141628e-01 3.69260967e-01 3.15310836e-01
-1.07294643e+00 -3.99293691e-01 8.52597207e-02 1.17248148e-01
2.47196913e-01 -1.23219252e-01 -5.05528927e-01 -9.42957163e-01
-1.87149704e-01 -8.39637637e-01 6.23921633e-01 -1.19764209e+00
-1.07269132e+00 7.02996910e-01 2.76838958e-01 -9.17009592e-01
-1.22488916e+00 3.47022057e-01 -1.10214841e+00 5.76689482e-01
-9.85936224e-01 -1.37108839e+00 -3.45817432e-02 2.33258933e-01
8.33607912e-01 -1.11542679e-01 7.84100235e-01 -1.25944108e-01
-5.58458030e-01 4.58715111e-01 -6.87309980e-01 1.51323214e-01
4.77858245e-01 -1.22849703e+00 9.99621689e-01 1.22470784e+00
1.22417040e-01 1.14331611e-01 8.75329316e-01 -1.26457071e+00
-1.07139993e+00 -1.37189043e+00 1.21344972e+00 -6.22029245e-01
6.57470763e-01 -3.81127596e-01 -6.16237521e-01 1.01426160e+00
6.32504821e-01 -7.27767408e-01 5.74898183e-01 1.13438100e-01
-1.09543264e-01 6.65429413e-01 -6.90707207e-01 1.12854350e+00
1.46985292e+00 -3.05731237e-01 -7.67385483e-01 3.95802349e-01
1.06405330e+00 -7.60639250e-01 -9.23011422e-01 2.49248911e-02
3.26691985e-01 -7.65509725e-01 8.05889785e-01 -7.87033319e-01
1.38923657e+00 -2.88950205e-01 2.22660795e-01 -1.64806342e+00
-5.21238387e-01 -9.89737868e-01 -4.92473930e-01 1.50474370e+00
7.53948569e-01 9.26523060e-02 5.66251397e-01 6.53824091e-01
-5.03553689e-01 -5.36370933e-01 -4.60708320e-01 -4.52003062e-01
-3.59129816e-01 -6.69690847e-01 9.70659137e-01 8.28814805e-01
3.60783845e-01 1.01930845e+00 -4.83444571e-01 -1.73111781e-02
3.50513905e-01 8.11351463e-02 6.07695460e-01 -8.47186387e-01
-5.17088771e-01 -5.83886147e-01 1.47119865e-01 -5.63256979e-01
3.29159766e-01 -1.04824924e+00 3.77045602e-01 -2.13974476e+00
1.31441459e-01 -2.03327965e-02 2.30276838e-01 5.56545377e-01
-5.24265409e-01 -1.68472871e-01 3.71682912e-01 -1.08281173e-01
-7.11083651e-01 7.47797608e-01 1.33055139e+00 6.96005896e-02
-4.85660493e-01 -2.60139316e-01 -8.14631879e-01 5.25579095e-01
5.68946004e-01 -5.07966638e-01 -6.90961242e-01 -4.86127436e-01
6.24508917e-01 7.60319889e-01 3.44843417e-01 -9.67676938e-01
3.62777889e-01 -3.60496044e-01 3.72586936e-01 -7.85835683e-01
3.11583042e-01 -2.49602273e-02 6.20001554e-01 -2.86505036e-02
-8.49756837e-01 2.04004794e-01 -1.12156142e-02 6.09025896e-01
-2.37748832e-01 1.02873212e-02 2.44760230e-01 -2.61807799e-01
-5.78789175e-01 3.04506630e-01 -5.12310803e-01 3.14068079e-01
9.76257026e-01 -1.78798527e-01 -3.84832829e-01 -9.50003684e-01
-4.20707822e-01 2.09003359e-01 1.28046587e-01 4.32241142e-01
6.30829811e-01 -1.61022902e+00 -1.16462731e+00 -3.36698979e-01
1.86821952e-01 4.54315424e-01 -1.07833520e-02 3.62391829e-01
-2.83196270e-01 1.49869725e-01 -1.43940106e-01 -7.15736076e-02
-5.12071788e-01 4.35476303e-01 -2.08270131e-03 -8.99672866e-01
-9.37935293e-01 6.25013709e-01 1.14782453e-01 -3.45658101e-02
3.32432091e-02 -6.01354167e-02 -2.79486537e-01 1.27497464e-01
7.08828509e-01 3.71731162e-01 -2.19818190e-01 -2.17157260e-01
2.27988288e-01 -3.14917743e-01 -1.36480823e-01 -7.53541291e-01
1.40115213e+00 1.95867121e-01 3.48306477e-01 2.41565064e-01
4.94168818e-01 -3.19018036e-01 -1.48762441e+00 1.56089049e-02
1.75982520e-01 -4.26690429e-02 -7.75564089e-02 -1.01739955e+00
-5.97449243e-01 3.44872564e-01 -5.82388222e-01 3.30889076e-01
8.88106763e-01 -1.03547886e-01 1.19947886e+00 1.43656120e-01
1.45211905e-01 -1.03155971e+00 2.65645474e-01 7.60927200e-01
1.27802348e+00 -6.93698168e-01 -6.65965676e-02 -3.80023748e-01
-1.09686542e+00 9.06942248e-01 6.29360914e-01 -1.26445100e-01
-4.50558774e-02 2.72991717e-01 -4.62200075e-01 -1.60544425e-01
-1.17260635e+00 -4.09982428e-02 3.33324760e-01 4.79756504e-01
5.10325074e-01 5.27225062e-02 -1.41683146e-01 7.65963018e-01
-8.16091597e-01 1.53167322e-01 7.66881049e-01 9.07230616e-01
-6.29320741e-02 -9.48756635e-01 -2.04624087e-01 3.95150274e-01
-3.11550319e-01 -2.76822925e-01 -5.02855539e-01 6.37941420e-01
-2.56302893e-01 1.22022378e+00 3.58703397e-02 -2.88739115e-01
4.60398614e-01 2.41131678e-01 4.81617510e-01 -8.50831628e-01
-9.02671993e-01 -5.02341893e-03 8.32492709e-01 -6.08379722e-01
-5.21601774e-02 -8.54196489e-01 -1.45900667e+00 -3.96776348e-01
-3.14322039e-02 2.65955515e-02 1.37575224e-01 1.00107241e+00
4.43823040e-01 1.08287132e+00 3.49784076e-01 -8.82846653e-01
-1.41946390e-01 -1.01787829e+00 -2.01062813e-01 6.38496220e-01
-2.07621247e-01 -4.08680707e-01 2.33335346e-01 3.38311285e-01]
|
[11.660073280334473, 8.892730712890625]
|
ed85c86f-284a-4e0a-9cdd-d74f7ac5d97f
|
discrete-cosine-transform-network-for-guided
|
2104.06977
| null |
https://arxiv.org/abs/2104.06977v3
|
https://arxiv.org/pdf/2104.06977v3.pdf
|
Discrete Cosine Transform Network for Guided Depth Map Super-Resolution
|
Guided depth super-resolution (GDSR) is an essential topic in multi-modal image processing, which reconstructs high-resolution (HR) depth maps from low-resolution ones collected with suboptimal conditions with the help of HR RGB images of the same scene. To solve the challenges in interpreting the working mechanism, extracting cross-modal features and RGB texture over-transferred, we propose a novel Discrete Cosine Transform Network (DCTNet) to alleviate the problems from three aspects. First, the Discrete Cosine Transform (DCT) module reconstructs the multi-channel HR depth features by using DCT to solve the channel-wise optimization problem derived from the image domain. Second, we introduce a semi-coupled feature extraction module that uses shared convolutional kernels to extract common information and private kernels to extract modality-specific information. Third, we employ an edge attention mechanism to highlight the contours informative for guided upsampling. Extensive quantitative and qualitative evaluations demonstrate the effectiveness of our DCTNet, which outperforms previous state-of-the-art methods with a relatively small number of parameters. The code is available at \url{https://github.com/Zhaozixiang1228/GDSR-DCTNet}.
|
['Hanspeter Pfister', 'Zudi Lin', 'Shuang Xu', 'Jiangshe Zhang', 'Zixiang Zhao']
|
2021-04-14
| null |
http://openaccess.thecvf.com//content/CVPR2022/html/Zhao_Discrete_Cosine_Transform_Network_for_Guided_Depth_Map_Super-Resolution_CVPR_2022_paper.html
|
http://openaccess.thecvf.com//content/CVPR2022/papers/Zhao_Discrete_Cosine_Transform_Network_for_Guided_Depth_Map_Super-Resolution_CVPR_2022_paper.pdf
|
cvpr-2022-1
|
['depth-map-super-resolution']
|
['computer-vision']
|
[ 4.41161633e-01 6.37172014e-02 -8.79785568e-02 -3.12408954e-01
-1.30611205e+00 -1.22115612e-01 2.99126953e-01 -5.52854180e-01
-3.93629849e-01 6.24483168e-01 4.67664123e-01 1.33155182e-01
-7.31521696e-02 -9.95696127e-01 -4.64149982e-01 -8.97814333e-01
2.79654294e-01 -2.59709507e-01 3.89933735e-01 -2.93541521e-01
3.51306230e-01 5.24905086e-01 -1.66748977e+00 4.45413679e-01
7.99535871e-01 1.30394101e+00 3.63138229e-01 6.11399412e-01
1.55321583e-01 7.40672052e-01 4.75584678e-02 -1.39426649e-01
3.18342656e-01 -3.84669334e-01 -8.00907433e-01 -3.48759182e-02
3.93277705e-01 -7.27841556e-01 -7.35033751e-01 1.09081268e+00
6.66749120e-01 7.11227953e-02 3.53702009e-01 -8.76519024e-01
-6.87492967e-01 1.01305008e-01 -8.59226286e-01 4.27973807e-01
3.46235067e-01 1.19313978e-01 7.91333318e-01 -9.67002749e-01
7.40600348e-01 9.24111009e-01 2.84242809e-01 6.45850182e-01
-9.65044141e-01 -6.63851857e-01 -2.96637826e-02 1.50651053e-01
-1.45932376e+00 -5.60853302e-01 9.77722228e-01 -1.19852990e-01
6.85317636e-01 1.92764819e-01 5.61428666e-01 9.21539187e-01
1.08994998e-01 5.45576155e-01 1.32993054e+00 -2.52884477e-01
6.46196306e-02 -7.06902146e-02 -3.10790956e-01 7.93315113e-01
-8.46543312e-02 4.90358651e-01 -7.29455769e-01 1.88626274e-01
1.36076951e+00 6.33885264e-02 -7.50732303e-01 -7.52929449e-02
-1.18808985e+00 7.24182189e-01 6.16424561e-01 3.13616067e-01
-3.08227271e-01 1.47289902e-01 8.76527578e-02 1.85521822e-02
7.00974226e-01 -2.93993242e-02 -4.50371146e-01 8.84993151e-02
-6.93773746e-01 -1.04828961e-01 1.60236552e-01 8.56526613e-01
1.09438503e+00 -3.17936502e-02 2.57101213e-03 7.22811401e-01
2.26727754e-01 3.22712511e-01 3.07754070e-01 -1.19969785e+00
3.38391453e-01 3.99609864e-01 -1.57111749e-01 -7.25719571e-01
-3.34461808e-01 -1.47457674e-01 -9.75853503e-01 2.75327772e-01
2.79705793e-01 1.08503453e-01 -9.35457706e-01 1.38261819e+00
5.45328498e-01 1.86150536e-01 9.74460319e-02 1.33507395e+00
1.23801970e+00 3.70523870e-01 -1.57944232e-01 -9.58758965e-02
1.42068434e+00 -7.97502875e-01 -6.12797976e-01 -1.70915857e-01
-5.63012296e-03 -5.75053632e-01 1.01555204e+00 2.29276970e-01
-1.04703534e+00 -4.93737191e-01 -1.03559804e+00 -7.54915953e-01
-2.92812467e-01 1.16392992e-01 7.92620420e-01 4.04338628e-01
-1.09318030e+00 4.14001286e-01 -7.55374134e-01 -4.54295240e-02
6.13754749e-01 5.39624766e-02 -4.89288300e-01 -5.77519655e-01
-1.12537491e+00 4.76868987e-01 8.28008354e-02 2.90631711e-01
-7.78155744e-01 -8.65092039e-01 -1.06295931e+00 -2.93711931e-01
1.19732201e-01 -6.66096032e-01 8.07966411e-01 -7.33602822e-01
-1.52929568e+00 9.77592945e-01 -3.35231990e-01 2.53519714e-01
4.05723214e-01 -2.72557810e-02 -4.25718278e-01 7.61947513e-01
1.77814633e-01 5.44829786e-01 9.51615274e-01 -1.35276401e+00
-7.75047541e-01 -6.42909586e-01 3.14505510e-02 3.94966364e-01
-3.13062817e-02 -7.98671767e-02 -7.83826351e-01 -6.07990563e-01
5.25412381e-01 -3.28307986e-01 -2.30452478e-01 7.09866732e-02
-3.85349929e-01 2.55295336e-01 6.04721010e-01 -8.75724196e-01
1.03074706e+00 -2.21258950e+00 2.13477686e-01 -3.91393714e-02
4.90950108e-01 -3.46295148e-01 4.54109386e-02 -4.66306247e-02
-5.01979776e-02 -1.20629981e-01 -4.71312135e-01 -4.07331198e-01
-4.14242089e-01 -9.05655026e-02 -1.64810002e-01 7.38470376e-01
2.61666566e-01 1.03133965e+00 -7.43290246e-01 -5.42945623e-01
5.31785846e-01 9.51626360e-01 -2.81240821e-01 2.68591702e-01
8.36313590e-02 8.23616326e-01 -6.26190603e-01 1.02867460e+00
1.02841675e+00 -2.86399603e-01 -1.28130287e-01 -7.70906627e-01
-3.34517509e-01 -2.25837901e-02 -1.08897078e+00 2.12065244e+00
-4.20026183e-01 4.95814025e-01 1.25234753e-01 -6.89383626e-01
7.12952256e-01 8.20038393e-02 7.43672967e-01 -1.26530361e+00
2.42241189e-01 1.72477365e-01 -7.34058201e-01 -5.75715125e-01
6.49917424e-01 -2.31836289e-01 1.95073523e-02 4.67481971e-01
1.37644717e-02 -2.79580295e-01 -4.00685161e-01 1.04447544e-01
8.58922482e-01 1.76405951e-01 1.12370484e-01 1.24911264e-01
5.81435025e-01 -2.73185462e-01 5.67420483e-01 2.45160118e-01
-1.16745025e-01 1.26147830e+00 1.57295868e-01 -4.04178411e-01
-1.00346470e+00 -1.05712581e+00 -3.27499479e-01 6.96786582e-01
5.18294215e-01 -8.68366361e-02 -4.38374758e-01 -3.64552468e-01
-2.81777620e-01 1.05072975e-01 -8.05164158e-01 -8.86416994e-03
-5.20591378e-01 -7.56955624e-01 3.21278811e-01 5.25390148e-01
9.56138790e-01 -9.59810376e-01 -7.44604349e-01 -9.70405042e-02
-6.48037732e-01 -1.44415820e+00 -3.90640110e-01 8.11445192e-02
-7.62192905e-01 -1.15810657e+00 -8.70666265e-01 -6.38148785e-01
6.51218712e-01 4.65256393e-01 9.78332281e-01 2.13508122e-02
-6.88134551e-01 4.30905521e-01 -3.82733554e-01 2.69061685e-01
3.41810882e-01 -1.12741724e-01 -5.07567227e-01 9.25303623e-02
2.53760695e-01 -7.00466752e-01 -1.04988718e+00 1.75966814e-01
-1.20036018e+00 3.70069861e-01 6.96794987e-01 7.89043844e-01
9.55498517e-01 2.81364888e-01 1.85124263e-01 -6.85091853e-01
1.57794565e-01 -2.74906337e-01 -5.28638005e-01 8.47231746e-02
-3.67713392e-01 -7.55245313e-02 2.94360071e-01 -1.15764096e-01
-1.32557130e+00 1.80244133e-01 -1.85174480e-01 -4.78739023e-01
-2.37699047e-01 1.78643949e-02 -3.83843333e-01 -4.06011820e-01
3.24278384e-01 6.78429544e-01 -9.54793915e-02 -3.56720537e-01
5.42423368e-01 5.84130704e-01 6.72685325e-01 -4.93260860e-01
7.61398435e-01 1.20830286e+00 -3.09077203e-02 -7.86493838e-01
-1.00531089e+00 -5.12454271e-01 -6.87685251e-01 -2.83948153e-01
1.20385623e+00 -1.27818537e+00 -6.09972537e-01 6.11275494e-01
-7.70201385e-01 -4.63158250e-01 -1.79387122e-01 5.04810452e-01
-7.30683386e-01 3.96674812e-01 -8.76717567e-01 -6.00590885e-01
-3.93187165e-01 -1.01961601e+00 1.44480073e+00 6.04725957e-01
5.49187303e-01 -8.20071220e-01 -1.20894805e-01 5.97264588e-01
4.97695506e-01 4.37256545e-01 6.14945173e-01 3.76221776e-01
-1.12621510e+00 3.64945590e-01 -7.14524567e-01 3.51476699e-01
1.22830205e-01 -2.59262681e-01 -1.37318420e+00 -2.16460288e-01
3.57933044e-02 -3.15608978e-01 1.01937425e+00 6.41561568e-01
1.29281378e+00 7.00056776e-02 -2.64190342e-02 1.39677703e+00
1.99311590e+00 -2.41343856e-01 1.18759918e+00 4.99239057e-01
9.39795673e-01 5.20138502e-01 8.24624479e-01 4.99873191e-01
7.53748894e-01 5.14417768e-01 5.84095955e-01 -6.13744676e-01
-3.38279992e-01 -2.91722596e-01 1.13749832e-01 3.08015406e-01
-3.19591761e-01 3.33023936e-01 -4.03770536e-01 4.66509283e-01
-1.49203897e+00 -7.58576691e-01 -6.73621614e-03 2.11281991e+00
7.76573300e-01 -3.23777437e-01 -1.47405207e-01 1.12758584e-01
4.73396361e-01 3.34991634e-01 -6.48317039e-01 2.52141535e-01
-4.26938921e-01 5.14429927e-01 7.19594121e-01 6.75742865e-01
-1.03043079e+00 9.86386061e-01 5.22139692e+00 8.91123414e-01
-1.28572202e+00 1.88758150e-01 8.25699866e-01 3.89784724e-02
-5.06956339e-01 -5.96844181e-02 -6.59779310e-01 2.79041260e-01
3.33548814e-01 7.61302486e-02 5.46949446e-01 4.66388881e-01
1.43317461e-01 -4.76411879e-01 -7.17933595e-01 1.27085626e+00
2.32086495e-01 -1.31268585e+00 -2.59350240e-01 1.09645560e-01
7.79678881e-01 1.99639410e-01 2.98313707e-01 -2.14244142e-01
3.57925147e-02 -1.06584764e+00 5.38593054e-01 7.63180375e-01
1.35743737e+00 -8.33166063e-01 4.20083553e-01 -1.88611522e-01
-1.58610260e+00 -3.19390334e-02 -5.07601559e-01 4.02316064e-01
1.91549107e-01 7.40923703e-01 -6.51222393e-02 9.25690234e-01
1.16968846e+00 9.88688231e-01 -4.60368931e-01 6.62907958e-01
-3.09713602e-01 -1.64489657e-01 -3.76894213e-02 8.22124064e-01
-1.79226428e-01 -2.64471203e-01 2.24610895e-01 9.29872096e-01
2.81290710e-01 6.77784026e-01 -3.22084785e-01 1.14130151e+00
1.80466041e-01 -2.73201257e-01 -4.46742326e-01 2.51202285e-01
2.77841687e-01 1.60220301e+00 -6.82791531e-01 -2.59656161e-02
-5.12906551e-01 1.32740688e+00 2.50436366e-01 6.70050323e-01
-6.80605173e-01 -4.23148245e-01 7.55865991e-01 2.08331779e-01
4.39410299e-01 -2.40889683e-01 -3.33664805e-01 -1.36331105e+00
1.47998929e-01 -6.46418333e-01 4.66062963e-01 -1.09815490e+00
-1.19334626e+00 7.53239393e-01 -1.81450948e-01 -1.41546428e+00
1.03133261e-01 -4.20655608e-01 -3.19279373e-01 1.08009219e+00
-2.38422227e+00 -1.27698553e+00 -9.12757277e-01 1.20428193e+00
3.51820529e-01 2.83392370e-01 5.54185212e-01 4.54682887e-01
-4.18207705e-01 4.35980707e-01 -1.93709046e-01 1.51866212e-01
6.33972108e-01 -8.98413897e-01 -8.48686136e-03 8.48911464e-01
-5.23374319e-01 3.42615038e-01 1.84685290e-01 -4.23588932e-01
-1.52173340e+00 -1.02885270e+00 4.48395401e-01 -1.90990061e-01
2.27117181e-01 -2.14578196e-01 -8.03637624e-01 5.06408274e-01
-1.80284068e-01 5.77493250e-01 4.29680347e-01 -5.65610230e-01
-3.52891833e-01 -1.09451465e-01 -1.38663816e+00 2.32442841e-01
1.06738555e+00 -8.99659395e-01 -8.64060521e-02 -4.55144458e-02
8.82574856e-01 -6.90356255e-01 -1.26706958e+00 3.73968542e-01
6.07516170e-01 -1.43015885e+00 1.25292230e+00 1.26854494e-01
9.16535020e-01 -5.21022856e-01 -5.03070891e-01 -9.75940287e-01
-1.14105053e-01 -2.51471758e-01 -1.75499376e-02 9.95497644e-01
7.77089968e-02 -7.72974849e-01 7.89522290e-01 5.68216383e-01
-2.76513547e-02 -8.13784242e-01 -1.00035214e+00 -2.51539737e-01
-1.35512054e-01 -4.13436383e-01 5.44278145e-01 1.00709486e+00
-2.59016722e-01 -4.17362489e-02 -2.61436522e-01 4.39107478e-01
1.10490847e+00 2.85326660e-01 4.40977901e-01 -7.39031374e-01
-8.45201239e-02 -1.81983888e-01 -2.32826516e-01 -1.05975080e+00
-1.24440596e-01 -6.33415699e-01 -1.01343900e-01 -1.58585072e+00
2.12845355e-01 -5.82871318e-01 -2.08696678e-01 1.98369339e-01
-8.34646747e-02 6.38994098e-01 -1.99978620e-01 1.17929004e-01
-3.63808066e-01 8.42018127e-01 1.71854722e+00 1.55393973e-01
-2.94846855e-02 -3.16879332e-01 -9.22542810e-01 4.10954207e-01
5.89678109e-01 -1.57842711e-01 -2.73461670e-01 -4.45444316e-01
8.04558173e-02 3.25446993e-01 6.41979337e-01 -7.29129851e-01
3.18021268e-01 -1.81964472e-01 6.88958466e-01 -5.66589117e-01
6.18184328e-01 -7.99465358e-01 9.66427550e-02 -7.88793433e-03
-3.97853404e-02 -1.84281573e-01 3.10922991e-02 5.05320609e-01
-3.68840694e-01 4.28444415e-01 9.15532589e-01 -3.90770644e-01
-9.95936632e-01 7.26636171e-01 1.05298907e-01 2.68318076e-02
8.41697931e-01 -4.42347765e-01 -6.05405569e-01 -2.53229588e-01
-5.26502371e-01 2.42553800e-02 7.88649380e-01 3.26458156e-01
1.26781702e+00 -1.26190138e+00 -6.46808684e-01 5.12331903e-01
2.75662482e-01 4.33668077e-01 8.68733287e-01 9.71672535e-01
-5.47241688e-01 1.77900597e-01 -4.54305112e-01 -6.01876557e-01
-9.33393896e-01 2.40226671e-01 5.93733549e-01 1.68352351e-02
-1.13631248e+00 8.69068503e-01 4.41372186e-01 -3.04787308e-01
3.62457782e-02 -2.19555393e-01 -2.56153584e-01 -9.10832435e-02
6.44195378e-01 2.10988745e-01 5.95684052e-02 -6.96369588e-01
-3.23521167e-01 1.11556673e+00 4.28005122e-02 -3.17878664e-01
1.62634528e+00 -6.13672972e-01 -1.83715314e-01 1.37952298e-01
1.39279521e+00 -8.34972560e-02 -1.61411822e+00 -4.52772290e-01
-6.02783561e-01 -9.99989629e-01 4.92336899e-01 -5.71947753e-01
-1.66313624e+00 6.89101994e-01 7.58104026e-01 -3.43166560e-01
1.66679180e+00 1.67103454e-01 7.79867947e-01 -4.50523138e-01
4.60023522e-01 -9.52297986e-01 1.18012942e-01 1.73444986e-01
8.12993050e-01 -1.52785540e+00 1.74595729e-01 -6.50360882e-01
-7.20612586e-01 1.19381404e+00 6.94709837e-01 -4.38105464e-02
8.23374927e-01 2.44344935e-01 1.78025752e-01 -5.19231677e-01
-4.38173711e-01 -4.28412467e-01 2.08594844e-01 8.16585124e-01
2.75972515e-01 -1.25936791e-01 -3.66477016e-03 3.94923568e-01
-1.00451641e-01 9.52908769e-02 5.70151389e-01 7.12645710e-01
-2.29982525e-01 -7.31967568e-01 -1.50187135e-01 1.01757519e-01
-2.81519592e-01 -2.94041067e-01 6.93089217e-02 7.04571545e-01
8.54537413e-02 7.73511708e-01 2.81420182e-02 -4.86641735e-01
1.70487478e-01 -5.15526414e-01 6.73728526e-01 -3.01225722e-01
-6.94091022e-02 1.82336524e-01 -2.53886461e-01 -1.12438571e+00
-7.37376332e-01 -4.75460798e-01 -1.25224054e+00 -2.37632185e-01
1.03971191e-01 -4.33918566e-01 6.94195330e-01 6.00396216e-01
2.70678163e-01 6.82130516e-01 8.00915658e-01 -1.11378551e+00
2.05815673e-01 -6.65924609e-01 -9.35633361e-01 1.49888337e-01
7.11560547e-01 -5.45192838e-01 -6.50361300e-01 -9.30674523e-02]
|
[9.786229133605957, -2.4010820388793945]
|
16337c6d-a4ac-4ad1-b5c3-77b5b4090878
|
color-inference-from-semantic-labeling-for
|
1911.13114
| null |
https://arxiv.org/abs/1911.13114v2
|
https://arxiv.org/pdf/1911.13114v2.pdf
|
Color inference from semantic labeling for person search in videos
|
We propose an explainable model to generate semantic color labels for person search. In this context, persons are described from their semantic parts, such as hat, shirt, etc. Person search consists in looking for people based on these descriptions. In this work, we aim to improve the accuracy of color labels for people. Our goal is to handle the high variability of human perception. Existing solutions are based on hand-crafted features or learnt features that are not explainable. Moreover most of them only focus on a limited set of colors. We propose a method based on binary search trees and a large peer-labelled color name dataset. This allows us to synthesize the human perception of colors. Using semantic segmentation and our color labeling method, we label segments of pedestrians with their associated colors. We evaluate our solution on person search on datasets such as PCN, and show a precision as high as 80.4%.
|
['Guillaume-Alexandre Bilodeau', 'Harshad Mahadik', 'Jules Simon', 'David Steele']
|
2019-11-29
| null | null | null | null |
['person-search']
|
['computer-vision']
|
[-1.63285416e-02 -1.93681106e-01 4.09077927e-02 -6.78391397e-01
-3.81245404e-01 -8.34416509e-01 6.84113145e-01 8.53134543e-02
-4.61689770e-01 5.68560481e-01 -1.11027136e-01 9.27594230e-02
-4.58181463e-02 -8.04776609e-01 -5.41674256e-01 -2.84965605e-01
4.99399275e-01 1.08153665e+00 2.30594471e-01 6.21816963e-02
2.24908605e-01 2.81576663e-01 -1.82008910e+00 3.50330055e-01
9.10397768e-01 9.42755163e-01 8.86598155e-02 5.93550920e-01
-4.98760074e-01 2.57859319e-01 -8.35729063e-01 -6.91470206e-01
2.70963252e-01 -4.96873885e-01 -9.22687113e-01 2.38349020e-01
5.92378318e-01 -1.36146411e-01 1.09700911e-01 1.07093847e+00
2.18657255e-01 4.11083028e-02 9.26266909e-01 -1.58828187e+00
-7.56140053e-01 6.26930833e-01 -3.30322683e-01 -3.93204212e-01
4.75930870e-01 7.14758486e-02 1.04867613e+00 -4.44429725e-01
5.81541657e-01 1.41610754e+00 5.67114949e-01 1.03081965e+00
-1.37338042e+00 -5.42967141e-01 3.00600260e-01 5.08038759e-01
-1.41641831e+00 -2.86830049e-02 8.04851294e-01 -5.16932189e-01
3.11326295e-01 6.26198530e-01 8.59093845e-01 1.03258967e+00
-7.51147389e-01 8.03538084e-01 1.59876263e+00 -6.88509524e-01
3.89819443e-01 4.73901778e-01 3.24029028e-01 7.55630076e-01
2.98883021e-01 -1.04277074e-01 -3.42088282e-01 -1.54280663e-01
6.07260168e-01 1.06515415e-01 6.64040670e-02 -4.05773401e-01
-1.01130724e+00 7.39942193e-01 5.89908004e-01 1.71723247e-01
-1.10660166e-01 3.67774546e-01 -1.28106996e-02 -2.36783996e-01
3.00365597e-01 6.07101560e-01 -3.82881403e-01 6.62045851e-02
-9.59197104e-01 4.48475212e-01 8.47005427e-01 9.19439018e-01
1.07075357e+00 -6.00475311e-01 -4.11317468e-01 1.10864711e+00
3.14556301e-01 8.88632596e-01 1.02122977e-01 -1.14341152e+00
3.80218066e-02 7.60089278e-01 4.89560157e-01 -1.08055449e+00
-4.90467519e-01 -2.12637603e-01 -3.66015732e-01 2.80768573e-01
8.30110133e-01 -6.20745495e-02 -1.34446192e+00 1.69150555e+00
2.40290523e-01 -1.48586273e-01 -2.83775449e-01 1.38247466e+00
7.64507532e-01 2.80024737e-01 2.43815601e-01 5.69196284e-01
1.71577156e+00 -1.10002077e+00 -4.23419803e-01 -2.37216055e-01
1.06320038e-01 -6.54264987e-01 1.27888227e+00 4.96009976e-01
-7.43436396e-01 -6.70032918e-01 -5.58120370e-01 2.55526286e-02
-7.77296185e-01 4.39172149e-01 6.43318236e-01 1.08998597e+00
-9.31086004e-01 5.15733898e-01 -4.75702703e-01 -6.49069965e-01
1.33603841e-01 2.57675767e-01 -3.69274579e-02 -1.40616894e-01
-9.45490003e-01 5.91342509e-01 1.74212009e-01 6.22954965e-02
-6.94466591e-01 -1.88994139e-01 -4.31439012e-01 6.76699877e-02
5.16267419e-01 -9.26756501e-01 9.85213816e-01 -1.08058047e+00
-1.22060823e+00 1.02902031e+00 -2.11480245e-01 -3.58623087e-01
6.68181539e-01 -1.47511169e-01 -2.07767859e-01 2.90674418e-01
2.83478647e-01 1.13172722e+00 8.90268505e-01 -1.88170230e+00
-7.79695094e-01 -3.78586382e-01 1.93294525e-01 -6.17732480e-02
-3.01767111e-01 8.65998864e-02 -9.22951400e-01 -4.87503588e-01
-3.95863503e-02 -1.34633100e+00 -2.90143818e-01 -9.74522531e-03
-8.95914912e-01 -2.79722005e-01 3.44069898e-01 -8.27988505e-01
6.87328398e-01 -1.71601117e+00 -3.67468558e-02 7.19682097e-01
2.25271747e-01 6.89463615e-02 -7.41010234e-02 6.86571524e-02
4.20677096e-01 4.22848493e-01 -8.68540630e-02 -4.90444720e-01
5.03146052e-01 1.79634616e-01 6.85748011e-02 -1.03877895e-02
-2.76006833e-02 7.16284871e-01 -8.19167137e-01 -6.95812047e-01
7.90427476e-02 4.42973226e-01 -4.60398614e-01 1.45123929e-01
-5.15304625e-01 4.28867102e-01 -6.16702735e-01 7.14362919e-01
6.23220265e-01 -1.73900113e-01 1.32308409e-01 -1.41386837e-01
1.21978059e-01 -2.05329597e-01 -1.16817129e+00 1.57425594e+00
-2.90515095e-01 3.52449775e-01 -3.67360860e-01 -5.08693397e-01
1.07090056e+00 -1.82828784e-01 3.42005104e-01 -6.40201330e-01
8.71704593e-02 2.23642349e-01 -3.15893590e-01 -3.23773026e-01
5.21322608e-01 2.71410197e-01 -2.32236862e-01 5.11204720e-01
-3.76287997e-01 2.05334965e-02 4.50530708e-01 7.54446909e-02
8.06944311e-01 2.79532790e-01 -2.80663520e-01 -9.02364030e-02
5.34416795e-01 3.47008407e-01 2.97766656e-01 9.29602385e-01
-1.45493969e-01 9.25318301e-01 4.92982417e-01 -6.40565276e-01
-1.27056885e+00 -8.93293083e-01 1.48341283e-01 1.12644899e+00
3.93916428e-01 -4.64442551e-01 -1.29248714e+00 -8.40634525e-01
6.09698966e-02 6.01358831e-01 -6.38883770e-01 2.22136423e-01
-4.21411663e-01 -5.01671493e-01 3.93760264e-01 3.67542684e-01
6.15630805e-01 -1.13792849e+00 -7.09127247e-01 -1.09498940e-01
-3.87585253e-01 -1.09822404e+00 -3.84657770e-01 -1.81222111e-01
-2.42683232e-01 -1.24454618e+00 -1.09371352e+00 -5.96999049e-01
9.73501325e-01 9.46650803e-02 1.23278844e+00 3.96241546e-01
-5.55225670e-01 6.87941432e-01 -5.38205922e-01 -2.05250353e-01
-1.24516509e-01 1.10181339e-01 -2.27015957e-01 2.79318660e-01
5.79065800e-01 6.15830123e-02 -8.42222035e-01 5.96543729e-01
-4.92882758e-01 1.34681627e-01 4.85695362e-01 5.51179171e-01
4.39461797e-01 1.52969345e-01 -1.65772423e-01 -9.60031092e-01
5.18841386e-01 -4.04661447e-02 -4.77658570e-01 6.63670421e-01
-7.59727776e-01 4.09504503e-01 3.33116889e-01 -4.57049429e-01
-1.14053857e+00 4.27202791e-01 1.11371145e-01 -1.94673300e-01
-6.99507833e-01 -2.11007386e-01 -1.29055470e-01 -4.71525602e-02
6.65898681e-01 7.48232901e-02 -3.07041019e-01 -7.60844886e-01
7.42296994e-01 5.55742323e-01 4.78304148e-01 -1.00816309e+00
8.88295293e-01 5.04679620e-01 -1.29859298e-01 -3.47413808e-01
-5.50935924e-01 -4.65696365e-01 -5.43781400e-01 -4.12191272e-01
1.16837740e+00 -5.66102028e-01 -9.84658599e-01 1.87937737e-01
-1.17287660e+00 -3.38087618e-01 2.11157370e-02 2.90252209e-01
-4.61399972e-01 3.97617370e-01 -4.28355515e-01 -1.05676281e+00
-2.24629864e-01 -9.25572157e-01 1.28672922e+00 5.61718583e-01
-3.73013854e-01 -6.31490111e-01 -1.33925289e-01 6.21545374e-01
3.19359094e-01 2.87285984e-01 8.23372722e-01 -6.38143718e-01
-7.13571906e-01 -7.74558336e-02 -6.11083388e-01 7.71463066e-02
-9.89839807e-02 -7.55574740e-03 -9.19554055e-01 7.94399232e-02
-7.79925108e-01 -1.84261322e-01 1.04349816e+00 2.34005660e-01
1.28967011e+00 -2.25445569e-01 -5.77402771e-01 3.42414558e-01
1.49319196e+00 9.31205153e-02 3.49836379e-01 3.84354860e-01
8.03323328e-01 1.03851736e+00 5.19890368e-01 3.81353289e-01
4.77819353e-01 8.81848752e-01 2.77733177e-01 -1.90632269e-01
-3.62287611e-01 -4.36014056e-01 -2.37739563e-01 -1.01179488e-01
-4.94878918e-01 -2.42491648e-01 -1.00173676e+00 4.75038826e-01
-1.96754110e+00 -7.19261110e-01 -3.43247265e-01 2.01561260e+00
7.37941027e-01 2.81127766e-02 5.66880167e-01 5.42966742e-03
9.41821635e-01 -4.51617301e-01 -2.26926371e-01 -1.35628760e-01
4.04439978e-02 4.67363074e-02 5.53039312e-01 5.28736830e-01
-9.92745578e-01 1.09656870e+00 5.90255547e+00 6.71638072e-01
-6.99736953e-01 -5.85764125e-02 6.79353416e-01 2.33927831e-01
-4.65118617e-01 1.19941451e-01 -9.65572774e-01 6.19883358e-01
2.65791416e-01 2.89389223e-01 7.35519588e-01 8.31480384e-01
-8.50413963e-02 -1.53237179e-01 -9.76653755e-01 1.26619291e+00
2.45710447e-01 -7.29022563e-01 4.25139852e-02 -3.79379876e-02
6.91296995e-01 -7.77319074e-01 1.43350706e-01 -9.37954523e-04
4.57715005e-01 -1.07520258e+00 1.03422761e+00 8.35721135e-01
6.15322530e-01 -5.47652125e-01 4.79910225e-01 4.66392785e-02
-9.75237191e-01 -1.28229484e-01 -3.30013186e-01 2.28668272e-01
4.44223322e-02 3.59656811e-01 -1.04583061e+00 1.96998969e-01
8.45367253e-01 2.45784864e-01 -1.15852904e+00 1.32113707e+00
-3.93436909e-01 3.54854047e-01 -4.14839476e-01 -4.63131607e-01
-3.17225009e-02 -1.64020717e-01 4.95706275e-02 1.28165758e+00
3.70967299e-01 -1.87682450e-01 3.02872270e-01 1.38878047e+00
3.07704002e-01 1.41608536e-01 -2.84310095e-02 4.65921052e-02
3.37304711e-01 1.44277251e+00 -1.16440022e+00 -4.41072017e-01
-1.94276512e-01 1.25669098e+00 8.90540034e-02 5.74455976e-01
-9.73946452e-01 -2.55494863e-01 4.14133281e-01 1.86599389e-01
1.48020193e-01 -2.97122132e-02 -3.67174000e-01 -9.20882344e-01
-1.21047522e-03 -7.34373808e-01 3.67935032e-01 -9.81663942e-01
-1.36577165e+00 7.59063900e-01 1.54781774e-01 -9.44702089e-01
-2.39897504e-01 -6.34362340e-01 -1.42332956e-01 8.24772954e-01
-1.28427494e+00 -1.56898367e+00 -7.60841787e-01 5.89416742e-01
3.84289354e-01 -6.47934005e-02 7.83383012e-01 2.68482685e-01
-2.36004323e-01 4.31739897e-01 -3.65990281e-01 2.90335923e-01
7.67266452e-01 -1.55664778e+00 4.23352599e-01 4.73533094e-01
4.35116708e-01 6.33209288e-01 9.40785050e-01 -6.81561351e-01
-6.91640794e-01 -7.07146287e-01 9.81723309e-01 -6.28794670e-01
8.40423182e-02 -5.34006894e-01 -5.35729170e-01 2.52258271e-01
-1.07940065e-03 -1.57687470e-01 5.08703232e-01 4.06168997e-01
-5.55410445e-01 -2.09298924e-01 -1.06922364e+00 5.57202041e-01
1.15522873e+00 -3.00580382e-01 -3.09220284e-01 3.16398919e-01
2.47855455e-01 -1.02229968e-01 -2.88681149e-01 3.11943721e-02
6.93057954e-01 -1.00902605e+00 1.18278134e+00 -5.71375966e-01
1.04869150e-01 -5.42350888e-01 9.19907317e-02 -1.24671733e+00
-2.17125267e-01 -1.29473776e-01 5.65630257e-01 1.32626247e+00
5.50550282e-01 -3.74776542e-01 1.15962040e+00 1.24266624e+00
2.61198163e-01 2.30616760e-02 -4.40961063e-01 -7.89387941e-01
-4.53670353e-01 -2.69728810e-01 9.25249159e-01 6.28438890e-01
-3.46189409e-01 1.20171323e-01 -5.17676532e-01 1.02759808e-01
1.02338660e+00 4.00464922e-01 9.63056147e-01 -1.49691725e+00
-3.99840444e-01 -4.71021205e-01 -2.72675097e-01 -7.31336653e-01
7.78643042e-02 -6.37157440e-01 1.74315110e-01 -1.68039525e+00
7.08437502e-01 -9.41449285e-01 -2.45370120e-01 7.14316130e-01
-2.43405059e-01 4.87411082e-01 5.01882493e-01 1.59681678e-01
-9.10916984e-01 1.97547376e-02 1.04705131e+00 -4.66072410e-01
-8.50819796e-02 5.38548492e-02 -5.20555854e-01 7.03501701e-01
1.02553105e+00 -4.67507750e-01 -7.42382631e-02 -4.40399170e-01
3.12041909e-01 -3.88371259e-01 8.92158031e-01 -1.10270560e+00
1.31311566e-01 -1.76414266e-01 7.02889442e-01 -4.19701517e-01
5.65498173e-01 -1.07655478e+00 3.21595252e-01 4.86264735e-01
-3.58600289e-01 -2.90124446e-01 -3.76136780e-01 5.32913446e-01
2.65843403e-02 -3.43734980e-01 4.13765430e-01 -5.19346893e-01
-9.84805405e-01 2.68224068e-02 7.83513114e-02 -1.13200068e-01
7.99670756e-01 -9.78115946e-02 -4.54635471e-01 -4.96589512e-01
-9.62083697e-01 1.55416772e-01 7.17725813e-01 3.77795279e-01
4.24291849e-01 -1.43157363e+00 -4.69703585e-01 -2.30971677e-03
2.52003074e-01 -4.60498393e-01 6.78682625e-02 1.35331452e-01
-6.20865762e-01 5.24777949e-01 -3.87374401e-01 -5.59500933e-01
-1.50600135e+00 7.34799385e-01 3.23150814e-01 9.77758020e-02
-2.51821667e-01 7.36181855e-01 -1.13891112e-02 -4.19806123e-01
2.37418726e-01 -2.62357950e-01 -4.28640217e-01 -1.34703726e-01
1.41835451e-01 4.69446778e-01 -5.62971950e-01 -6.98107779e-01
-4.26650882e-01 1.11024797e+00 4.34884459e-01 -2.68400162e-01
9.01469707e-01 -6.32874072e-02 -6.93315342e-02 2.46414784e-02
9.01610672e-01 1.49923652e-01 -1.23077714e+00 -1.52765825e-01
1.68457940e-01 -7.16182053e-01 -5.82697034e-01 -1.11006641e+00
-9.97630000e-01 6.71403289e-01 8.05222869e-01 4.64915246e-01
1.02401543e+00 2.85674304e-01 7.87013412e-01 2.70261437e-01
6.28903270e-01 -1.34491789e+00 1.87271863e-01 -8.46280754e-02
4.58034068e-01 -1.33902192e+00 -1.22748271e-01 -7.40045905e-01
-6.78839505e-01 1.06194544e+00 5.69602013e-01 1.42686471e-01
1.88698113e-01 -2.59932697e-01 1.32453740e-01 -3.45558748e-02
-1.96401358e-01 -8.57699275e-01 6.25680625e-01 8.52434576e-01
4.23510782e-02 4.22394365e-01 -4.15608823e-01 7.77952611e-01
-4.06323850e-01 -1.30080059e-01 8.43690857e-02 4.11542296e-01
-5.92084944e-01 -1.44141150e+00 -7.44738340e-01 1.67855009e-01
-2.21927196e-01 3.35173495e-02 -9.17208076e-01 5.38336337e-01
5.87694764e-01 1.21717596e+00 7.24078109e-03 -3.93724889e-01
2.89014280e-01 2.00089723e-01 5.20164907e-01 -2.89220572e-01
-2.67625958e-01 -8.06456506e-02 3.00255895e-01 -4.60675210e-01
-5.55321991e-01 -5.28120458e-01 -1.17182100e+00 -2.09539860e-01
1.33342087e-01 2.46393442e-01 9.27997828e-01 6.66786432e-01
-3.48404609e-02 1.23966552e-01 3.84871215e-01 -4.76149976e-01
-1.39968708e-01 -6.06007338e-01 -6.70988262e-01 1.05830956e+00
-3.35888490e-02 -6.80767655e-01 -1.73414528e-01 3.78975898e-01]
|
[9.059786796569824, 0.13283585011959076]
|
9b059526-46aa-4535-bc53-3d2bf13127bc
|
learnable-graph-matching-incorporating-graph
|
2103.16178
| null |
https://arxiv.org/abs/2103.16178v1
|
https://arxiv.org/pdf/2103.16178v1.pdf
|
Learnable Graph Matching: Incorporating Graph Partitioning with Deep Feature Learning for Multiple Object Tracking
|
Data association across frames is at the core of Multiple Object Tracking (MOT) task. This problem is usually solved by a traditional graph-based optimization or directly learned via deep learning. Despite their popularity, we find some points worth studying in current paradigm: 1) Existing methods mostly ignore the context information among tracklets and intra-frame detections, which makes the tracker hard to survive in challenging cases like severe occlusion. 2) The end-to-end association methods solely rely on the data fitting power of deep neural networks, while they hardly utilize the advantage of optimization-based assignment methods. 3) The graph-based optimization methods mostly utilize a separate neural network to extract features, which brings the inconsistency between training and inference. Therefore, in this paper we propose a novel learnable graph matching method to address these issues. Briefly speaking, we model the relationships between tracklets and the intra-frame detections as a general undirected graph. Then the association problem turns into a general graph matching between tracklet graph and detection graph. Furthermore, to make the optimization end-to-end differentiable, we relax the original graph matching into continuous quadratic programming and then incorporate the training of it into a deep graph network with the help of the implicit function theorem. Lastly, our method GMTracker, achieves state-of-the-art performance on several standard MOT datasets. Our code will be available at https://github.com/jiaweihe1996/GMTracker .
|
['Zhaoxiang Zhang', 'Naiyan Wang', 'Zehao Huang', 'JiaWei He']
|
2021-03-30
| null |
http://openaccess.thecvf.com//content/CVPR2021/html/He_Learnable_Graph_Matching_Incorporating_Graph_Partitioning_With_Deep_Feature_Learning_CVPR_2021_paper.html
|
http://openaccess.thecvf.com//content/CVPR2021/papers/He_Learnable_Graph_Matching_Incorporating_Graph_Partitioning_With_Deep_Feature_Learning_CVPR_2021_paper.pdf
|
cvpr-2021-1
|
['online-multi-object-tracking', 'graph-partitioning']
|
['computer-vision', 'graphs']
|
[-2.83076048e-01 -1.59772560e-01 -3.62760007e-01 -2.83671767e-01
-5.07658362e-01 -3.61923784e-01 1.62476137e-01 -6.23373874e-02
-2.31781155e-01 5.33691347e-01 -1.87223747e-01 -1.33256689e-01
-1.46780610e-01 -6.04688883e-01 -1.02978146e+00 -7.14738369e-01
2.67782602e-02 4.09983963e-01 3.80149543e-01 6.60748333e-02
-2.60684341e-01 4.02310818e-01 -1.18307984e+00 -3.04338038e-01
7.40705788e-01 1.15294695e+00 1.54288724e-01 4.08783108e-01
-1.18862487e-01 7.19612837e-01 -2.61614084e-01 -3.79036069e-01
4.57223058e-01 -1.70201063e-01 -2.34055772e-01 1.53736308e-01
9.31665421e-01 -3.20983022e-01 -9.72903848e-01 1.28170991e+00
4.79904085e-01 1.23039812e-01 1.76707774e-01 -1.62907422e+00
-7.15198815e-01 2.95974314e-01 -7.83839345e-01 2.24075422e-01
1.47084901e-02 1.92225739e-01 8.98808837e-01 -7.01259196e-01
4.08154130e-01 1.08735752e+00 9.72762167e-01 3.36186856e-01
-9.77417529e-01 -8.84236455e-01 4.98795986e-01 2.79932529e-01
-1.34579420e+00 -3.85382861e-01 7.87947297e-01 -5.91292739e-01
4.83407289e-01 9.62744653e-02 8.93723428e-01 6.43785775e-01
1.92066148e-01 8.14784884e-01 5.09274960e-01 1.13441199e-01
-2.96719402e-01 -2.39611536e-01 2.71845251e-01 1.10845673e+00
6.34169221e-01 3.03846776e-01 -2.16318816e-01 -6.20055757e-02
8.85014474e-01 2.80596584e-01 -2.91557968e-01 -5.88867307e-01
-1.19161296e+00 7.28715301e-01 7.94659734e-01 5.18565774e-02
-2.35354275e-01 4.61679041e-01 2.61317194e-01 2.03114584e-01
3.24946672e-01 -1.19855359e-01 -2.09436521e-01 3.59346181e-01
-8.00799251e-01 3.24357808e-01 6.06435180e-01 1.18487096e+00
1.03951979e+00 1.83876693e-01 -2.42107421e-01 3.49788219e-01
8.19914877e-01 6.57977939e-01 -1.85616817e-02 -7.33871639e-01
4.65222597e-01 6.08482718e-01 7.89111629e-02 -1.38767409e+00
-4.70068961e-01 -7.68793583e-01 -7.70764470e-01 7.63202179e-03
5.69189370e-01 -3.13383996e-01 -7.16322720e-01 1.73468494e+00
7.25175977e-01 5.51177919e-01 -4.37942773e-01 1.18473971e+00
1.02133667e+00 3.62105519e-01 -4.60573733e-02 -2.63531834e-01
1.29666424e+00 -1.28819752e+00 -8.94135118e-01 -2.05153212e-01
5.80707133e-01 -6.46725655e-01 5.15103579e-01 -4.10616174e-02
-8.41980994e-01 -6.39646292e-01 -1.03734946e+00 -7.30260536e-02
-1.88224614e-01 2.69188732e-01 7.52476931e-01 2.96873271e-01
-9.22858655e-01 3.42084587e-01 -9.74286556e-01 -2.73156077e-01
4.62981045e-01 6.45850062e-01 -2.00652272e-01 -3.76911052e-02
-9.32220280e-01 5.76040924e-01 3.94334018e-01 4.69229281e-01
-7.55719721e-01 -4.54064995e-01 -9.23153639e-01 -1.13460928e-01
7.31140494e-01 -1.11743510e+00 1.01634967e+00 -7.57302821e-01
-1.30378485e+00 7.17187822e-01 -1.75515577e-01 -3.08896720e-01
6.09845519e-01 -3.73943865e-01 -3.67052764e-01 -2.78635234e-01
1.94135278e-01 3.90519261e-01 7.10939527e-01 -9.91269529e-01
-7.00740457e-01 -2.28064001e-01 2.36920297e-01 -6.71543479e-02
-2.17263341e-01 -3.10560521e-02 -9.42252934e-01 -7.44202077e-01
1.05371252e-01 -1.07955337e+00 -1.46585792e-01 4.51214463e-01
-4.68886703e-01 -3.96561742e-01 1.11069298e+00 -4.98685092e-01
1.17292547e+00 -2.09350753e+00 1.06677808e-01 -8.86581913e-02
6.24391079e-01 2.91763097e-01 -1.54196769e-01 1.28493056e-01
1.58717647e-01 -4.36781049e-01 6.65318295e-02 -4.53046709e-01
1.49356127e-01 1.63168371e-01 -1.62357669e-02 9.81786549e-01
4.28252444e-02 1.15168118e+00 -1.09771764e+00 -7.09793687e-01
2.91940480e-01 4.70931798e-01 -4.85013008e-01 2.14741796e-01
-3.82501781e-01 5.73987126e-01 -8.18426609e-01 6.42582357e-01
8.04659724e-01 -5.50399959e-01 -9.46459640e-03 -6.32894039e-01
-3.08099985e-01 2.21940186e-02 -1.30570912e+00 1.81771958e+00
1.37337521e-01 6.30846977e-01 2.18999878e-01 -1.18930459e+00
8.68577898e-01 1.08418830e-01 9.27652955e-01 -3.97841424e-01
2.79334515e-01 1.47528779e-02 3.24459211e-03 -3.95814896e-01
3.29420269e-01 1.46272331e-01 2.10678056e-01 1.13431357e-01
-1.83601901e-02 5.26771247e-01 2.17797548e-01 1.04338691e-01
9.93838131e-01 3.38914394e-01 1.79036167e-02 -1.35976851e-01
4.47212845e-01 3.12391631e-02 1.04726315e+00 6.99149311e-01
-3.47602993e-01 3.59238565e-01 1.18911281e-01 -6.10236883e-01
-6.28268778e-01 -9.80885148e-01 -9.10902978e-04 8.92452955e-01
5.78126431e-01 -4.06915784e-01 -3.03630620e-01 -7.81906664e-01
1.70939893e-01 -5.58998957e-02 -4.03652281e-01 -7.32171386e-02
-8.35520983e-01 -6.66037858e-01 3.69868845e-01 4.53779072e-01
4.87089127e-01 -7.73059011e-01 -1.71070501e-01 3.83072913e-01
-2.07543552e-01 -1.34422708e+00 -8.53929698e-01 -2.26876989e-01
-8.21434975e-01 -1.24937391e+00 -5.10739744e-01 -7.66440988e-01
6.86689734e-01 5.73566020e-01 9.41955388e-01 4.59096104e-01
-1.13654450e-01 3.73062402e-01 -1.69626176e-01 -3.50871414e-01
2.38598228e-01 1.15731388e-01 9.41003263e-02 2.11761862e-01
3.84611338e-01 -4.40199584e-01 -6.99278712e-01 4.86859411e-01
-4.44105834e-01 9.25798416e-02 5.20852029e-01 6.77639961e-01
8.58038247e-01 -1.67187646e-01 2.74097472e-01 -4.60443795e-01
9.67191756e-02 -4.64124173e-01 -1.09050417e+00 2.48508021e-01
-2.48687238e-01 -9.62082818e-02 3.27407062e-01 -5.97236454e-01
-4.69526201e-01 4.42793518e-01 8.03842768e-02 -9.83367085e-01
2.50178456e-01 5.26946604e-01 -3.44940662e-01 -6.32283986e-01
1.14117950e-01 1.46078974e-01 3.17781083e-02 -3.67259622e-01
2.78959334e-01 1.76672995e-01 5.89925766e-01 -5.19889414e-01
1.16375518e+00 5.93799233e-01 5.73890619e-02 -5.54667175e-01
-1.18072844e+00 -6.52922750e-01 -4.36326206e-01 -5.73112071e-01
9.35185790e-01 -1.03400588e+00 -1.09073651e+00 3.29774022e-01
-1.19006693e+00 -3.99070472e-01 -5.56261465e-02 8.08025301e-01
-2.74611652e-01 6.75345302e-01 -5.56289673e-01 -6.42266095e-01
-2.43271440e-01 -1.08343101e+00 1.12194860e+00 5.49985945e-01
3.65655005e-01 -1.02139401e+00 1.56498358e-01 3.68478298e-01
1.92839295e-01 4.12563503e-01 1.80039406e-01 -3.45341653e-01
-1.18225551e+00 -2.19555154e-01 -5.55561662e-01 -7.25952014e-02
1.76214322e-01 -2.36705039e-03 -5.91611326e-01 -6.25071585e-01
-7.35865459e-02 -5.58213927e-02 8.12420964e-01 6.98379338e-01
1.14218378e+00 -2.48456642e-01 -7.09803343e-01 1.08882391e+00
1.36738253e+00 -6.85096607e-02 3.35945755e-01 4.25098896e-01
1.28029513e+00 2.47944266e-01 7.71980464e-01 2.45447144e-01
7.81835616e-01 9.94418740e-01 6.77042365e-01 -1.84245467e-01
-2.34236747e-01 -2.14634806e-01 4.06330138e-01 9.72367585e-01
-3.54465768e-02 -2.77171940e-01 -6.59042358e-01 3.40687215e-01
-2.49952340e+00 -8.60690415e-01 -4.40688461e-01 2.17634439e+00
3.14091772e-01 1.41988739e-01 3.31548095e-01 -4.78203684e-01
1.07042003e+00 1.84878483e-01 -7.11055994e-01 4.12936717e-01
-5.15377223e-02 -4.17682618e-01 7.52564609e-01 3.95666242e-01
-1.43002427e+00 8.91290426e-01 5.10825253e+00 7.38585174e-01
-1.14088631e+00 1.98328137e-01 -1.76369399e-02 -5.69262207e-02
3.34946811e-02 1.55515730e-01 -1.12992454e+00 5.70524454e-01
3.56855452e-01 -1.07680067e-01 4.19372141e-01 6.75836861e-01
1.59950793e-01 3.98513883e-01 -1.02527547e+00 1.15668571e+00
-1.71281621e-01 -1.28958309e+00 -1.85411647e-01 2.08684713e-01
4.42462385e-01 3.53280455e-01 -1.41617253e-01 4.20598328e-01
2.28670880e-01 -6.47211850e-01 7.25731015e-01 5.79377294e-01
4.29271460e-01 -2.35331595e-01 5.65248132e-01 2.50630885e-01
-1.86606157e+00 1.56646878e-01 -6.20274425e-01 -6.64516017e-02
4.48678255e-01 6.56626284e-01 -3.53885174e-01 1.01338100e+00
6.43410265e-01 1.08826959e+00 -3.98554921e-01 1.57328606e+00
-7.35487044e-02 4.31025356e-01 -5.54561913e-01 1.27992928e-01
1.74945012e-01 -3.96778613e-01 8.87686074e-01 1.10969138e+00
2.83353448e-01 -1.06881477e-01 7.23598719e-01 8.13249648e-01
-1.25101686e-01 -1.62330359e-01 -5.22085130e-01 -3.27555649e-02
4.25708443e-01 1.44908786e+00 -6.67152643e-01 -2.05338702e-01
-7.41534233e-01 4.98333424e-01 5.32139421e-01 4.28078622e-01
-1.25112569e+00 -1.40575305e-01 6.60241544e-01 9.39801261e-02
5.41715741e-01 -4.32325631e-01 1.97173990e-02 -1.34486282e+00
3.43220294e-01 -6.45079136e-01 5.34618378e-01 -5.29330254e-01
-1.42057240e+00 3.99234742e-01 -1.43439248e-01 -1.48012984e+00
2.24375173e-01 -4.71211523e-01 -7.18468904e-01 5.47018945e-01
-1.49440229e+00 -1.29577124e+00 -6.36099219e-01 6.81895852e-01
1.36069208e-01 2.03996040e-02 2.16058820e-01 9.81204152e-01
-9.13182199e-01 7.55573869e-01 -6.71285614e-02 5.37777960e-01
6.59118652e-01 -1.00972569e+00 3.18390399e-01 9.30998147e-01
1.05453350e-01 4.86575603e-01 5.96242964e-01 -8.25470865e-01
-1.69115436e+00 -1.47627306e+00 6.05192125e-01 -3.73622090e-01
9.05075192e-01 -3.37210953e-01 -9.73718762e-01 9.57724571e-01
-2.96047386e-02 5.82219362e-01 3.35781902e-01 6.82893321e-02
-1.61508650e-01 -2.76059210e-01 -5.94018281e-01 4.46817130e-01
1.36394417e+00 -2.85703838e-01 -2.98623830e-01 5.85094869e-01
8.04034472e-01 -7.80443728e-01 -8.82286549e-01 5.41014314e-01
5.12549818e-01 -5.71125627e-01 1.03717303e+00 -5.12019098e-01
-2.61693150e-01 -9.48337734e-01 -3.44989297e-04 -8.62834156e-01
-6.07648015e-01 -6.50676847e-01 -5.73038936e-01 1.19913745e+00
1.03250802e-01 -8.18642735e-01 8.88965666e-01 2.76125938e-01
-3.51962388e-01 -7.55797505e-01 -8.20455611e-01 -9.30411518e-01
-3.94289106e-01 -1.08021244e-01 5.36015809e-01 1.15605831e+00
-5.34232974e-01 3.17634165e-01 -6.04641318e-01 5.45409024e-01
1.15401912e+00 2.90429235e-01 1.14655781e+00 -1.18566442e+00
-5.45914352e-01 -3.51430178e-01 -5.43970585e-01 -1.48777819e+00
8.62321928e-02 -8.89171839e-01 1.63790196e-01 -1.56723344e+00
1.66532248e-01 -7.91576743e-01 -3.74739498e-01 5.08634090e-01
-3.96197587e-01 1.38516247e-01 4.23421472e-01 4.08791751e-01
-1.04439282e+00 6.96051002e-01 1.46011853e+00 -3.04677606e-01
-2.30447645e-03 1.50498152e-01 -4.92305160e-01 5.21558344e-01
6.91558182e-01 -8.28046560e-01 -1.42194331e-01 -6.54445291e-01
1.88752905e-01 1.83628887e-01 6.76905811e-01 -9.58449423e-01
6.51187062e-01 -1.58081219e-01 1.61259919e-01 -7.52641022e-01
2.63081104e-01 -9.06166553e-01 4.50431824e-01 5.30071080e-01
2.19977051e-01 1.81790769e-01 2.79269814e-01 8.38044524e-01
-1.56209290e-01 8.45529363e-02 4.83981758e-01 1.58424303e-01
-6.73044384e-01 1.07687628e+00 1.34211421e-01 1.11300036e-01
9.42971051e-01 -2.89806843e-01 -3.90947044e-01 -1.69085175e-01
-6.24443114e-01 6.83897674e-01 4.72546875e-01 4.67617124e-01
2.20066816e-01 -1.66229570e+00 -7.00484753e-01 -1.71489120e-01
-1.08260952e-01 3.00211936e-01 2.36060411e-01 1.46660781e+00
-2.88703322e-01 1.63679883e-01 -1.43977103e-03 -1.01299739e+00
-1.13174653e+00 6.96397483e-01 5.52773595e-01 -2.00133279e-01
-8.01273942e-01 6.41776741e-01 5.55825531e-01 -2.86512703e-01
4.34981644e-01 -2.21969739e-01 1.91089232e-02 -2.51731008e-01
2.09007949e-01 1.82628229e-01 -2.35719919e-01 -6.79784060e-01
-4.84831691e-01 8.61215889e-01 2.92835850e-02 5.30564666e-01
1.02444065e+00 -1.19828627e-01 -1.10388428e-01 1.70151010e-01
1.11014497e+00 -1.51004300e-01 -1.48168826e+00 -3.76247436e-01
-2.22172678e-01 -4.09591734e-01 6.23521470e-02 -1.28836736e-01
-1.55826294e+00 6.10870063e-01 5.47825813e-01 1.81496054e-01
8.35340738e-01 -2.94352211e-02 9.76338446e-01 4.40580487e-01
2.68302619e-01 -7.44224012e-01 -5.07790595e-02 4.65469986e-01
4.79118347e-01 -1.38989627e+00 1.80752859e-01 -5.16512692e-01
-6.41383007e-02 1.04704201e+00 8.41704309e-01 -3.27913463e-01
7.56598353e-01 1.34675786e-01 1.44040361e-01 -4.77782428e-01
-4.04644608e-01 -3.93268168e-01 5.01289904e-01 4.23186213e-01
3.28021079e-01 -6.43668026e-02 -2.17678204e-01 3.17313224e-01
1.30757526e-01 1.06372386e-01 -5.98184094e-02 7.71865666e-01
-3.48953456e-01 -9.64385390e-01 -3.01506609e-01 4.01516944e-01
-3.35052222e-01 2.23843440e-01 -1.19633451e-01 8.69512796e-01
2.67784864e-01 8.09876323e-01 -1.01834700e-01 -3.75307441e-01
3.50781322e-01 -6.20535374e-01 5.30629516e-01 -3.72966796e-01
-4.34746414e-01 2.11192146e-01 -1.02621965e-01 -6.34255886e-01
-6.68079793e-01 -6.83429718e-01 -1.17698133e+00 -4.65007812e-01
-7.91901827e-01 1.17353223e-01 3.39692533e-01 1.09620619e+00
4.18854624e-01 7.58056641e-01 2.07585260e-01 -8.58191252e-01
-2.71997422e-01 -6.65506423e-01 -2.93537080e-01 2.21123815e-01
5.03275931e-01 -1.07466888e+00 -1.04766935e-01 -1.26999676e-01]
|
[6.3659443855285645, -2.08918833732605]
|
2308c416-c26a-4aa6-99b0-9d06b0af0fc0
|
egcn-an-ensemble-based-learning-framework-for
| null | null |
https://www.ijcai.org/proceedings/2022/511
|
https://www.ijcai.org/proceedings/2022/0511.pdf
|
EGCN: An Ensemble-based Learning Framework for Exploring Effective Skeleton-based Rehabilitation Exercise Assessment
|
Recently, some skeleton-based physical therapy systems have been attempted to automatically evaluate the correctness or quality of an exercise performed by rehabilitation subjects. However, in terms of algorithms and evaluation criteria, the task remains not fully explored regarding making full use of different skeleton features. To advance the prior work, we propose a learning framework called Ensemble-based Graph Convolutional Network (EGCN) for skeleton-based rehabilitation exercise assessment. As far as we know, this is the first attempt that utilizes both two skeleton feature groups and investigates different ensemble strategies for the task. We also examine the properness of existing evaluation criteria and focus on evaluating the prediction ability of our proposed method. We then conduct extensive cross-validation experiments on two latest public datasets: UI-PRMD and KIMORE. Results indicate that the model-level ensemble scheme of our EGCN achieves better performance than existing methods. Code is available: https://github.com/bruceyo/EGCN.
|
['Keith C.C. Chan', 'Gong Chen', 'Xiang Zhang', 'Yan Liu', 'Bruce X.B. Yu']
|
2022-07-01
| null | null | null |
ijcai-2022-7
|
['action-assessment']
|
['computer-vision']
|
[ 3.39961052e-01 -6.57908246e-02 -4.91352260e-01 -1.83355600e-01
-6.18546188e-01 1.15785547e-01 2.01950192e-01 -1.30228117e-01
-3.52052003e-01 8.30767214e-01 6.59649134e-01 -1.58064976e-01
-4.22126502e-01 -8.37986588e-01 -2.22229823e-01 -3.57958317e-01
-2.47538164e-01 2.03562498e-01 1.94284022e-01 -1.89452752e-01
2.03002185e-01 3.30592930e-01 -1.48172235e+00 2.79317796e-01
8.29312503e-01 7.26629734e-01 -2.08247647e-01 6.93904161e-01
4.52831179e-01 6.77863896e-01 -3.40539277e-01 -2.29013160e-01
1.25554636e-01 -7.32582629e-01 -1.11374605e+00 4.32690606e-03
1.25377893e-01 -3.75696570e-01 -5.93993664e-01 5.24197578e-01
1.17814970e+00 2.70637482e-01 6.03879154e-01 -1.00197804e+00
-4.83988434e-01 5.21928549e-01 -3.37219745e-01 3.88481766e-01
3.31073314e-01 4.76574093e-01 9.31089282e-01 -5.07081032e-01
4.77361202e-01 6.47246301e-01 9.14546072e-01 7.01007783e-01
-7.98376679e-01 -5.50933361e-01 -1.70106858e-01 5.06682158e-01
-1.28162324e+00 -4.39268053e-01 9.04541671e-01 -3.61416966e-01
7.52991736e-01 2.52643794e-01 8.66061687e-01 1.16716564e+00
1.99111268e-01 8.21362078e-01 1.24279094e+00 -4.83190149e-01
1.09159209e-01 -8.06260169e-01 2.95575947e-01 8.52032006e-01
3.13668191e-01 7.72425383e-02 -5.80096304e-01 4.93781082e-02
7.14332640e-01 -2.32968822e-01 -4.03502136e-01 -1.67569026e-01
-9.23466384e-01 5.02787709e-01 3.65157694e-01 4.77933377e-01
-5.21620512e-01 4.99249727e-01 5.86415112e-01 -1.98811269e-03
5.61245620e-01 1.06253065e-01 -3.31145078e-01 -4.46332812e-01
-9.26056743e-01 2.20704138e-01 6.01196945e-01 2.75248587e-01
1.56131133e-01 6.16128615e-04 -3.14880133e-01 8.86369467e-01
1.99273452e-01 -2.43991390e-02 6.44852400e-01 -1.00914693e+00
2.41430551e-01 7.94917047e-01 -5.13496101e-01 -8.87730002e-01
-5.81050754e-01 -5.55559933e-01 -7.26009250e-01 5.00559032e-01
3.51540446e-01 -3.68399829e-01 -6.73306525e-01 1.46684003e+00
3.76096368e-01 3.65463287e-01 -2.93524593e-01 1.13746560e+00
1.11685526e+00 -4.03765947e-01 3.01751465e-01 2.44950905e-01
1.29956973e+00 -9.63178456e-01 -5.01366317e-01 1.36027962e-01
8.04298520e-01 -3.30264121e-01 8.90231788e-01 3.74012798e-01
-9.15580690e-01 -5.95097661e-01 -1.03132939e+00 1.83184505e-01
3.91447963e-03 4.48948473e-01 8.87020111e-01 8.06447804e-01
-9.56535757e-01 1.07800579e+00 -1.08494043e+00 -4.60057974e-01
6.08228862e-01 3.50358158e-01 -5.67949593e-01 -4.87566143e-02
-1.20849383e+00 8.78732860e-01 2.75292784e-01 3.72250050e-01
-7.36620605e-01 -4.49992090e-01 -5.98555684e-01 -2.65398532e-01
3.83869261e-01 -1.12263429e+00 1.03678167e+00 -7.59621382e-01
-1.56664944e+00 7.11611211e-01 1.28144696e-01 -2.69594640e-01
6.52048647e-01 -4.46161091e-01 -3.66678804e-01 2.65457392e-01
-1.54552251e-01 1.99342147e-01 1.61663249e-01 -9.51852858e-01
-1.40289888e-01 -6.67247355e-01 1.12202920e-01 5.48449934e-01
-3.82317126e-01 9.72407982e-02 -3.48739713e-01 -5.81581652e-01
1.59410201e-02 -9.48890388e-01 -2.88464874e-01 2.45783716e-01
-4.77976710e-01 -3.97029340e-01 4.51216072e-01 -8.59791756e-01
1.57041740e+00 -1.73710489e+00 2.38573313e-01 8.48731175e-02
3.45348537e-01 3.46686274e-01 3.67770605e-02 7.13012397e-01
-2.69125760e-01 3.28930080e-01 -3.66849452e-01 -1.88701063e-01
-1.73087806e-01 3.47198009e-01 5.72997332e-01 5.12655318e-01
-5.89546934e-02 9.76444960e-01 -6.62395358e-01 -7.98007488e-01
2.48537138e-01 5.52282631e-01 -4.90441322e-01 -9.94318426e-02
2.42877349e-01 6.48930967e-01 -7.88430870e-01 7.59407103e-01
1.42193615e-01 -2.00726673e-01 3.59825343e-01 -3.60433340e-01
2.94011444e-01 -3.47995141e-04 -1.05530131e+00 2.03561711e+00
-8.65364969e-02 3.03932279e-01 -3.83086383e-01 -1.02961993e+00
5.64493179e-01 3.73248726e-01 1.15432680e+00 -4.25465107e-01
2.40545064e-01 5.69955297e-02 2.09060594e-01 -7.30899215e-01
3.23555321e-01 -3.72698456e-02 2.28443891e-01 4.96200800e-01
4.61874343e-02 5.63169479e-01 2.17566490e-01 -1.83179509e-02
1.67010343e+00 7.61103213e-01 4.90820706e-01 1.38719734e-02
4.56278443e-01 -6.74284548e-02 5.72762847e-01 4.17453349e-01
-6.37313604e-01 8.72662187e-01 7.95369595e-02 -2.13668659e-01
-6.02439463e-01 -7.74835885e-01 -8.32233671e-03 9.47830915e-01
-8.37474912e-02 -6.70338392e-01 -7.53213942e-01 -8.74019206e-01
1.01712488e-01 2.88058162e-01 -6.72097147e-01 -2.34505519e-01
-6.73217595e-01 -8.80709350e-01 1.02022576e+00 8.94543707e-01
7.21011579e-01 -1.22549784e+00 -5.34673691e-01 3.42996344e-02
-5.14964819e-01 -7.64199018e-01 -4.11743432e-01 -3.31157655e-01
-1.27793550e+00 -1.50525808e+00 -7.08323777e-01 -3.62336546e-01
2.81964809e-01 -8.45219567e-02 1.08870471e+00 6.45034254e-01
-3.37904751e-01 6.18956208e-01 -6.27780437e-01 -2.32981518e-02
-1.15847439e-01 1.69164643e-01 -1.25658393e-01 -2.80054778e-01
2.09001511e-01 -9.20105457e-01 -1.23384881e+00 2.50989646e-01
-6.94752991e-01 -1.96805242e-02 6.77452803e-01 5.87676287e-01
5.50482512e-01 7.54569545e-02 6.67517900e-01 -6.85539365e-01
7.87538707e-01 -4.35214907e-01 4.92226928e-01 1.73841506e-01
-8.22203577e-01 -2.10508585e-01 1.01182666e-02 -2.12377265e-01
-7.99920917e-01 -4.77279648e-02 -4.94523048e-01 -1.63958117e-01
-2.92944998e-01 8.56536567e-01 8.50171819e-02 -1.76444739e-01
7.46483505e-01 4.34088148e-02 1.09022759e-01 -7.71352291e-01
1.42021984e-01 6.74554586e-01 5.33088386e-01 -8.24508131e-01
3.51627141e-01 3.08634907e-01 1.31998703e-01 -6.32956088e-01
-5.92639685e-01 -5.90535998e-01 -7.67090261e-01 -8.16921592e-01
8.45633090e-01 -7.41645515e-01 -6.67866945e-01 5.61712086e-01
-5.54024339e-01 -4.67136860e-01 -2.27092877e-01 6.28946364e-01
-7.50166893e-01 9.11067963e-01 -7.24560261e-01 -7.05535710e-01
-8.26157629e-01 -7.77575552e-01 9.85006809e-01 -9.29605030e-03
-3.65471482e-01 -9.76251841e-01 5.24926007e-01 9.70589519e-01
3.71049345e-01 7.33595967e-01 7.11123586e-01 -5.20971298e-01
2.89184134e-02 -2.01547951e-01 6.27690600e-03 3.84843051e-01
2.53236949e-01 -5.07568605e-02 -6.83077991e-01 -1.64226919e-01
-4.49481189e-01 -3.08638513e-01 9.55540419e-01 4.56451088e-01
1.44918966e+00 1.20119847e-01 -2.78707772e-01 3.72484654e-01
1.31971765e+00 -3.50490004e-01 1.04785967e+00 4.28744078e-01
6.48212075e-01 3.36404413e-01 4.45107341e-01 4.34113532e-01
4.86149549e-01 7.35447943e-01 4.17520970e-01 -1.76963612e-01
-7.58429885e-01 -6.45633787e-02 2.88112193e-01 6.84822440e-01
-1.12785220e+00 -1.59072280e-01 -1.00367725e+00 4.73968297e-01
-1.84233475e+00 -1.11183560e+00 -2.91826338e-01 1.86069429e+00
8.76951575e-01 1.91293601e-02 4.33505654e-01 4.25154805e-01
5.49329877e-01 7.36545548e-02 -4.38860685e-01 -7.83998333e-03
2.41229370e-01 6.43185496e-01 3.03240746e-01 -1.73142418e-01
-1.05395889e+00 4.74667132e-01 6.18548679e+00 7.65055835e-01
-6.37912273e-01 1.89855978e-01 3.21351886e-01 4.95292582e-02
6.72833845e-02 5.04519138e-03 -2.70614386e-01 2.38068491e-01
8.17569256e-01 9.56614390e-02 9.04738009e-02 5.76205730e-01
5.05749881e-01 -1.22269951e-01 -7.64046788e-01 4.63691384e-01
4.04246710e-02 -1.00149608e+00 -3.78345281e-01 1.65793315e-01
3.83621603e-01 2.67897725e-01 -3.83047730e-01 3.40697259e-01
2.82364905e-01 -1.04339266e+00 2.41081253e-01 8.86868358e-01
6.71980143e-01 -3.93544316e-01 7.34864056e-01 2.09826380e-01
-1.20712924e+00 1.64305151e-01 9.81940180e-02 -1.53212294e-01
9.25507396e-02 3.78772765e-01 -3.46139252e-01 1.22234142e+00
7.46709824e-01 8.79050374e-01 -7.40014553e-01 1.33624482e+00
-3.09285164e-01 1.12259805e+00 -8.11067149e-02 1.78021118e-01
-2.30168402e-01 6.26388043e-02 4.71577346e-01 9.27426398e-01
2.62460679e-01 2.61343539e-01 2.69195408e-01 4.91582006e-01
-3.96548137e-02 4.07538652e-01 -4.45856273e-01 -3.77523750e-02
-1.10644802e-01 1.24839008e+00 -6.07370913e-01 -3.68678570e-02
-3.27232152e-01 8.45175505e-01 3.44253123e-01 2.16716871e-01
-7.43009388e-01 -6.00452535e-02 4.50209320e-01 2.36402273e-01
-1.36218190e-01 -3.01107675e-01 -4.50701326e-01 -1.14428246e+00
-3.07274964e-02 -9.62072194e-01 7.33850539e-01 -9.04808462e-01
-1.28956485e+00 3.29745054e-01 -6.00216500e-02 -1.26458931e+00
4.13485803e-02 -5.43306231e-01 -8.38128805e-01 7.15711832e-01
-1.25733435e+00 -1.24919736e+00 -5.42654276e-01 6.61526382e-01
4.35194910e-01 -1.88659411e-02 1.00867331e+00 4.93487626e-01
-9.08586085e-01 4.07686323e-01 -2.46385813e-01 5.18266797e-01
6.02724254e-01 -1.07722235e+00 -1.11691080e-01 7.16092706e-01
-2.77146846e-01 4.73714501e-01 5.38819611e-01 -9.87969160e-01
-1.21166217e+00 -9.17087436e-01 5.97336113e-01 -4.58847433e-01
5.19124269e-01 4.88093257e-01 -7.91094720e-01 5.19755721e-01
2.47070149e-01 9.96367410e-02 1.06890905e+00 3.62273782e-01
3.47958356e-02 1.60673559e-01 -1.02796996e+00 6.38219178e-01
1.57225621e+00 -1.63123667e-01 -4.70969200e-01 3.84368032e-01
8.19099769e-02 -4.24223840e-01 -1.59632003e+00 8.44309151e-01
9.25419748e-01 -9.17710066e-01 8.77083242e-01 -9.38845575e-01
8.77241492e-01 -2.12994069e-01 -8.27926993e-02 -1.04806972e+00
-2.88418651e-01 -5.51873334e-02 -2.12063566e-01 9.87873852e-01
8.90509412e-02 -4.55983490e-01 1.16706705e+00 4.94761854e-01
-5.04697800e-01 -1.16904461e+00 -8.75705242e-01 -7.12179780e-01
1.58861250e-01 -6.14432514e-01 3.66566598e-01 8.52657378e-01
1.29449487e-01 1.19203031e-01 -5.44394314e-01 -1.02790833e-01
6.40332997e-01 9.87527668e-02 7.38585591e-01 -1.08453357e+00
-4.78281617e-01 -4.76662278e-01 -7.61405051e-01 -3.85143757e-01
-6.79449039e-03 -1.22113299e+00 -2.58277923e-01 -2.07592201e+00
3.88980538e-01 -1.69208959e-01 -6.14915252e-01 8.73571157e-01
-2.69142687e-01 4.72738087e-01 -1.87730901e-02 1.81706205e-01
-5.50736785e-01 6.80558741e-01 1.48103619e+00 -3.80403502e-03
4.63406779e-02 2.09452212e-01 -8.08766365e-01 6.40248358e-01
1.24571991e+00 -3.98332477e-01 -3.25474113e-01 -3.04694861e-01
-7.89299533e-02 1.88107103e-01 6.78537309e-01 -1.29010701e+00
2.08211760e-03 -1.82301372e-01 4.14793909e-01 -5.42376935e-01
1.43890858e-01 -4.50602680e-01 5.04942358e-01 6.27069056e-01
-4.25045580e-01 -8.75022858e-02 -1.08028442e-01 5.61805248e-01
6.70159757e-02 -1.18974090e-01 3.23371053e-01 -1.47712201e-01
-7.55894303e-01 4.71449316e-01 -3.39458674e-01 -4.85721491e-02
7.99647570e-01 -5.18031061e-01 -2.30519459e-01 -2.91083932e-01
-9.75374043e-01 2.27441639e-01 2.21184477e-01 2.78866738e-01
8.05766642e-01 -1.52427709e+00 -9.38003778e-01 -4.69946772e-01
1.97993547e-01 -5.69683075e-01 4.72123563e-01 1.30011845e+00
-5.28419435e-01 2.68254299e-02 -4.11413848e-01 -3.75275522e-01
-1.54171848e+00 1.30355014e-02 7.06247330e-01 -6.40457928e-01
-9.12476659e-01 2.88898408e-01 -5.45158446e-01 -5.36805391e-01
-8.81936550e-02 1.41902462e-01 -6.15379214e-01 -4.49017793e-01
6.26861826e-02 7.81759501e-01 1.82796955e-01 -6.36976659e-01
-3.34037900e-01 4.82372254e-01 3.65659267e-01 5.95037416e-02
1.48304069e+00 1.72642730e-02 4.83702533e-02 1.54611036e-01
8.18980932e-01 -1.84942544e-01 -8.50675404e-01 -1.75803453e-01
-3.02892439e-02 -3.26571584e-01 2.05822632e-01 -1.03536510e+00
-1.41107523e+00 6.58541024e-01 1.02271700e+00 -2.67897666e-01
1.47748518e+00 -3.08850586e-01 8.19350660e-01 4.45333719e-02
3.63784015e-01 -1.05442727e+00 1.94298625e-01 8.68851095e-02
9.13066924e-01 -1.07891023e+00 4.24843252e-01 -2.86298364e-01
-5.65663218e-01 1.10839331e+00 8.26187074e-01 -2.70312697e-01
8.27650666e-01 -1.75058603e-01 -1.40194967e-01 -4.97951537e-01
-5.17034471e-01 -6.25407159e-01 5.94481111e-01 6.02339149e-01
8.58653247e-01 1.88085422e-01 -1.21928644e+00 7.38124728e-01
-3.11331954e-02 6.34250283e-01 2.76421189e-01 1.05178297e+00
-3.62537026e-01 -1.37590396e+00 6.53249696e-02 7.14091897e-01
-8.01306844e-01 1.54881896e-02 -5.09782076e-01 9.33822572e-01
2.79420950e-02 9.41252589e-01 -7.63802767e-01 -7.06258297e-01
5.70554674e-01 2.06014052e-01 6.42416239e-01 -5.88032186e-01
-8.19485784e-01 -1.58437908e-01 6.71587884e-01 -6.00918472e-01
-9.04870629e-01 -6.59266055e-01 -1.31432438e+00 -3.74420196e-01
-3.17734033e-01 -1.46603867e-01 3.68990928e-01 1.03573549e+00
2.95685172e-01 1.00976956e+00 2.55634725e-01 -6.66874170e-01
-6.28245294e-01 -1.17520523e+00 -6.23323202e-01 5.25926232e-01
-4.21213239e-01 -9.19771016e-01 5.80562092e-02 -9.32221040e-02]
|
[7.320581912994385, 0.22588685154914856]
|
9d390157-fbb3-4fa0-814d-b7c04ed4f848
|
iteratively-selecting-an-easy-reference-frame
|
2112.12402
| null |
https://arxiv.org/abs/2112.12402v1
|
https://arxiv.org/pdf/2112.12402v1.pdf
|
Iteratively Selecting an Easy Reference Frame Makes Unsupervised Video Object Segmentation Easier
|
Unsupervised video object segmentation (UVOS) is a per-pixel binary labeling problem which aims at separating the foreground object from the background in the video without using the ground truth (GT) mask of the foreground object. Most of the previous UVOS models use the first frame or the entire video as a reference frame to specify the mask of the foreground object. Our question is why the first frame should be selected as a reference frame or why the entire video should be used to specify the mask. We believe that we can select a better reference frame to achieve the better UVOS performance than using only the first frame or the entire video as a reference frame. In our paper, we propose Easy Frame Selector (EFS). The EFS enables us to select an 'easy' reference frame that makes the subsequent VOS become easy, thereby improving the VOS performance. Furthermore, we propose a new framework named as Iterative Mask Prediction (IMP). In the framework, we repeat applying EFS to the given video and selecting an 'easier' reference frame from the video than the previous iteration, increasing the VOS performance incrementally. The IMP consists of EFS, Bi-directional Mask Prediction (BMP), and Temporal Information Updating (TIU). From the proposed framework, we achieve state-of-the-art performance in three UVOS benchmark sets: DAVIS16, FBMS, and SegTrack-V2.
|
['Euntai Kim', 'Hongje Seong', 'Youngjo Lee']
|
2021-12-23
| null | null | null | null |
['unsupervised-video-object-segmentation']
|
['computer-vision']
|
[ 4.91417915e-01 -1.40300974e-01 -4.58082378e-01 -1.88790664e-01
-4.04900938e-01 -3.33257645e-01 3.83811504e-01 -1.29553899e-01
-5.82351506e-01 5.36248863e-01 -1.79304019e-01 -3.39351356e-01
1.51354179e-01 -8.23145688e-01 -6.61305726e-01 -9.16924417e-01
4.54315603e-01 2.75275141e-01 1.14940608e+00 1.23487450e-01
2.34597921e-01 4.36730772e-01 -1.45194459e+00 3.17883551e-01
9.50637758e-01 1.11284220e+00 7.25926876e-01 4.91155565e-01
-3.89266163e-01 8.44507277e-01 -5.64632833e-01 -1.67637691e-01
4.18399245e-01 -6.76763117e-01 -1.03824437e+00 5.54217577e-01
2.70897508e-01 -3.92839164e-01 -2.24111542e-01 1.06692457e+00
1.08218879e-01 3.76881331e-01 3.91981989e-01 -1.26775622e+00
4.41098027e-02 5.52598655e-01 -7.77954459e-01 4.40537512e-01
8.70274901e-02 1.85168758e-01 8.31564605e-01 -5.98685801e-01
9.00743961e-01 9.95081067e-01 5.37022464e-02 4.29435462e-01
-9.08114135e-01 -4.29599226e-01 5.57100713e-01 3.62856686e-01
-1.51955485e+00 -3.63719583e-01 8.16166639e-01 -4.66951370e-01
4.23385173e-01 4.43241596e-01 7.40123808e-01 5.41003406e-01
1.22294575e-01 1.01759064e+00 7.73072243e-01 -4.78920788e-01
4.34417337e-01 -1.79398134e-01 2.15041474e-01 7.59251535e-01
6.11034129e-03 -1.55439064e-01 -4.19375211e-01 4.15483825e-02
6.97953165e-01 1.07435912e-01 -5.73834956e-01 -1.37207597e-01
-1.17868149e+00 4.69482571e-01 1.69875592e-01 3.46101165e-01
-4.74787951e-01 2.21223637e-01 1.48386970e-01 -8.46711099e-02
2.65451640e-01 2.54206336e-03 -2.81663418e-01 7.03085586e-03
-1.37567961e+00 -4.82488349e-02 4.73250449e-01 6.42905414e-01
8.96383107e-01 2.97303847e-03 -3.86182368e-01 7.07402289e-01
2.74559736e-01 1.73573926e-01 2.64405131e-01 -1.14284611e+00
3.51084471e-01 6.74585164e-01 2.57172942e-01 -7.16483653e-01
-4.10705432e-03 -3.75736803e-01 -3.50225687e-01 3.03820014e-01
5.82197964e-01 -2.57257491e-01 -1.38033473e+00 1.47520721e+00
6.67729020e-01 6.03840351e-01 -7.58100003e-02 1.10507357e+00
7.40899622e-01 1.05245280e+00 1.94271859e-02 -5.96523702e-01
1.18977618e+00 -1.08902645e+00 -6.25295997e-01 -2.89899677e-01
4.23317164e-01 -7.65397966e-01 6.75424337e-01 5.56926906e-01
-8.49205375e-01 -6.78796649e-01 -9.51601684e-01 2.33793497e-01
8.80780071e-02 3.56015682e-01 3.05630833e-01 5.88970900e-01
-7.99728692e-01 3.76032084e-01 -1.00360489e+00 -2.30243146e-01
2.84778416e-01 3.50455940e-01 -2.24773735e-01 -7.34393895e-02
-8.66019726e-01 3.47016126e-01 6.06686056e-01 1.45906657e-01
-1.13436604e+00 -4.41198379e-01 -7.51828015e-01 2.93859169e-02
1.02234173e+00 -5.29695868e-01 9.13472652e-01 -1.52593696e+00
-1.39136589e+00 5.70173323e-01 -5.21573544e-01 -5.31326175e-01
6.30415380e-01 -1.57046303e-01 -1.23507835e-01 4.18473780e-01
1.01464633e-02 8.82298052e-01 1.03238285e+00 -1.35119963e+00
-1.11376059e+00 -5.18006040e-04 3.30495268e-01 2.54185736e-01
1.05792366e-01 2.80996412e-02 -1.12880421e+00 -6.09628975e-01
3.36788833e-01 -1.02056026e+00 -1.76908031e-01 -1.97761029e-01
-6.32981658e-01 -2.40224138e-01 1.07673740e+00 -6.83244228e-01
1.56223190e+00 -2.46088839e+00 2.77634680e-01 1.05563857e-01
5.38479909e-02 4.36843634e-01 -2.11542696e-02 -9.70599279e-02
8.92135724e-02 3.67327370e-02 -2.32586429e-01 -2.08471864e-01
-4.21258301e-01 3.25104147e-01 -2.74450593e-02 3.66215944e-01
-7.39918724e-02 5.08205116e-01 -7.44106591e-01 -8.89762044e-01
3.53780657e-01 2.34845698e-01 -5.27692556e-01 1.81832552e-01
-4.54219520e-01 4.91143346e-01 -4.21922743e-01 5.55126488e-01
7.07446277e-01 -2.42243364e-01 1.41211435e-01 -3.39861482e-01
-3.25848341e-01 -4.18853760e-02 -1.71289408e+00 1.27886117e+00
1.21621817e-01 4.90722805e-01 -2.27613375e-02 -7.49710262e-01
8.75876367e-01 1.51121467e-01 7.00354874e-01 -3.37897182e-01
2.23706618e-01 5.42047434e-02 1.62263647e-01 -4.16780710e-01
3.75100702e-01 2.02877209e-01 5.17970443e-01 1.47691980e-01
-1.60199359e-01 2.78830022e-01 6.66003406e-01 2.60896266e-01
8.88246298e-01 3.98705274e-01 3.24657671e-02 -1.18671812e-01
8.31408739e-01 -4.26856875e-02 1.04025102e+00 6.52142286e-01
-3.20886761e-01 8.02288115e-01 5.85642457e-01 -3.53677928e-01
-6.27367020e-01 -9.10035253e-01 1.39920518e-01 8.86464059e-01
6.52528465e-01 -3.63899082e-01 -1.04543388e+00 -9.41520512e-01
-4.52297121e-01 7.35112906e-01 -4.28849936e-01 1.61942765e-01
-6.68752253e-01 -6.57323241e-01 -1.10018542e-02 2.79499412e-01
8.20757687e-01 -9.65742886e-01 -7.24261761e-01 2.55415171e-01
-3.99027228e-01 -1.25007260e+00 -8.52490485e-01 1.68218434e-01
-6.77628636e-01 -1.10876632e+00 -6.54020727e-01 -6.67752922e-01
8.42627466e-01 4.97105151e-01 5.87589264e-01 3.88523459e-01
1.97920188e-01 -2.65373960e-02 -5.46009183e-01 3.88809592e-02
-3.59584838e-01 -2.36580223e-02 -2.44283333e-01 5.86618185e-01
-8.07360336e-02 -1.50154904e-02 -7.18132734e-01 5.88870585e-01
-1.09232008e+00 4.74151075e-01 2.68129438e-01 5.13681531e-01
1.02530384e+00 4.14008766e-01 1.58156425e-01 -9.30638492e-01
-1.58271447e-01 -2.66608834e-01 -7.78294206e-01 3.41584980e-01
-2.84969330e-01 -1.54997125e-01 4.82293248e-01 -4.97100770e-01
-1.06700790e+00 3.29142690e-01 -2.01678798e-01 -5.97475588e-01
-8.47431272e-02 2.00903714e-01 -2.93088347e-01 8.11824054e-02
1.74597606e-01 3.09675694e-01 -3.47648263e-01 -4.46175396e-01
6.29528835e-02 5.67540288e-01 5.62638342e-01 -4.17033285e-01
4.82451022e-01 4.75382358e-01 -2.47380078e-01 -7.10801721e-01
-5.62459230e-01 -5.19514859e-01 -7.57940888e-01 -4.61512119e-01
1.13071704e+00 -6.69813514e-01 -3.77829850e-01 3.63188386e-01
-1.07035553e+00 -4.95090783e-01 -5.43993376e-02 3.92878920e-01
-1.95003137e-01 5.39245367e-01 -3.93735141e-01 -8.71330917e-01
-2.11281657e-01 -1.72482979e+00 1.08609772e+00 5.06013036e-01
-1.95467487e-01 -5.68534315e-01 -4.70305175e-01 4.61886495e-01
-2.40898371e-01 2.61143506e-01 5.49828589e-01 -3.84196758e-01
-1.02865279e+00 1.96832404e-01 -1.52222663e-01 3.69682282e-01
2.67523348e-01 4.14852887e-01 -6.51630044e-01 -7.52413869e-02
-6.81264177e-02 4.18814570e-01 1.00036848e+00 6.09803498e-01
1.06873071e+00 -6.24430701e-02 -5.95756531e-01 7.39620209e-01
1.48062229e+00 9.47943747e-01 7.81565189e-01 4.41577435e-01
8.90687525e-01 2.48524338e-01 9.65524137e-01 2.32148722e-01
2.77511716e-01 7.83802986e-01 3.27425569e-01 -1.22282617e-01
-2.13549346e-01 1.13770142e-02 5.82510054e-01 3.82098824e-01
-1.14012286e-01 -5.52766263e-01 -7.56094098e-01 5.37350059e-01
-1.84462249e+00 -7.94847429e-01 -2.04257712e-01 2.20538473e+00
5.46277583e-01 4.08100754e-01 1.28895432e-01 2.44747832e-01
9.92485940e-01 2.88774997e-01 -5.04277885e-01 4.60898280e-02
1.09290713e-02 -1.52600273e-01 4.56060767e-01 7.21151173e-01
-1.20257282e+00 1.01510990e+00 4.96231890e+00 1.05184615e+00
-1.25268185e+00 4.36687805e-02 1.06826639e+00 -1.00383505e-01
-1.00276448e-01 3.18332911e-01 -1.06676817e+00 8.16085339e-01
5.08483946e-01 1.94950968e-01 4.40360218e-01 6.57632351e-01
5.18917382e-01 -7.54348278e-01 -9.55308557e-01 8.52144957e-01
-6.42237961e-02 -1.38021982e+00 7.75959715e-02 -1.73036352e-01
7.25236177e-01 -1.93138212e-01 -3.69757086e-01 -9.88321304e-02
-6.67357370e-02 -4.52158153e-01 1.12121773e+00 3.42685848e-01
5.61810076e-01 -6.36954784e-01 7.12532461e-01 4.19952273e-01
-1.45598292e+00 -1.11330871e-03 -7.45955482e-02 3.44916344e-01
3.12386155e-01 4.84495014e-01 -6.59697354e-01 6.43666267e-01
7.90372908e-01 4.05948728e-01 -4.70853448e-01 1.15923941e+00
-1.95485085e-01 7.70984113e-01 -3.80612314e-01 2.91195095e-01
3.80854398e-01 -4.00076717e-01 6.00190580e-01 1.08836389e+00
1.27394766e-01 2.38685235e-01 5.65323889e-01 5.68535984e-01
4.94381003e-02 1.42805502e-01 -1.00037167e-02 1.45001292e-01
4.37286317e-01 1.03396475e+00 -1.49490106e+00 -5.54183125e-01
-3.63358706e-01 1.03834140e+00 -1.36600927e-01 5.47778249e-01
-9.65583980e-01 -1.26991555e-01 3.24749321e-01 3.10809344e-01
6.75021350e-01 -7.80185461e-02 -9.20076221e-02 -9.26043749e-01
-6.34260029e-02 -7.55011976e-01 5.18509388e-01 -8.24478626e-01
-5.12241006e-01 7.04925835e-01 4.73501161e-02 -1.19785702e+00
1.15515806e-01 -2.11124599e-01 -6.02273166e-01 6.58676505e-01
-1.38636899e+00 -7.32991755e-01 -3.66134852e-01 3.36061776e-01
9.60831285e-01 1.17848828e-01 -6.36892542e-02 4.02733296e-01
-8.23797762e-01 2.13403225e-01 -1.16197079e-01 2.50570118e-01
3.94641489e-01 -1.07815325e+00 -2.99021415e-02 1.26057982e+00
1.54846713e-01 3.41222018e-01 6.45628393e-01 -9.20536757e-01
-1.06773484e+00 -1.15449798e+00 6.36369765e-01 -3.61257941e-02
2.43369192e-01 -1.40191838e-01 -8.75499249e-01 5.94590843e-01
4.30641472e-02 1.16339214e-01 1.54506147e-01 -5.73265910e-01
3.28000873e-01 -3.37665379e-01 -9.83925819e-01 7.24313259e-01
8.21351290e-01 1.79741811e-02 -2.70940721e-01 1.56549007e-01
8.48131299e-01 -6.52248025e-01 -5.05350411e-01 5.63693285e-01
3.59282613e-01 -9.85548198e-01 7.94229746e-01 4.66935635e-02
3.39082986e-01 -1.09867954e+00 -1.09366022e-01 -8.31003964e-01
-1.60788208e-01 -6.79840624e-01 -7.28761777e-02 1.41839945e+00
1.77342847e-01 -2.95387566e-01 8.94517541e-01 4.40100491e-01
-1.47603810e-01 -1.05980718e+00 -9.39512074e-01 -6.01238191e-01
-5.98186433e-01 -4.67945307e-01 5.51988661e-01 6.86331153e-01
-5.96512198e-01 2.08291388e-03 -3.54666710e-01 4.10835147e-01
3.62397254e-01 2.47372717e-01 7.61542380e-01 -9.32781696e-01
-3.71774524e-01 -2.69442707e-01 -2.42231742e-01 -1.34760308e+00
-6.57174438e-02 -5.67602456e-01 1.72195494e-01 -1.71650231e+00
2.05750704e-01 -3.62696618e-01 -2.74485528e-01 3.87682915e-01
-5.10337949e-01 3.03453177e-01 5.46736240e-01 1.91366225e-01
-7.99953878e-01 1.74675435e-01 1.38987088e+00 -1.87870592e-01
-4.96398389e-01 -2.12948024e-02 -3.10889155e-01 7.55399048e-01
5.73626399e-01 -4.80277210e-01 -4.91784006e-01 -3.43711972e-01
-3.00579727e-01 2.77268261e-01 1.36589691e-01 -8.70480120e-01
2.34679222e-01 -5.01120925e-01 4.44891900e-01 -6.99594080e-01
2.93014646e-01 -8.50841701e-01 4.21504557e-01 5.96785903e-01
1.38367459e-01 -2.09236011e-01 1.03330538e-01 4.39171404e-01
-1.19055092e-01 -6.20767236e-01 9.77342427e-01 -6.34122416e-02
-1.07897353e+00 3.83684516e-01 -3.94202143e-01 -1.80457175e-01
1.29945111e+00 -5.56847274e-01 -2.03675106e-01 -6.23951294e-02
-7.93908358e-01 3.69239002e-01 6.20535612e-01 2.38776550e-01
5.87031841e-01 -9.51387703e-01 -5.27821600e-01 1.55667871e-01
-2.26815090e-01 1.88938931e-01 2.86298066e-01 8.88514936e-01
-7.59115577e-01 1.31584797e-02 2.09404707e-01 -8.07766199e-01
-1.59257007e+00 5.89637280e-01 3.79326910e-01 -1.52340740e-01
-7.24886596e-01 7.98160970e-01 5.79526603e-01 3.95291179e-01
1.88941523e-01 -4.33788210e-01 -3.68178308e-01 -4.33806218e-02
5.51283896e-01 3.83043379e-01 -1.96662471e-01 -8.40454221e-01
-3.16363215e-01 4.84693706e-01 3.70890945e-02 -6.90910295e-02
9.59029019e-01 -4.06982899e-01 -9.42037627e-02 3.77958775e-01
8.39064419e-01 1.17739059e-01 -1.44013572e+00 -1.32145599e-01
-8.94146692e-03 -6.74979329e-01 1.66920915e-01 -5.68850636e-01
-1.43434596e+00 4.08344299e-01 6.26391113e-01 5.19338958e-02
1.44974756e+00 -1.82804186e-02 8.46884131e-01 -1.32300645e-01
3.24079186e-01 -1.08053637e+00 -2.50223838e-02 3.31573308e-01
4.39552903e-01 -9.07475471e-01 8.30606669e-02 -7.88520277e-01
-8.61509740e-01 9.75552559e-01 7.54474461e-01 1.02252364e-01
4.80618656e-01 1.85902923e-01 -3.40025276e-02 1.06700696e-01
-4.97421741e-01 -4.02217418e-01 3.63443762e-01 1.07929610e-01
1.33955568e-01 -2.06870019e-01 -5.15830040e-01 4.12019402e-01
1.74792260e-01 1.53704450e-01 4.59730744e-01 7.90437758e-01
-7.01063156e-01 -1.12159681e+00 -6.31910920e-01 5.02906919e-01
-6.01958156e-01 6.23645596e-02 -1.59222782e-01 5.38459122e-01
6.97948039e-01 9.94072199e-01 7.86117390e-02 -2.69460410e-01
4.47459035e-02 1.17835946e-01 3.47398579e-01 -6.43496096e-01
-2.81161994e-01 6.71066105e-01 4.57963459e-02 -5.22883117e-01
-5.70781350e-01 -7.59762108e-01 -1.59663165e+00 -2.97076739e-02
-5.23008823e-01 2.38888696e-01 2.67486721e-01 1.25114715e+00
3.14113721e-02 7.48060226e-01 5.29255509e-01 -6.65971518e-01
3.02715778e-01 -6.36489153e-01 -4.34484720e-01 2.62497038e-01
2.50760198e-01 -6.97694480e-01 -1.62791088e-01 4.88289595e-01]
|
[9.144041061401367, -0.30029332637786865]
|
7cfc273c-71cb-49bb-9cf8-3e8ecb12ecc1
|
can-chatgpt-pass-an-introductory-level
|
2305.02230
| null |
https://arxiv.org/abs/2305.02230v2
|
https://arxiv.org/pdf/2305.02230v2.pdf
|
Can ChatGPT Pass An Introductory Level Functional Language Programming Course?
|
The recent introduction of ChatGPT has drawn significant attention from both industry and academia due to its impressive capabilities in solving a diverse range of tasks, including language translation, text summarization, and computer programming. Its capability for writing, modifying, and even correcting code together with its ease of use and access is already dramatically impacting computer science education. This paper aims to explore how well ChatGPT can perform in an introductory-level functional language programming course. In our systematic evaluation, we treated ChatGPT as one of our students and demonstrated that it can achieve a grade B- and its rank in the class is 155 out of 314 students overall. Our comprehensive evaluation provides valuable insights into ChatGPT's impact from both student and instructor perspectives. Additionally, we identify several potential benefits that ChatGPT can offer to both groups. Overall, we believe that this study significantly clarifies and advances our understanding of ChatGPT's capabilities and potential impact on computer science education.
|
['Yihan Zhang', 'Xujie Si', 'Brigitte Pientka', 'Chuqin Geng']
|
2023-04-29
| null | null | null | null |
['text-summarization']
|
['natural-language-processing']
|
[-2.85636038e-02 1.05568446e-01 -2.31762081e-01 -1.30489171e-01
-1.03326368e+00 -9.90213096e-01 3.37672353e-01 8.69718313e-01
-2.02059731e-01 4.26190078e-01 3.66054364e-02 -9.18013275e-01
5.60508706e-02 -6.52643681e-01 -7.23659158e-01 -1.21166319e-01
1.15164585e-01 -3.54273915e-02 2.45657027e-01 -3.64483684e-01
8.52666974e-01 4.33436543e-01 -1.57764316e+00 3.23583186e-01
1.47390008e+00 -9.83101223e-03 2.92841673e-01 6.19600236e-01
-5.00573337e-01 1.22748280e+00 -9.31913733e-01 -6.03291690e-01
-4.65174526e-01 -3.96074474e-01 -1.16082203e+00 -5.53053856e-01
5.45972347e-01 1.23572387e-01 8.86869133e-02 7.85486221e-01
4.85402644e-01 1.31108105e-01 -6.67265430e-02 -1.09809089e+00
-6.92313790e-01 9.24619555e-01 -5.13995826e-01 2.93884903e-01
8.59198511e-01 2.35342294e-01 8.86024833e-01 -4.89332020e-01
6.40428126e-01 7.65963554e-01 8.92575145e-01 4.41670120e-01
-1.19805288e+00 -7.12594092e-01 -6.63324147e-02 -4.86796014e-02
-1.21469867e+00 -8.87768269e-02 2.97523737e-01 -6.60623848e-01
1.18941271e+00 2.91471988e-01 1.07396185e+00 4.67792094e-01
6.13630176e-01 7.53039420e-01 8.53451967e-01 -6.53586209e-01
-1.67127430e-01 4.83249158e-01 1.43685639e-01 8.00637603e-01
1.34373307e-01 -5.45209646e-01 -6.55322909e-01 1.76997744e-02
4.62960452e-01 -2.86467731e-01 -1.16359018e-01 8.85794777e-03
-1.32637870e+00 5.56236982e-01 -1.38781801e-01 4.91043866e-01
3.59159052e-01 4.64753240e-01 3.75729263e-01 5.56412876e-01
-2.37247143e-02 8.11643422e-01 -3.43895435e-01 -1.16679895e+00
-7.94301033e-01 1.26887575e-01 1.27101779e+00 1.27467108e+00
2.52744645e-01 2.05819383e-01 1.48439959e-01 7.30957627e-01
1.79693401e-01 2.36974582e-01 3.89185369e-01 -9.56597388e-01
5.34011483e-01 8.78697872e-01 -5.85721493e-01 -9.22476649e-01
1.29710019e-01 -4.68568593e-01 -1.51259363e-01 2.86616057e-01
4.34947938e-01 -2.63304234e-01 -2.57675588e-01 1.60624444e+00
-1.33413389e-01 -9.58942622e-02 -4.48883809e-02 2.48061165e-01
1.22229016e+00 7.66305506e-01 2.41821334e-01 8.64992887e-02
1.28984666e+00 -9.57586288e-01 -2.67002881e-01 -1.56621203e-01
1.27764320e+00 -1.48022139e+00 1.13235939e+00 6.92337513e-01
-1.48169804e+00 -4.06184614e-01 -9.53788519e-01 -2.96349376e-01
-2.96705775e-03 9.41084139e-03 5.13879657e-01 1.20568407e+00
-1.23261201e+00 7.12691009e-01 -7.80788362e-01 -4.20220882e-01
1.99005589e-01 3.06799501e-01 -3.01802903e-01 -1.93090573e-01
-5.93407989e-01 9.77059245e-01 2.89039854e-02 -6.18809700e-01
-2.53836751e-01 -1.28815889e+00 -6.40706539e-01 4.54187810e-01
2.88936198e-01 -6.70185804e-01 1.80626583e+00 -1.55641109e-01
-1.68831134e+00 9.16387618e-01 1.07461557e-01 -1.82594790e-03
3.64494622e-01 -1.34918066e-02 1.10727996e-01 -7.88238049e-02
9.73671973e-02 3.82204086e-01 -1.92812830e-01 -4.68479425e-01
-5.50144017e-01 2.41338864e-01 2.14212045e-01 4.90392387e-01
-5.98449290e-01 1.66560084e-01 -3.76322120e-01 -5.59774518e-01
8.05871487e-02 -8.30577672e-01 -3.79792079e-02 -4.79828380e-02
1.00975275e-01 -4.30814534e-01 4.27741170e-01 -3.88418347e-01
1.46462917e+00 -1.96580863e+00 -1.75114453e-01 5.57448454e-02
3.88204664e-01 1.35221854e-01 1.34072408e-01 1.10823441e+00
-1.05487354e-01 5.40457368e-01 1.40517518e-01 1.19361496e-02
8.12710449e-02 -2.18387067e-01 -2.37135917e-01 1.91547349e-02
-1.68221951e-01 7.70124733e-01 -1.06327140e+00 -5.52208960e-01
3.87358606e-01 3.48700166e-01 -7.36864507e-01 -1.81930527e-01
1.06606716e-02 3.85636866e-01 -4.27204877e-01 3.07734013e-01
2.38990366e-01 -1.70002043e-01 3.26749474e-01 8.21599841e-01
-7.10494518e-01 7.24811733e-01 -9.56624091e-01 1.77173698e+00
-8.95229638e-01 1.11833465e+00 -6.53463155e-02 -6.79311812e-01
9.96586323e-01 4.36298877e-01 1.77346632e-01 -3.82651001e-01
-1.11302309e-01 2.59778470e-01 2.73636311e-01 -4.16483104e-01
7.50723541e-01 4.05128598e-02 -5.53883193e-03 1.05684483e+00
4.64026406e-02 -8.66433501e-01 5.78283608e-01 7.16446102e-01
1.12919414e+00 6.38701245e-02 2.67361701e-01 -5.19653380e-01
4.77185041e-01 1.46172673e-01 5.33281974e-02 7.76399910e-01
5.81862405e-02 2.85643250e-01 8.25884879e-01 3.72640640e-02
-6.26332581e-01 -6.69466734e-01 1.82980657e-01 1.32117128e+00
-3.05798590e-01 -1.06284940e+00 -6.09143436e-01 -1.96430475e-01
-2.92768717e-01 9.00717199e-01 1.00450285e-01 -2.70015210e-01
-6.05929554e-01 -2.97179699e-01 7.21690297e-01 5.25009215e-01
3.80318791e-01 -9.65784729e-01 -5.53232133e-01 1.64949223e-01
-1.55765384e-01 -9.14947212e-01 -5.69494486e-01 4.17729877e-02
-1.05773497e+00 -6.68069959e-01 -3.18422377e-01 -1.49931324e+00
6.58578753e-01 5.58508992e-01 1.15521836e+00 7.09424973e-01
-3.53357166e-01 7.22990811e-01 -2.27264285e-01 -5.00365257e-01
-9.61998522e-01 3.46271694e-01 -4.92007107e-01 -1.16175973e+00
2.51005262e-01 -7.76587605e-01 6.38794750e-02 -1.62852824e-01
-6.59156084e-01 4.11113620e-01 5.21661341e-01 3.07170480e-01
-2.06552640e-01 5.52569404e-02 3.64925236e-01 -1.08227122e+00
9.47300434e-01 -5.77934831e-02 -6.57001376e-01 3.51416796e-01
-5.91487944e-01 -8.26638415e-02 4.87814665e-01 -1.31299555e-01
-8.40422034e-01 -4.08973843e-01 -4.09190387e-01 6.03439510e-01
1.52776748e-01 9.80014205e-01 2.42337629e-01 -8.38467658e-01
7.48206735e-01 1.39488503e-01 -2.00031493e-02 -2.49165818e-02
-1.16141900e-01 3.44996125e-01 2.03793764e-01 -1.10048640e+00
8.14979196e-01 -5.51895320e-01 -2.44778588e-01 -9.84922528e-01
-2.33031526e-01 -4.53894943e-01 -2.56815672e-01 -3.05959880e-01
2.20805377e-01 -1.02660596e+00 -1.29357958e+00 3.96047890e-01
-1.05455327e+00 -5.38910747e-01 -2.03685284e-01 4.57925111e-01
5.85143343e-02 5.03559649e-01 -8.48517537e-01 -4.46204841e-01
-1.19926028e-01 -1.43052375e+00 3.97611439e-01 6.24559164e-01
-6.63738191e-01 -1.37863433e+00 3.45941968e-02 7.25796640e-01
5.69558322e-01 1.08199164e-01 1.16170597e+00 -3.83977085e-01
-6.81285501e-01 -2.91031867e-01 2.11241215e-01 5.29496744e-02
-2.66377896e-01 4.54123765e-01 -6.01512194e-01 -1.68177187e-01
-9.02365074e-02 -1.86810851e-01 1.90429881e-01 3.34522240e-02
7.72435069e-01 -1.16858579e-01 -1.22872211e-01 1.51944175e-01
1.44310725e+00 1.16937429e-01 2.95042604e-01 4.56692159e-01
4.37476188e-01 4.63227153e-01 2.81078219e-01 1.72059178e-01
2.34556332e-01 3.93048882e-01 -9.60228816e-02 5.21241367e-01
-1.32089689e-01 -3.87743711e-01 5.06527722e-01 1.54986084e+00
1.10584283e-02 -5.13343588e-02 -1.41435659e+00 5.00838757e-01
-1.43173409e+00 -8.21212769e-01 -5.32728791e-01 2.00234389e+00
9.11425710e-01 3.16951185e-01 -6.76668882e-02 1.02958947e-01
5.85397542e-01 -2.12815210e-01 1.20387211e-01 -1.17328465e+00
3.84785593e-01 7.37339079e-01 1.15359031e-01 4.66530621e-01
-1.06044412e-01 8.14738572e-01 6.55528069e+00 7.19580472e-01
-1.24594033e+00 -1.34639874e-01 1.52662218e-01 2.86616176e-01
-5.86241186e-01 9.71938446e-02 -7.28881359e-01 3.63353416e-02
1.03167856e+00 -1.07767916e+00 2.37264827e-01 9.07387018e-01
7.05803111e-02 -1.97059840e-01 -9.86453533e-01 6.15946233e-01
-6.10148767e-03 -1.59169805e+00 -1.08964898e-01 9.43278894e-02
1.04628634e+00 -3.59230071e-01 1.31856471e-01 6.80077434e-01
5.90023637e-01 -1.09800553e+00 7.80326903e-01 -2.12399960e-01
5.18963158e-01 -8.35985124e-01 5.59731126e-01 5.55723369e-01
-1.14115155e+00 9.32166949e-02 9.44371447e-02 -8.21557462e-01
-3.63943070e-01 1.74920633e-01 -1.29819262e+00 4.93135005e-01
2.20216081e-01 5.17143071e-01 -7.03583479e-01 1.42483473e+00
-4.30057943e-01 8.19344878e-01 -1.08405009e-01 -6.39008820e-01
-5.46050444e-03 -1.55520633e-01 3.77128989e-01 1.25646985e+00
5.66129148e-01 5.50120711e-01 5.05526543e-01 7.68312633e-01
-5.06886765e-02 4.98841763e-01 -2.52466887e-01 -3.72993767e-01
7.47977495e-01 1.21562719e+00 -9.89114702e-01 -2.25061044e-01
-3.60791355e-01 3.88990909e-01 1.81357369e-01 5.06297648e-02
-6.40309632e-01 -9.08663869e-01 5.21278620e-01 1.59607798e-01
3.70198414e-02 -6.57857597e-01 -9.13282275e-01 -8.79116535e-01
2.52028722e-02 -9.73811269e-01 2.91841733e-03 -7.70733893e-01
-4.83375818e-01 3.41767758e-01 -2.20247611e-01 -1.10708869e+00
1.65498778e-02 -3.55776548e-01 -1.05630064e+00 1.04373240e+00
-8.45385253e-01 -5.50686181e-01 -4.95749861e-01 -4.27864380e-02
4.75317508e-01 -2.35199183e-02 8.92838120e-01 2.13706970e-01
-4.64970201e-01 8.78078759e-01 1.95498802e-02 -1.36211604e-01
8.17626178e-01 -1.34697282e+00 5.78293741e-01 1.04485285e+00
-4.63134097e-03 1.23404026e+00 8.72323871e-01 -3.31287652e-01
-1.90782499e+00 -6.42496705e-01 1.29982257e+00 -4.82615530e-01
8.55174422e-01 9.37494114e-02 -7.10942745e-01 7.14749336e-01
3.22914153e-01 -5.25024891e-01 8.88671696e-01 1.69260085e-01
-2.33166352e-01 1.19530052e-01 -7.41765797e-01 6.67968035e-01
7.27907300e-01 -5.91085076e-01 -7.31274784e-01 3.37895393e-01
5.32111943e-01 -1.06304252e+00 -1.19817209e+00 -3.22939813e-01
4.22960609e-01 -8.74216259e-01 6.66148722e-01 -1.38103604e-01
1.03373694e+00 1.70000449e-01 3.99874032e-01 -1.20160079e+00
2.78180167e-02 -1.02176738e+00 5.65601289e-01 1.52044725e+00
2.61138856e-01 -7.25381076e-01 8.72496307e-01 1.01605701e+00
-7.55920351e-01 -6.02734506e-01 -3.68217558e-01 -5.83351493e-01
6.10854924e-01 -4.67290252e-01 2.01858118e-01 1.13780308e+00
8.20789456e-01 3.18393946e-01 2.28290752e-01 -3.35758239e-01
1.11792862e-01 2.99081296e-01 1.01543558e+00 -1.25849795e+00
-3.72296065e-01 -8.06022704e-01 -4.74645257e-01 -1.03820658e+00
-4.64461856e-02 -1.29994106e+00 -1.53099984e-01 -1.62663007e+00
1.82210982e-01 -3.23883146e-01 6.28869593e-01 7.05632210e-01
8.52669105e-02 2.43676692e-01 3.75978559e-01 -1.41259313e-01
-3.89741838e-01 -2.00100809e-01 1.35313857e+00 2.03972787e-01
-3.71001393e-01 1.18086874e-01 -1.22243929e+00 4.78380054e-01
7.26568103e-01 -2.67322242e-01 -3.19106430e-01 -5.27779341e-01
5.48473835e-01 2.64934510e-01 -1.46069810e-01 -1.18544626e+00
6.30233884e-01 -4.32258695e-01 -1.49698406e-01 8.22738335e-02
-2.89439946e-01 -3.47436488e-01 1.17150703e-02 6.19919538e-01
-6.21287346e-01 4.25601602e-01 8.31374109e-01 -1.13039933e-01
-2.46883199e-01 -7.03683853e-01 6.05308354e-01 -1.87753692e-01
-5.33019841e-01 -3.95805776e-01 -1.09210765e+00 4.29203153e-01
1.03623128e+00 -3.41940999e-01 -5.97247660e-01 -5.15814602e-01
-3.23945642e-01 2.76573420e-01 4.60487217e-01 1.98201537e-01
2.93945462e-01 -7.99552083e-01 -4.84760135e-01 7.68692344e-02
-2.65845563e-02 -1.92397803e-01 1.58460021e-01 9.95488405e-01
-1.24333262e+00 7.33604789e-01 -3.89445782e-01 -5.85747719e-01
-1.91254652e+00 -1.91163763e-01 -3.45219612e-01 -3.36490870e-01
-6.86310351e-01 9.95262742e-01 -3.14938843e-01 -7.47169852e-01
1.55253887e-01 -4.71159846e-01 6.56552091e-02 -1.91055447e-01
6.06389761e-01 3.68698746e-01 3.84776562e-01 1.09132014e-01
-1.15854092e-01 4.46533769e-01 -2.30317280e-01 -1.78597614e-01
1.32103765e+00 7.52413198e-02 -2.35359371e-01 5.61899483e-01
1.02372837e+00 6.09694302e-01 -1.77441090e-01 2.10791141e-01
7.91258439e-02 -2.48902872e-01 -2.57705003e-01 -8.42094958e-01
-6.44333720e-01 1.09093964e+00 -1.26015559e-01 1.65200472e-01
6.88885033e-01 -4.39534158e-01 6.11189306e-01 5.28701246e-01
3.90737742e-01 -5.10801494e-01 1.98767915e-01 9.44676399e-01
5.35533845e-01 -7.11156189e-01 6.22742064e-02 -7.59132922e-01
-4.35981065e-01 1.67777014e+00 7.47967184e-01 2.04374686e-01
1.69793546e-01 2.93856621e-01 -1.63378969e-01 8.55107382e-02
-1.00347364e+00 5.51710784e-01 9.67629906e-03 2.69326746e-01
1.45954669e+00 -1.91156104e-01 -5.42389750e-01 1.43726140e-01
-8.73391807e-01 1.24175437e-01 1.49266672e+00 1.46772993e+00
-7.88343728e-01 -1.45253730e+00 -4.34432715e-01 1.62134513e-01
-5.36922753e-01 -4.95159239e-01 -4.33590502e-01 8.79968345e-01
-4.60423559e-01 8.78666878e-01 -3.22961181e-01 -3.05608392e-01
1.39201388e-01 3.24677318e-01 7.21878052e-01 -1.13559914e+00
-1.32129323e+00 -5.40063798e-01 7.03617558e-02 4.12510633e-02
7.07304850e-02 -6.81323886e-01 -1.40025520e+00 -1.22733331e+00
-3.20362926e-01 8.27519834e-01 8.64794970e-01 9.13466156e-01
3.97437423e-01 7.35456824e-01 -3.60861532e-02 -1.51017278e-01
-5.78900754e-01 -7.35277593e-01 -8.87157768e-02 -5.07740140e-01
-1.39817268e-01 2.46948153e-02 -6.28653467e-02 1.55877203e-01]
|
[9.750856399536133, 7.311288833618164]
|
3747ee65-2299-43dc-a494-770162704002
|
efficient-video-semantic-segmentation-with
|
1912.11844
| null |
https://arxiv.org/abs/1912.11844v1
|
https://arxiv.org/pdf/1912.11844v1.pdf
|
Efficient Video Semantic Segmentation with Labels Propagation and Refinement
|
This paper tackles the problem of real-time semantic segmentation of high definition videos using a hybrid GPU / CPU approach. We propose an Efficient Video Segmentation(EVS) pipeline that combines: (i) On the CPU, a very fast optical flow method, that is used to exploit the temporal aspect of the video and propagate semantic information from one frame to the next. It runs in parallel with the GPU. (ii) On the GPU, two Convolutional Neural Networks: A main segmentation network that is used to predict dense semantic labels from scratch, and a Refiner that is designed to improve predictions from previous frames with the help of a fast Inconsistencies Attention Module (IAM). The latter can identify regions that cannot be propagated accurately. We suggest several operating points depending on the desired frame rate and accuracy. Our pipeline achieves accuracy levels competitive to the existing real-time methods for semantic image segmentation(mIoU above 60%), while achieving much higher frame rates. On the popular Cityscapes dataset with high resolution frames (2048 x 1024), the proposed operating points range from 80 to 1000 Hz on a single GPU and CPU.
|
['Luc van Gool', 'Radu Timofte', 'Matthieu Paul', 'Christoph Mayer']
|
2019-12-26
| null | null | null | null |
['2048']
|
['playing-games']
|
[ 3.87954652e-01 1.05169132e-01 8.89495090e-02 -2.05405563e-01
-6.08065665e-01 -3.22060794e-01 3.32954347e-01 6.93494007e-02
-7.53708124e-01 4.61093575e-01 -1.85901999e-01 -2.67636538e-01
5.30265749e-01 -8.99774015e-01 -6.30970836e-01 -4.90924031e-01
4.16629612e-02 6.10460162e-01 1.12123656e+00 1.23228682e-02
3.97545576e-01 4.81167257e-01 -1.82426035e+00 5.10279179e-01
6.54689431e-01 1.25392568e+00 3.43079597e-01 1.10301900e+00
-1.93846107e-01 1.14417279e+00 -3.28314811e-01 -2.78564245e-02
2.75893480e-01 -2.24456489e-01 -1.14827478e+00 2.58884490e-01
5.02448916e-01 -6.10400379e-01 -1.31762281e-01 7.96956360e-01
1.61597282e-01 2.28126615e-01 1.70683473e-01 -1.06002951e+00
2.32116073e-01 2.82381922e-01 -5.95578790e-01 5.73687255e-01
2.60718256e-01 1.96188405e-01 6.12397194e-01 -5.59639335e-01
8.52328777e-01 1.19029522e+00 6.49954021e-01 4.56685066e-01
-9.48884487e-01 -4.56930369e-01 1.27789527e-01 2.80974358e-01
-1.14598739e+00 -2.75105149e-01 4.33809429e-01 -5.50510228e-01
1.03074634e+00 5.67474514e-02 9.73857701e-01 6.31961584e-01
2.46105324e-02 7.21156061e-01 7.45423675e-01 -1.66703328e-01
4.50380236e-01 -1.57691523e-01 2.74087340e-01 8.38133812e-01
-1.44776225e-01 2.38959901e-02 -4.31336224e-01 1.56884305e-02
9.70507920e-01 -2.08566085e-01 -3.09794575e-01 2.33602766e-02
-1.02956903e+00 7.38454938e-01 3.89539361e-01 2.68634439e-01
-5.14622688e-01 4.46395934e-01 6.18588388e-01 -9.41378102e-02
4.78000998e-01 4.34476845e-02 -5.06000042e-01 -3.07954073e-01
-1.48307204e+00 1.18884437e-01 8.30837011e-01 6.93540692e-01
1.07717919e+00 -3.62836085e-02 -1.87067911e-01 3.88507664e-01
2.37874836e-01 1.37800276e-01 3.71388108e-01 -1.45221651e+00
2.78838903e-01 3.27294618e-01 1.76033750e-01 -9.49716032e-01
-5.36077678e-01 -4.94915396e-02 -4.60396886e-01 3.21657866e-01
5.53509474e-01 -1.18776843e-01 -9.54714715e-01 1.20196640e+00
6.76302433e-01 8.40648234e-01 -1.87575668e-01 1.25463164e+00
7.12397158e-01 8.66786540e-01 3.49904209e-01 -4.63226028e-02
1.49136829e+00 -1.40667593e+00 -3.22176605e-01 -2.57634580e-01
6.46354735e-01 -7.54365623e-01 7.19395936e-01 4.77190703e-01
-1.31032944e+00 -7.91383803e-01 -9.02960956e-01 -4.36444491e-01
-2.32148886e-01 -4.03399058e-02 5.75227797e-01 6.38836622e-01
-1.48826683e+00 7.93818355e-01 -1.18719685e+00 -2.45623514e-01
5.56974053e-01 4.89248842e-01 -4.88954373e-02 2.02029813e-02
-8.82499516e-01 4.42737222e-01 6.44617617e-01 4.26332504e-02
-7.96130657e-01 -8.35527599e-01 -9.30231631e-01 8.03530663e-02
2.21209094e-01 -7.46456206e-01 1.02397609e+00 -1.71244204e+00
-1.76684892e+00 8.49045753e-01 -1.92566380e-01 -6.54234827e-01
6.35945916e-01 -3.06746185e-01 1.08835086e-01 7.63890922e-01
1.46169141e-01 1.13372886e+00 8.59337091e-01 -9.97905612e-01
-1.12715936e+00 -1.89665839e-01 1.26701981e-01 6.54667169e-02
7.35868365e-02 3.98623087e-02 -1.09813964e+00 -3.86290371e-01
-1.47036970e-01 -9.20853734e-01 -3.48208249e-01 -9.56207961e-02
-1.17701404e-01 -7.46096820e-02 1.13350618e+00 -9.36682284e-01
9.78332639e-01 -1.95206428e+00 6.70123622e-02 2.75824159e-01
8.30628797e-02 6.39762282e-01 7.61469826e-02 -1.99486181e-01
1.03882924e-01 -3.83401632e-01 -2.29701132e-01 -5.20757198e-01
-5.26373565e-01 2.46171609e-01 -6.27775192e-02 3.87255251e-01
1.03962079e-01 8.73296738e-01 -9.81827021e-01 -6.97642744e-01
5.62629521e-01 6.71389401e-01 -7.27334380e-01 3.06867242e-01
-2.52123982e-01 5.87822020e-01 -3.46178770e-01 2.65162855e-01
7.76953101e-01 -4.17200953e-01 8.22737962e-02 -1.46579280e-01
-4.51295406e-01 1.35129571e-01 -1.30307782e+00 1.94908023e+00
-3.17999125e-01 7.70691097e-01 2.53856152e-01 -1.05911386e+00
5.90551674e-01 2.18057245e-01 7.28446245e-01 -5.72186291e-01
3.03522378e-01 1.43133581e-01 -5.40942430e-01 -5.14671206e-01
6.42687738e-01 3.96236956e-01 3.29824328e-01 3.26745868e-01
5.43138795e-02 -5.29187173e-02 5.99341094e-01 1.91688702e-01
1.07220268e+00 5.73187649e-01 -5.92746213e-02 -4.61677432e-01
8.70529413e-01 2.97838122e-01 6.43429279e-01 4.52864230e-01
-2.48125657e-01 7.17140198e-01 6.03827834e-01 -9.03931201e-01
-1.23013353e+00 -6.40276790e-01 1.20910361e-01 1.14948463e+00
5.41377604e-01 -4.97475326e-01 -1.19283772e+00 -5.68577111e-01
-3.97977620e-01 3.15349221e-01 -5.06567061e-01 2.67433554e-01
-9.34330404e-01 -4.16487217e-01 2.10189357e-01 6.67060614e-01
8.13514888e-01 -1.08277237e+00 -1.35375369e+00 4.78652626e-01
-2.35082239e-01 -1.48706758e+00 -3.75630856e-01 -9.78104100e-02
-9.59007263e-01 -1.27216017e+00 -6.22660220e-01 -6.85257494e-01
4.69140500e-01 2.60777444e-01 1.21461415e+00 2.92952240e-01
-2.68580884e-01 2.56158292e-01 -2.99344808e-01 1.75595909e-01
-4.16825056e-01 9.58252475e-02 -6.25207543e-01 9.35664624e-02
1.78381816e-01 -3.32759440e-01 -9.22450602e-01 1.59786403e-01
-9.77779686e-01 7.04048872e-01 1.33277208e-01 5.15968442e-01
7.48455524e-01 -2.72997439e-01 -8.83439779e-02 -9.95177984e-01
-2.74807960e-01 -3.68206888e-01 -9.88721490e-01 -5.96151687e-02
-2.00610802e-01 -9.31155458e-02 6.54179335e-01 -1.09063305e-01
-9.91821051e-01 5.22345841e-01 -5.42567492e-01 -4.60945606e-01
-3.44616026e-01 -1.49165705e-01 3.02446812e-01 -9.81580615e-02
2.23381281e-01 3.06387451e-02 5.79876937e-02 -2.10465804e-01
3.49278092e-01 5.33428311e-01 5.96587181e-01 -3.28838229e-01
1.71163306e-01 7.20181346e-01 -1.96397364e-01 -8.81823242e-01
-7.84949422e-01 -6.94287837e-01 -8.68630409e-01 -4.82096344e-01
1.34368944e+00 -1.02051485e+00 -6.19208038e-01 5.43896496e-01
-1.27411079e+00 -8.21794212e-01 -2.49902874e-01 3.48601609e-01
-8.35483313e-01 3.30476046e-01 -9.80417430e-01 -4.25637275e-01
-5.08940578e-01 -1.37924230e+00 1.45710218e+00 4.80562896e-01
-1.11793831e-01 -9.73330021e-01 -1.22308448e-01 5.70327699e-01
3.75258535e-01 4.52136487e-01 2.63456136e-01 -1.36883378e-01
-9.42600965e-01 2.48884499e-01 -6.52906120e-01 1.87744081e-01
-4.01923150e-01 2.35557541e-01 -9.32814777e-01 -1.65411711e-01
-1.45032123e-01 -4.60103229e-02 1.12421107e+00 7.39918590e-01
1.25663209e+00 1.08089652e-02 -4.46325719e-01 9.00453627e-01
1.62335598e+00 2.39869431e-01 9.38332438e-01 4.18150336e-01
9.55642700e-01 3.93259525e-01 6.20011687e-01 3.71622443e-01
5.48526227e-01 6.81772232e-01 5.01025259e-01 -3.50840598e-01
-3.88732105e-01 2.44357273e-01 2.86020964e-01 5.94379187e-01
-2.60490507e-01 -2.71688253e-02 -8.71701658e-01 5.38843334e-01
-2.03591657e+00 -7.88931787e-01 -4.74479139e-01 2.08020282e+00
4.41798627e-01 1.29655853e-01 3.36174935e-01 1.54221937e-01
7.87230492e-01 1.44461527e-01 -2.41667509e-01 -6.46399260e-01
2.39929080e-01 4.37729299e-01 5.99741995e-01 7.54757404e-01
-1.23748887e+00 1.33130729e+00 5.96363115e+00 7.64467120e-01
-1.22777331e+00 3.87510449e-01 9.85651314e-01 8.14914703e-02
1.03123531e-01 4.74440455e-02 -7.61729240e-01 6.01600587e-01
1.01305223e+00 5.41942894e-01 3.21634829e-01 9.63950098e-01
3.58036220e-01 -5.39545834e-01 -7.18773246e-01 9.20675933e-01
5.22608906e-02 -1.57812369e+00 -1.49893329e-01 -1.31049782e-01
7.38720953e-01 2.78538615e-01 -4.87608790e-01 -6.75296709e-02
3.74073684e-02 -6.35007501e-01 7.80132174e-01 3.70283365e-01
6.77862287e-01 -9.11559284e-01 6.48395061e-01 1.97368786e-01
-1.39237964e+00 7.39422143e-02 -2.36970469e-01 1.60400607e-02
4.10305858e-01 6.14255548e-01 -5.30598164e-01 3.05113465e-01
8.79498959e-01 8.35159183e-01 -4.11946416e-01 8.33531380e-01
-8.56346637e-02 4.13275182e-01 -4.68600929e-01 3.56328428e-01
7.02353716e-01 -2.55101621e-01 1.98191524e-01 1.46580219e+00
2.35738471e-01 1.34363621e-01 3.70375156e-01 6.15149736e-01
3.38286102e-01 5.96989952e-02 -8.36079642e-02 5.76509714e-01
-2.19069459e-02 1.43768680e+00 -1.33725369e+00 -7.38161325e-01
-4.98022109e-01 1.39231408e+00 2.10288569e-01 2.44282991e-01
-1.09662724e+00 -1.61159903e-01 5.68508625e-01 1.94989800e-01
6.27080977e-01 -3.02514017e-01 -2.12775469e-01 -1.09554935e+00
-3.66498023e-01 -3.71813089e-01 5.31689823e-01 -8.29815209e-01
-4.82765406e-01 7.25775301e-01 -4.26270247e-01 -8.27043056e-01
-1.32516563e-01 -4.83167708e-01 -4.12291020e-01 5.93523383e-01
-1.73158252e+00 -9.85879302e-01 -6.22043908e-01 7.62744427e-01
8.70820165e-01 2.91998297e-01 5.37781715e-01 4.69042301e-01
-5.57068288e-01 7.61506567e-03 -3.04530889e-01 2.42252141e-01
1.30106091e-01 -1.13511431e+00 4.61947024e-01 1.01035678e+00
6.22160546e-02 -1.62705645e-01 3.69131595e-01 -6.03430986e-01
-9.39327717e-01 -1.33951128e+00 7.14183748e-01 -6.88715056e-02
2.82054394e-01 -1.28764331e-01 -9.01532590e-01 4.59343463e-01
6.39525503e-02 3.69438589e-01 2.03500137e-01 -6.22256994e-01
7.49598444e-02 1.04895167e-01 -1.02800226e+00 2.45200187e-01
7.11772501e-01 -1.52264386e-01 -2.20308781e-01 2.32389495e-01
6.73392713e-01 -7.84576833e-01 -7.28030145e-01 1.74126793e-02
4.27723616e-01 -1.34649622e+00 1.00485182e+00 -2.86187008e-02
4.78854656e-01 -4.92217630e-01 2.50782132e-01 -6.68330193e-01
-1.55156642e-01 -7.54226148e-01 -1.34514466e-01 8.24906707e-01
-8.45648870e-02 -3.03393662e-01 1.01192319e+00 6.44451618e-01
-1.09879740e-01 -5.88388979e-01 -9.64612544e-01 -3.16364259e-01
-4.79478121e-01 -6.70097530e-01 2.25924417e-01 6.13871992e-01
-4.50555265e-01 1.64803639e-01 -1.17438905e-01 1.87544182e-01
5.58550239e-01 1.44197583e-01 6.29627764e-01 -9.90917623e-01
-2.36987010e-01 -3.71891081e-01 -5.97054482e-01 -1.29969192e+00
1.86225504e-01 -5.58673501e-01 6.04934953e-02 -1.33566153e+00
-4.58312081e-03 -4.55946505e-01 4.67649288e-02 3.50606054e-01
-2.12334976e-01 7.45867789e-01 2.09917560e-01 1.33605897e-01
-8.44565749e-01 4.22443226e-02 1.10874748e+00 1.65716112e-01
-3.46489340e-01 -1.78999856e-01 -7.50373229e-02 1.01168191e+00
5.18189192e-01 -4.26460594e-01 -1.91489547e-01 -5.97465694e-01
-2.19249099e-01 2.88810462e-01 4.67877626e-01 -1.36782157e+00
2.41775513e-01 9.72075239e-02 3.37363124e-01 -6.37611866e-01
2.91024804e-01 -7.26398110e-01 1.64815217e-01 8.16771150e-01
-1.11581773e-01 -1.11399973e-02 1.10649779e-01 3.95081788e-01
-2.76426703e-01 -1.72880918e-01 1.18537354e+00 -2.44053051e-01
-1.18483341e+00 4.87584323e-01 -4.26850080e-01 5.78299873e-02
1.29154253e+00 -3.76751631e-01 1.30730709e-02 4.17358475e-03
-8.04906845e-01 2.69966722e-01 5.23310959e-01 3.65393966e-01
4.46495384e-01 -9.77644682e-01 -5.13593554e-01 3.00891817e-01
-4.10277247e-01 2.33873844e-01 3.17636877e-01 8.90360832e-01
-1.39893878e+00 2.80421168e-01 -1.85585201e-01 -1.00526822e+00
-1.17086768e+00 4.65901703e-01 4.18409199e-01 -2.41605446e-01
-9.86028910e-01 9.21782732e-01 1.58700123e-01 3.05845112e-01
5.62533690e-03 -4.74979788e-01 -3.63903582e-01 1.48294896e-01
8.52276981e-01 5.84332824e-01 1.42119363e-01 -8.79841685e-01
-3.32174361e-01 7.79341578e-01 3.00126761e-01 -1.55228406e-01
1.22973680e+00 -2.92917639e-01 -1.59875125e-01 9.46112573e-02
1.28812885e+00 -5.07985175e-01 -1.85090959e+00 8.12698305e-02
1.54184438e-02 -6.36369228e-01 3.59029353e-01 -2.22398445e-01
-1.50544512e+00 7.39419818e-01 6.60348773e-01 2.24451974e-01
1.29918289e+00 -2.13918164e-01 1.37559545e+00 -3.44001710e-01
4.33718532e-01 -1.39979327e+00 -9.95081812e-02 5.97198844e-01
2.54099339e-01 -1.03140736e+00 -1.92466870e-01 -6.74390316e-01
-5.52851081e-01 1.49885476e+00 4.29049522e-01 -5.55116832e-01
4.83014971e-01 4.37069029e-01 7.06679421e-04 -5.75339645e-02
-5.16382158e-01 -3.78698319e-01 7.89555609e-02 5.88982403e-01
2.17887551e-01 -2.83513784e-01 -2.67754346e-01 7.07843453e-02
8.71370211e-02 2.61204392e-01 5.70172310e-01 6.13716066e-01
-5.68483233e-01 -8.41028750e-01 -3.19427729e-01 1.56297892e-01
-4.68624026e-01 5.85108623e-02 3.02783966e-01 5.23378670e-01
4.34389651e-01 8.87585044e-01 6.41950846e-01 9.25344136e-03
5.19351754e-03 -1.67808965e-01 3.05050492e-01 -4.35773760e-01
-8.55374932e-01 2.33086094e-01 2.28872709e-02 -1.39758348e+00
-7.65555024e-01 -6.46092415e-01 -1.57144630e+00 -3.15829933e-01
-2.77497582e-02 -4.36312752e-03 7.30118454e-01 1.01529860e+00
3.47694010e-01 5.56080103e-01 4.30470765e-01 -1.39064097e+00
5.00498176e-01 -4.59107518e-01 -3.41283381e-01 4.65329260e-01
2.05327109e-01 -3.41164172e-01 -1.54638171e-01 4.84146655e-01]
|
[9.186528205871582, -0.17200618982315063]
|
ad9a0e05-023a-4762-bb5f-130d5609c54c
|
transferable-graph-backdoor-attack
|
2207.00425
| null |
https://arxiv.org/abs/2207.00425v3
|
https://arxiv.org/pdf/2207.00425v3.pdf
|
Transferable Graph Backdoor Attack
|
Graph Neural Networks (GNNs) have achieved tremendous success in many graph mining tasks benefitting from the message passing strategy that fuses the local structure and node features for better graph representation learning. Despite the success of GNNs, and similar to other types of deep neural networks, GNNs are found to be vulnerable to unnoticeable perturbations on both graph structure and node features. Many adversarial attacks have been proposed to disclose the fragility of GNNs under different perturbation strategies to create adversarial examples. However, vulnerability of GNNs to successful backdoor attacks was only shown recently. In this paper, we disclose the TRAP attack, a Transferable GRAPh backdoor attack. The core attack principle is to poison the training dataset with perturbation-based triggers that can lead to an effective and transferable backdoor attack. The perturbation trigger for a graph is generated by performing the perturbation actions on the graph structure via a gradient based score matrix from a surrogate model. Compared with prior works, TRAP attack is different in several ways: i) it exploits a surrogate Graph Convolutional Network (GCN) model to generate perturbation triggers for a blackbox based backdoor attack; ii) it generates sample-specific perturbation triggers which do not have a fixed pattern; and iii) the attack transfers, for the first time in the context of GNNs, to different GNN models when trained with the forged poisoned training dataset. Through extensive evaluations on four real-world datasets, we demonstrate the effectiveness of the TRAP attack to build transferable backdoors in four different popular GNNs using four real-world datasets.
|
['Salil S. Kanhere', 'Damith C. Ranasinghe', 'Seyit Camtepe', 'Tamas Abraham', 'Olivier De Vel', 'Paul Montague', 'Bao Gia Doan', 'Shuiqiao Yang']
|
2022-06-21
| null | null | null | null |
['graph-mining']
|
['graphs']
|
[ 4.52548563e-01 4.00553912e-01 -1.47836670e-01 1.90697595e-01
-3.33217591e-01 -1.17776418e+00 5.44266701e-01 1.89195022e-01
1.12152502e-01 5.50941885e-01 -5.18299490e-02 -6.28865004e-01
2.50322185e-02 -1.37176788e+00 -1.25230896e+00 -4.83368784e-01
-6.38696015e-01 1.17023043e-01 2.89444715e-01 -6.27695858e-01
1.86109796e-01 6.93266034e-01 -7.51046300e-01 1.21645994e-01
4.70381349e-01 3.11212063e-01 -5.25398910e-01 7.15973854e-01
2.84235686e-01 6.90604568e-01 -9.72193360e-01 -5.97564340e-01
8.13359261e-01 -4.34836447e-01 -7.72160888e-01 -4.65214789e-01
3.09712857e-01 -3.29209119e-01 -1.07353461e+00 1.37391162e+00
5.18022537e-01 -1.17119804e-01 2.29210272e-01 -1.61062884e+00
-7.53400922e-01 1.01341140e+00 -3.59923273e-01 2.11297959e-01
4.85166550e-01 5.26308239e-01 8.76497626e-01 -1.26876637e-01
6.58694267e-01 1.32846749e+00 7.27630019e-01 7.90916920e-01
-1.24802172e+00 -1.07728934e+00 -3.29856984e-02 -1.89296320e-01
-1.20805562e+00 2.76768841e-02 8.67487729e-01 -7.41759390e-02
8.19591165e-01 4.56576139e-01 5.33665419e-01 1.79264581e+00
3.79455507e-01 4.20335531e-01 5.48394382e-01 -9.02630314e-02
1.89431295e-01 -2.29379386e-01 -1.32190555e-01 8.29972327e-01
6.92204714e-01 6.88695669e-01 -2.65449524e-01 -8.10701072e-01
6.80326700e-01 1.31382763e-01 -5.33593595e-01 -2.21651465e-01
-7.09019065e-01 1.08968258e+00 1.03814316e+00 3.77132595e-02
-9.13022235e-02 6.25646770e-01 5.55160880e-01 5.83632648e-01
6.54305294e-02 9.05914307e-01 -2.86680222e-01 3.63803446e-01
-3.08013052e-01 1.54677793e-01 1.12325275e+00 5.27395368e-01
7.99075127e-01 3.08923364e-01 -2.42324874e-01 3.89226899e-02
1.15340382e-01 3.54180574e-01 3.35780531e-01 -1.79323196e-01
5.77687323e-01 7.94365942e-01 -5.34682393e-01 -1.51374543e+00
-2.60770351e-01 -6.19187057e-01 -8.99239123e-01 1.42192796e-01
3.15834612e-01 -3.89294088e-01 -1.04501247e+00 2.05470467e+00
3.86859715e-01 6.18865669e-01 5.79389781e-02 4.68948662e-01
6.99613690e-01 3.93198103e-01 -1.78433001e-01 5.25693357e-01
8.12851071e-01 -5.19042313e-01 -8.61233175e-02 -3.21963996e-01
8.35919023e-01 -4.43448685e-02 1.02491021e+00 1.74228579e-01
-4.06551808e-01 8.64110291e-02 -1.42820907e+00 5.24540901e-01
-7.49472201e-01 -9.23490644e-01 7.51715064e-01 1.13452601e+00
-1.06599212e+00 9.63610888e-01 -6.53439939e-01 -3.23937356e-01
6.57316506e-01 6.08280420e-01 -6.01674736e-01 -1.66154027e-01
-1.72387922e+00 3.77860904e-01 6.76517725e-01 -7.79399425e-02
-1.67522287e+00 -9.12046611e-01 -8.33961427e-01 1.52366892e-01
4.68061656e-01 -5.09390533e-01 6.58382773e-01 -7.88119674e-01
-1.32488990e+00 5.49270034e-01 7.18579412e-01 -9.91489172e-01
4.38061744e-01 2.84410268e-01 -5.02692223e-01 2.99628347e-01
-1.88709319e-01 1.95444822e-01 1.14039278e+00 -1.03819168e+00
1.39584482e-01 -1.28446117e-01 4.39916193e-01 -2.25105956e-01
-5.75042367e-01 -1.13115288e-01 -2.84491554e-02 -8.11143219e-01
-3.61267656e-01 -9.98947620e-01 -4.54484314e-01 -3.33468676e-01
-1.02169263e+00 3.21819901e-01 1.16059220e+00 -2.20915556e-01
1.08579028e+00 -2.00549054e+00 4.35663275e-02 6.10641837e-01
6.01447105e-01 6.76545680e-01 -4.84703958e-01 9.57919657e-01
-3.50344986e-01 6.28775358e-01 -3.91227156e-01 2.57091820e-01
-1.06097370e-01 1.26159355e-01 -8.56485486e-01 6.82204068e-01
9.46413949e-02 1.30143487e+00 -1.17289984e+00 2.26141408e-01
-3.16296704e-02 4.02960211e-01 -6.36436582e-01 3.30180191e-02
-3.83406103e-01 2.17023239e-01 -4.60318148e-01 6.40695095e-01
5.82827389e-01 -1.85597688e-01 2.26672694e-01 8.23922157e-02
8.33470881e-01 1.72968898e-02 -6.11149848e-01 1.09891760e+00
8.83086994e-02 3.42269182e-01 -1.72935680e-01 -8.27358007e-01
8.89717042e-01 2.62207091e-01 1.06260382e-01 -3.13581675e-01
2.08150357e-01 1.23957500e-01 1.44037083e-01 1.09683946e-01
6.26891851e-02 1.43200353e-01 -4.05006737e-01 6.41388237e-01
7.65333697e-02 1.19403437e-01 -3.13006565e-02 6.56897902e-01
2.16846299e+00 -5.06485283e-01 2.19914883e-01 -1.15925232e-02
3.62806082e-01 -1.69838443e-01 4.15405214e-01 1.11938453e+00
-1.12564065e-01 3.32668543e-01 8.23269010e-01 -6.00439489e-01
-5.73339045e-01 -1.30763745e+00 4.29302335e-01 5.52675068e-01
1.24405488e-01 -7.00208843e-01 -8.04602981e-01 -1.34130895e+00
3.11478645e-01 5.18306971e-01 -8.63015294e-01 -1.16301882e+00
-5.18897653e-01 -7.32468009e-01 1.44705796e+00 1.46192566e-01
5.71015835e-01 -1.15698850e+00 -1.93672592e-03 1.29399613e-01
3.40393811e-01 -1.01421869e+00 -5.57482600e-01 -1.43229505e-02
-6.96436226e-01 -1.51904547e+00 -9.30514187e-02 -4.25370216e-01
8.68714809e-01 2.13795677e-01 8.96646559e-01 6.68531001e-01
-4.60651666e-01 4.46726918e-01 -3.95608366e-01 -3.48907471e-01
-9.77153242e-01 2.28743464e-01 1.50442079e-01 -5.72848599e-03
7.72073194e-02 -1.11018336e+00 -3.85258555e-01 9.50547010e-02
-1.35902953e+00 -6.74778342e-01 5.22110045e-01 6.14327669e-01
1.74672142e-01 3.05054814e-01 5.07550299e-01 -1.29357016e+00
1.03969538e+00 -8.72138143e-01 -6.48201346e-01 8.37007761e-02
-4.93402600e-01 3.41704488e-02 1.15732574e+00 -6.99606001e-01
-3.80352810e-02 -1.76186770e-01 4.46773544e-02 -8.40239882e-01
1.86874628e-01 5.56493819e-01 -2.78153628e-01 -7.41894603e-01
1.28667283e+00 2.64486462e-01 -7.46941864e-02 -1.26594290e-01
4.98842597e-01 -1.50091469e-01 4.19448525e-01 -5.90765893e-01
1.69359219e+00 4.06357467e-01 4.90937233e-01 -5.68678081e-01
-3.87893498e-01 2.92995602e-01 3.26617248e-02 6.67076325e-04
2.82894611e-01 -3.93956035e-01 -8.83293748e-01 6.44458175e-01
-9.59393024e-01 -4.73689973e-01 -1.98683813e-01 -6.81178272e-02
-2.06189305e-02 6.44252717e-01 -7.49808609e-01 -3.18855226e-01
-6.87537432e-01 -8.68553042e-01 4.12803978e-01 3.43747474e-02
1.46097124e-01 -1.18903458e+00 1.79634035e-01 -1.91518024e-01
4.57492799e-01 1.01537287e+00 9.91664886e-01 -1.18678904e+00
-8.29472303e-01 -8.20224762e-01 1.12263650e-01 3.11323106e-01
3.75876278e-01 -9.44685712e-02 -7.10355222e-01 -8.62801850e-01
-1.46687627e-01 -3.43768805e-01 6.84015810e-01 -9.18634534e-02
1.08431327e+00 -8.76574874e-01 -5.49641669e-01 1.12669337e+00
1.54539561e+00 -9.15736556e-02 8.15281451e-01 2.65191823e-01
1.09310138e+00 2.37568542e-02 -2.20248297e-01 3.23300697e-02
-1.76055655e-01 2.68815845e-01 1.14636123e+00 1.46032106e-02
-1.08368164e-02 -9.55546498e-01 6.55391753e-01 1.68820724e-01
5.43159783e-01 -6.31160200e-01 -9.19458807e-01 2.53456056e-01
-1.51058888e+00 -9.17561412e-01 2.27755859e-01 2.19150281e+00
6.02453172e-01 3.91253024e-01 1.31442502e-01 1.10400077e-02
8.06115687e-01 3.58903199e-01 -7.63621926e-01 -4.69686389e-01
-1.42511547e-01 5.11711001e-01 1.00993454e+00 3.76393795e-01
-9.16811764e-01 1.12153351e+00 6.18429422e+00 6.97189450e-01
-1.22042048e+00 -2.19898716e-01 3.27232093e-01 -2.34915502e-02
-4.06099796e-01 3.46241593e-01 -5.32124281e-01 4.46892828e-01
8.84701014e-01 -5.50085008e-01 9.38194513e-01 7.28650212e-01
-4.08401787e-01 7.16119587e-01 -1.08544183e+00 4.62741822e-01
-1.44464388e-01 -1.57121694e+00 5.77540994e-01 4.20295954e-01
5.23100078e-01 2.71205157e-01 2.63210744e-01 4.24073458e-01
9.56555188e-01 -1.42466199e+00 2.20537275e-01 5.49357049e-02
7.01699913e-01 -1.02223778e+00 2.79009432e-01 2.87098408e-01
-1.05373335e+00 -2.01201558e-01 -3.02914351e-01 1.93113647e-02
-2.27542147e-01 5.40860713e-01 -1.26194751e+00 7.56054521e-01
4.34001893e-01 4.82820302e-01 -6.26048028e-01 6.23175561e-01
-5.93607724e-01 1.05480635e+00 -4.10498828e-01 4.39984314e-02
3.78923118e-01 -7.87357762e-02 1.15851796e+00 7.91926742e-01
3.95710170e-02 -3.72646540e-01 4.71774861e-02 1.17763853e+00
-6.32746458e-01 -2.06980020e-01 -1.33567870e+00 -5.44044435e-01
6.90132499e-01 1.26375473e+00 -5.57407379e-01 2.78053135e-01
1.55865908e-01 9.34828401e-01 5.16436934e-01 4.98110384e-01
-7.84843862e-01 -7.12434113e-01 7.58560061e-01 2.09878340e-01
3.63615513e-01 2.33277958e-02 3.00354242e-01 -9.38458741e-01
-1.17648914e-01 -1.24881136e+00 6.41388774e-01 -2.04068109e-01
-1.49997580e+00 7.18569100e-01 -1.38262570e-01 -9.77301180e-01
-9.33729932e-02 -4.87917989e-01 -1.18108249e+00 7.87377596e-01
-1.06864429e+00 -1.11351800e+00 6.90203533e-02 9.95351851e-01
-3.97691488e-01 -4.32373792e-01 9.46703434e-01 3.63350250e-02
-7.73780227e-01 1.15692937e+00 -2.77349174e-01 7.16522336e-01
3.38787585e-01 -1.01898313e+00 1.15759778e+00 1.33199656e+00
2.47592956e-01 8.73780429e-01 7.35008001e-01 -1.04201496e+00
-1.64880347e+00 -1.49324477e+00 7.85043463e-03 -4.77862895e-01
8.84654284e-01 -6.84687376e-01 -8.89799237e-01 9.89282250e-01
-2.03332126e-01 5.08825243e-01 4.72423822e-01 -2.82660156e-01
-9.16483641e-01 -1.40557736e-02 -1.62503469e+00 8.73933315e-01
1.24122965e+00 -6.55737579e-01 -4.64304052e-02 5.31943440e-01
1.26401818e+00 -5.20580053e-01 -5.23576200e-01 3.58715743e-01
8.02712068e-02 -7.41301000e-01 1.06368577e+00 -1.05273628e+00
1.77907273e-01 -2.88963079e-01 -2.20970307e-02 -1.44562638e+00
-1.92486882e-01 -1.29291117e+00 -6.60459042e-01 9.55478668e-01
3.74302924e-01 -1.21217072e+00 9.90432084e-01 1.31740138e-01
2.38410383e-01 -7.70889401e-01 -8.98878992e-01 -1.00271451e+00
9.03821290e-02 -1.60133407e-01 1.19934499e+00 1.10777164e+00
-2.12539151e-01 1.58382490e-01 -3.83552879e-01 6.16799116e-01
7.78973997e-01 -4.55725968e-01 9.83757854e-01 -1.02354825e+00
-4.94012952e-01 -2.47054964e-01 -9.18855846e-01 -2.32018799e-01
2.77273715e-01 -1.46126640e+00 -4.31319088e-01 -9.72837448e-01
-2.26110995e-01 -1.16600513e-01 -4.01945889e-01 7.16361701e-01
-2.14818642e-01 1.10386342e-01 1.42337590e-01 -1.06406882e-01
-3.35091054e-02 3.42901289e-01 9.04898226e-01 -3.62113714e-01
-1.50973126e-01 1.14574797e-01 -1.10492611e+00 3.37897509e-01
9.01560545e-01 -9.24654365e-01 -9.04523194e-01 1.82463042e-02
5.97536623e-01 -1.78133324e-01 6.74510777e-01 -1.02347910e+00
4.74110432e-02 1.00447036e-01 1.84362549e-02 -1.07147142e-01
-2.56578289e-02 -5.67841172e-01 3.07158977e-01 9.26782250e-01
5.00692688e-02 -1.32262777e-03 3.47304106e-01 1.12434733e+00
2.76485533e-01 1.39910638e-01 6.42493188e-01 -1.73277110e-01
-2.78804600e-01 9.34901834e-01 -1.71457957e-02 2.68775880e-01
1.08907306e+00 -2.74691761e-01 -7.19715178e-01 -4.81128663e-01
-2.54359066e-01 8.40531811e-02 6.84086502e-01 4.38948721e-01
6.82969093e-01 -1.27043509e+00 -6.52696431e-01 4.15303379e-01
3.11805531e-02 -1.61595136e-01 -2.90245600e-02 2.63312489e-01
-5.84268570e-01 -2.71757487e-02 -1.79056481e-01 -1.84939653e-01
-9.44070101e-01 9.84226227e-01 7.19657838e-01 -5.63458085e-01
-8.19030404e-01 9.39208984e-01 1.60370007e-01 -7.32215464e-01
7.83433542e-02 1.24533102e-01 3.01530629e-01 -6.68610930e-01
3.55137348e-01 1.54300839e-01 7.73256421e-02 -2.87560463e-01
-4.53749478e-01 6.02164716e-02 -2.33379290e-01 1.81499824e-01
1.07636809e+00 6.44926488e-01 -1.95607722e-01 -2.71661431e-01
1.40901852e+00 6.43259436e-02 -9.64568615e-01 -2.00005710e-01
-2.25975513e-01 -4.92631286e-01 -2.37601548e-01 -7.08449006e-01
-1.38510180e+00 4.22661245e-01 1.89449236e-01 4.71358508e-01
9.78185833e-01 -3.04877907e-01 1.04486632e+00 6.09105170e-01
4.43837553e-01 -2.77563244e-01 2.37796947e-01 3.93421799e-01
7.95767248e-01 -6.67550802e-01 -6.32018894e-02 -3.25170279e-01
-1.15930878e-01 8.86707008e-01 6.16846263e-01 -5.35431802e-01
7.56171107e-01 1.38842404e-01 -1.89538807e-01 -5.24898767e-01
-6.60272241e-01 5.22232771e-01 3.42390202e-02 8.86294365e-01
-3.99301767e-01 3.54521051e-02 1.99601814e-01 5.60286641e-01
-2.65936643e-01 -2.02378079e-01 8.00737560e-01 8.24714243e-01
-1.23842455e-01 -1.30944633e+00 -3.21551621e-01 5.69156826e-01
-5.14618278e-01 -1.43518537e-01 -9.51598942e-01 7.85345435e-01
-3.78459066e-01 7.89502442e-01 -7.09767997e-01 -1.03692079e+00
1.93357721e-01 -3.30651820e-01 3.28126490e-01 -7.78475583e-01
-1.05151665e+00 -9.85547721e-01 -1.62491232e-01 -8.76547515e-01
4.58880752e-01 -2.09582690e-02 -1.13574576e+00 -7.23215163e-01
-2.24190727e-01 1.12009607e-01 3.00301522e-01 5.87841868e-01
6.90189004e-01 4.36832607e-01 9.83464003e-01 -5.82011878e-01
-8.44767332e-01 -5.33124685e-01 -6.86510324e-01 5.33323467e-01
3.10545743e-01 -4.80093628e-01 -9.04254138e-01 -8.17696333e-01]
|
[6.107460975646973, 7.3392181396484375]
|
1a4cbd10-b166-4502-b17f-d3659f195842
|
deepore-a-deep-learning-workflow-for-rapid
|
2005.03759
| null |
https://arxiv.org/abs/2005.03759v2
|
https://arxiv.org/pdf/2005.03759v2.pdf
|
DeePore: a deep learning workflow for rapid and comprehensive characterization of porous materials
|
DeePore is a deep learning workflow for rapid estimation of a wide range of porous material properties based on the binarized micro-tomography images. By combining naturally occurring porous textures we generated 17700 semi-real 3-D micro-structures of porous geo-materials with size of 256^3 voxels and 30 physical properties of each sample are calculated using physical simulations on the corresponding pore network models. Next, a designed feed-forward convolutional neural network (CNN) is trained based on the dataset to estimate several morphological, hydraulic, electrical, and mechanical characteristics of the porous material in a fraction of a second. In order to fine-tune the CNN design, we tested 9 different training scenarios and selected the one with the highest average coefficient of determination (R^2) equal to 0.885 for 1418 testing samples. Additionally, 3 independent synthetic images as well as 3 realistic tomography images have been tested using the proposed method and results are compared with pore network modelling and experimental data, respectively. Tested absolute permeabilities had around 13 % relative error compared to the experimental data which is noticeable considering the accuracy of the direct numerical simulation methods such as Lattice Boltzmann and Finite Volume. The workflow is compatible with any physical size of the images due to its dimensionless approach and can be used to characterize large-scale 3-D images by averaging the model outputs for a sliding window that scans the whole geometry.
|
['Traiwit Chung', 'Ying Da Wang', 'Reza Shams', 'Arash Rabbani', 'Masoud Babaei']
|
2020-05-03
| null | null | null | null |
['physical-simulations']
|
['miscellaneous']
|
[-4.08500545e-02 2.40526304e-01 8.18049312e-01 1.77069604e-02
-5.10069370e-01 1.74369410e-01 5.52367687e-01 4.75579947e-01
-8.42993140e-01 1.19404185e+00 -3.96288261e-02 -2.67383456e-01
-4.23281074e-01 -1.46260381e+00 -1.10701466e+00 -1.11103594e+00
-6.17574692e-01 1.07779670e+00 4.57865566e-01 -4.70595770e-02
3.02960664e-01 9.44429457e-01 -1.60900664e+00 2.14039356e-01
6.72467411e-01 1.29589140e+00 2.31486708e-01 6.80795431e-01
1.20509587e-01 4.69705641e-01 -1.47073403e-01 1.94761664e-01
8.74455720e-02 1.28175929e-01 -8.89928937e-01 -1.84244305e-01
8.70048255e-02 -4.65187609e-01 -4.25508380e-01 4.98696685e-01
6.91118896e-01 4.64613736e-01 1.22121108e+00 -1.92139253e-01
-3.90523434e-01 4.71564591e-01 -6.31376356e-02 1.73174694e-01
3.22597031e-03 3.00808400e-01 3.16188753e-01 -9.89376366e-01
5.48770785e-01 1.15015709e+00 6.51592731e-01 1.01586789e-01
-1.22371304e+00 -2.75198251e-01 -8.27885330e-01 -4.77664694e-02
-1.35621834e+00 -1.43685281e-01 2.39734709e-01 -6.90695107e-01
1.19504392e+00 2.16746643e-01 1.08506513e+00 9.50892627e-01
8.31587315e-01 -1.61180109e-01 1.53704119e+00 -2.22424313e-01
6.84196591e-01 -5.35418801e-02 -4.53284174e-01 6.39436543e-01
3.47124070e-01 2.34261215e-01 1.42964497e-01 -1.12478897e-01
1.38836980e+00 -4.06302035e-01 -2.88141593e-02 -8.78360122e-02
-1.12527668e+00 6.86371326e-01 6.42234504e-01 2.14851618e-01
-6.79320097e-01 2.71091014e-01 5.26466668e-01 -7.04300180e-02
3.67589474e-01 6.45287454e-01 -5.00534952e-01 -7.70736411e-02
-8.65635276e-01 6.34677410e-01 1.08856905e+00 2.66445488e-01
7.08427966e-01 3.65138650e-01 7.72138610e-02 6.57969356e-01
3.75800312e-01 9.27683830e-01 3.54039967e-01 -9.13771868e-01
1.73589159e-02 1.18673049e-01 1.91625267e-01 -7.45001376e-01
-3.84079307e-01 -1.02660149e-01 -1.12811720e+00 3.13017577e-01
6.24620795e-01 4.15827073e-02 -1.06883454e+00 7.64429331e-01
3.97147685e-01 -7.70573877e-03 1.45092830e-01 7.43994832e-01
1.11463642e+00 8.34733069e-01 8.22362229e-02 3.92197967e-02
1.01829112e+00 -5.45094192e-01 8.03813413e-02 3.60977888e-01
1.73795342e-01 -2.90985733e-01 1.06256258e+00 2.58972794e-01
-1.10431588e+00 -2.87220448e-01 -9.69121456e-01 3.47478747e-01
-5.43425381e-01 -2.61393368e-01 4.67094094e-01 2.97260672e-01
-7.44388759e-01 1.25378633e+00 -9.83961046e-01 -2.57681906e-01
5.53588569e-01 3.21229786e-01 -4.81966794e-01 1.40581071e-01
-1.42014658e+00 7.83940971e-01 4.66244936e-01 3.17929447e-01
-1.35599828e+00 -7.73116708e-01 -5.90848863e-01 2.39167601e-01
-3.24411869e-01 -7.10473061e-01 7.25861788e-01 -1.35080203e-01
-1.76276004e+00 7.12308586e-01 3.39536607e-01 -5.22454619e-01
9.99823093e-01 -2.55016163e-02 -1.34466603e-01 3.66268814e-01
-7.63753382e-03 4.03609782e-01 4.63615686e-01 -1.34695733e+00
1.52993232e-01 1.13958873e-01 -1.43994734e-01 -5.00485301e-02
6.58457652e-02 -3.14248592e-01 -1.38222173e-01 -8.51014331e-02
7.76401684e-02 -5.63317776e-01 -6.06087804e-01 3.91048612e-03
-5.22279382e-01 4.81601268e-01 3.26677054e-01 -8.87707055e-01
4.07061070e-01 -1.80747283e+00 7.38386735e-02 6.04160428e-01
2.55723149e-01 -4.72694337e-02 3.35333019e-01 4.96832371e-01
1.63447887e-01 5.54078445e-02 -7.08996654e-01 9.99037325e-02
-2.56507814e-01 4.10049446e-02 9.24362428e-03 6.94687843e-01
2.18551839e-03 7.30276167e-01 -6.04895592e-01 -5.64031959e-01
6.55664802e-01 7.58059204e-01 -3.25493395e-01 2.08902493e-01
-2.74979413e-01 6.98858082e-01 -4.30748522e-01 5.71840823e-01
1.04519343e+00 -2.53037095e-01 -2.34201372e-01 7.62230856e-03
-2.63426989e-01 -6.75035194e-02 -9.51604486e-01 9.94711101e-01
-6.72945201e-01 2.62622058e-01 1.67713225e-01 -1.03604174e+00
1.11512637e+00 3.22820932e-01 5.09701312e-01 -8.61896932e-01
3.41024548e-01 6.27046525e-01 1.40605688e-01 -6.46163821e-01
2.83913493e-01 -5.17918706e-01 3.23363394e-01 1.75512880e-01
-1.17233552e-01 -7.90006280e-01 4.87996303e-02 -2.12925032e-01
9.16808128e-01 -9.96040925e-02 -1.17999092e-01 -7.80865788e-01
6.08521402e-01 -5.53922392e-02 -2.00588822e-01 8.89775455e-01
2.02903822e-01 7.08381951e-01 3.46339226e-01 -6.85175240e-01
-1.82008362e+00 -1.31823325e+00 -7.66397178e-01 1.74101502e-01
8.71111304e-02 3.03663790e-01 -7.53409684e-01 2.64292240e-01
3.59149784e-01 1.35243297e-01 -1.04303133e+00 1.61398977e-01
-4.35102642e-01 -1.03119719e+00 5.89352250e-01 4.53385711e-01
7.84923851e-01 -1.50488460e+00 -7.02302754e-01 5.39588690e-01
3.75273883e-01 -1.01247334e+00 9.31371510e-01 2.49878749e-01
-1.09648073e+00 -1.09063637e+00 -7.18139529e-01 -2.33660266e-01
4.56146866e-01 -6.82272673e-01 8.97729039e-01 9.54139605e-02
-2.69784540e-01 -6.96731657e-02 -6.78503215e-02 -1.05566695e-01
-7.70616293e-01 4.06479947e-02 -1.38081107e-02 -3.97366107e-01
-3.84684205e-01 -8.18746269e-01 -7.37184823e-01 1.72018021e-01
-8.93385530e-01 2.18551293e-01 4.75216448e-01 7.94852018e-01
8.93121839e-01 4.01463091e-01 1.68586105e-01 -6.38369858e-01
4.94076550e-01 -6.34774029e-01 -5.37163556e-01 -1.10946774e-01
-2.79141873e-01 -1.32547766e-01 8.48513901e-01 -3.72878969e-01
-1.11034262e+00 -4.13869649e-01 -4.09091800e-01 -1.29849032e-01
-3.91707063e-01 7.94410288e-01 2.11605102e-01 -2.29070991e-01
7.65392542e-01 2.58445740e-01 -6.04930259e-02 -3.03736418e-01
-1.02020457e-01 5.10882735e-01 3.53872776e-01 -1.21506560e+00
5.31277478e-01 7.45621681e-01 3.07337254e-01 -1.39801216e+00
-2.09850729e-01 1.30289704e-01 -5.81885636e-01 -5.04973471e-01
7.97612309e-01 -7.57611811e-01 -9.61868942e-01 8.63914490e-01
-8.07841480e-01 -8.43391895e-01 -3.30788970e-01 7.01002419e-01
-6.72054410e-01 2.72540003e-01 -1.03880644e+00 -7.37028480e-01
-6.96005583e-01 -1.37109673e+00 8.32653582e-01 2.07832590e-01
-8.90047231e-04 -1.11678898e+00 -3.37723154e-03 1.82343110e-01
9.35388088e-01 7.61571705e-01 1.01737916e+00 -1.61885291e-01
-3.02184880e-01 -9.82655436e-02 -4.54399168e-01 2.46317729e-01
-1.65660396e-01 3.86314094e-01 -1.21320629e+00 -1.24186791e-01
-1.23242736e-01 -4.16305184e-01 1.04239321e+00 8.67966473e-01
1.30685484e+00 1.80846155e-02 -1.41509563e-01 5.68947375e-01
1.75139439e+00 -4.72821034e-02 1.01340592e+00 6.75829232e-01
5.93408942e-01 2.52433628e-01 1.81279212e-01 7.59151995e-01
-2.57447481e-01 2.18650386e-01 6.67049706e-01 -8.47052559e-02
1.96203589e-01 8.93646702e-02 -9.24462751e-02 5.88840783e-01
-7.30302334e-01 -6.59148172e-02 -1.30794740e+00 3.84437829e-01
-1.15544951e+00 -7.52480626e-01 -3.85142595e-01 2.23515630e+00
5.93414843e-01 5.02011955e-01 -3.53161871e-01 2.23624930e-01
5.67596138e-01 -9.07462612e-02 -4.91403520e-01 -5.34372687e-01
-2.82707363e-01 8.43971968e-01 8.21431398e-01 4.77458090e-01
-9.73687947e-01 7.06152439e-01 6.38275003e+00 6.61381066e-01
-1.66381776e+00 -1.00911029e-01 7.13535547e-01 5.40614687e-02
-1.77032039e-01 -2.03049302e-01 -1.91989690e-01 5.39831817e-01
1.32550836e+00 2.91286502e-02 5.62726259e-01 5.51102102e-01
4.55315799e-01 -6.67740643e-01 -5.38784683e-01 4.88698900e-01
-7.16186941e-01 -1.72341037e+00 1.39368579e-01 3.53349090e-01
6.94357216e-01 7.45088875e-01 -2.89518833e-01 1.10613771e-01
2.03161940e-01 -1.32383215e+00 8.45278859e-01 1.05662596e+00
1.13816226e+00 -3.40133250e-01 9.21336889e-01 8.58272016e-02
-1.03862131e+00 2.46777125e-02 -6.89590394e-01 -2.05535382e-01
2.61970848e-01 9.84124660e-01 -7.61714101e-01 6.66185439e-01
1.20017922e+00 2.51140475e-01 -4.21899080e-01 1.13743782e+00
3.54321897e-01 7.09132433e-01 -9.03978229e-01 -5.60051985e-02
3.57809454e-01 -4.03131753e-01 2.11793110e-01 1.16241860e+00
5.39735854e-01 -8.57501104e-02 -5.02579153e-01 1.19510233e+00
1.86159059e-01 2.46561766e-01 -5.64261198e-01 -5.92271462e-02
4.77596402e-01 1.07242215e+00 -1.13876915e+00 -4.59059030e-01
8.60210583e-02 2.34043851e-01 1.42900780e-01 1.85146108e-01
-8.92566442e-01 -2.35673517e-01 1.18923284e-01 5.50143003e-01
3.11755061e-01 -3.01916778e-01 -3.00285935e-01 -7.01308310e-01
-2.76811749e-01 -1.76278710e-01 -1.59974694e-01 -1.06105995e+00
-1.17224944e+00 4.98409718e-01 1.65626541e-01 -5.57701349e-01
2.85402894e-01 -1.03875172e+00 -8.05772364e-01 1.08135223e+00
-1.17067373e+00 -8.28753531e-01 -5.94616234e-01 2.62067318e-01
-6.02901308e-03 8.10828730e-02 1.06437993e+00 1.57174081e-01
-2.89018899e-01 -1.72004834e-01 7.79387176e-01 -2.20348556e-02
1.67722508e-01 -1.13492262e+00 2.82467544e-01 2.22449884e-01
-8.26952338e-01 2.79101461e-01 9.72179532e-01 -6.55205965e-01
-1.10637391e+00 -9.80542302e-01 2.71454573e-01 -3.77843790e-02
6.86026514e-01 -1.59163907e-01 -1.35617805e+00 1.06187396e-01
-2.59673625e-01 4.33967710e-01 2.05182433e-01 -5.45691013e-01
2.01545313e-01 1.45055905e-01 -1.47652960e+00 1.90364033e-01
3.69815469e-01 -3.09968412e-01 -1.94400296e-01 2.66008556e-01
3.59081388e-01 -7.85043001e-01 -1.65299976e+00 6.80336297e-01
7.78557718e-01 -1.20622969e+00 8.90731692e-01 -3.13393265e-01
1.00992811e+00 5.54042198e-02 -2.21804857e-01 -8.91217589e-01
-4.01316356e-04 2.40619108e-01 1.07079953e-01 5.95158398e-01
4.18388426e-01 -7.41235554e-01 8.03246975e-01 5.38580775e-01
-1.56445354e-01 -1.29158199e+00 -1.24408019e+00 -4.73016292e-01
7.15040386e-01 -3.55357349e-01 7.35508740e-01 7.27610469e-01
-4.20054108e-01 -3.82841945e-01 1.11334249e-01 1.76923513e-01
5.84395885e-01 -2.00823426e-01 5.01795292e-01 -1.52832925e+00
-6.51919022e-02 -2.43505999e-01 -3.18215013e-01 -3.49707991e-01
6.14282936e-02 -5.54579198e-01 -4.35425229e-02 -1.62711203e+00
9.16887447e-02 -8.71424437e-01 -2.03866929e-01 7.09096342e-02
6.54243171e-01 1.40307829e-01 -5.70699990e-01 2.14399531e-01
1.21810041e-01 7.48062253e-01 1.68199003e+00 -2.51417607e-01
-3.17556038e-02 -3.32819462e-01 1.14314258e-01 3.47180665e-01
7.08975315e-01 -2.35766813e-01 1.40438214e-01 -1.60178244e-01
2.82172084e-01 4.54332352e-01 6.24714673e-01 -1.51223803e+00
-3.39486003e-01 -6.79261759e-02 7.46847212e-01 -3.62539053e-01
3.23715150e-01 -6.73950255e-01 5.57173729e-01 6.04605317e-01
-1.37488037e-01 -5.04046082e-01 4.25063044e-01 3.64401162e-01
-1.32267803e-01 -7.85594583e-02 1.15674829e+00 -5.62490821e-01
-2.93056369e-01 5.03787041e-01 -6.09265983e-01 -3.02616268e-01
7.12126970e-01 -5.12223840e-01 -1.11266486e-01 1.22630395e-01
-9.65442657e-01 -3.10033202e-01 6.48512542e-01 -3.61279994e-01
6.25573933e-01 -8.83230388e-01 -6.91590428e-01 2.26557910e-01
-2.51699120e-01 7.62231171e-01 8.19598615e-01 7.67261505e-01
-1.65685332e+00 1.91637099e-01 -6.37097597e-01 -5.87155879e-01
-3.63871992e-01 7.26623088e-02 9.70005453e-01 -2.85387844e-01
-8.83864403e-01 7.25695312e-01 -2.45648682e-01 -6.36461914e-01
-5.11532307e-01 -4.90966022e-01 -2.38834321e-01 -3.70068312e-01
-8.55535120e-02 5.69450915e-01 4.30492967e-01 -6.73191607e-01
-7.57208914e-02 3.81795287e-01 2.17570916e-01 -1.05851345e-01
1.95571613e+00 2.66812474e-01 -3.56446475e-01 4.90711063e-01
1.02190185e+00 -3.58465075e-01 -1.51652324e+00 5.59247024e-02
-5.77333748e-01 -1.52152002e-01 1.13412462e-01 -5.17615199e-01
-1.12590349e+00 9.22258437e-01 5.31554401e-01 1.47345677e-01
3.71320516e-01 -1.12268448e-01 5.32244802e-01 3.86354297e-01
3.63293588e-01 -9.81872857e-01 -1.88850909e-01 8.43413234e-01
1.04336524e+00 -9.82460439e-01 3.16405892e-01 -2.96420127e-01
2.65683187e-03 1.46828830e+00 6.48248374e-01 -4.15317386e-01
9.97921765e-01 2.74681330e-01 -2.51287043e-01 -6.70233786e-01
-3.20866734e-01 5.60903013e-01 -3.37242991e-01 4.01434511e-01
3.54408890e-01 1.69072673e-01 -7.46010477e-03 1.51301846e-01
-3.82542104e-01 1.61612257e-01 6.34335637e-01 8.35604966e-01
-3.50587636e-01 -2.71084458e-01 -4.89771754e-01 8.04028332e-01
-2.07296267e-01 -9.43477899e-02 3.84231955e-01 8.91023934e-01
-7.44217187e-02 1.54248670e-01 4.04804945e-01 -4.68805656e-02
1.21057473e-01 -3.11371833e-01 6.02143347e-01 -3.59180897e-01
-1.97550252e-01 -2.28040934e-01 1.55982569e-01 -2.52881855e-01
-3.80150855e-01 -4.43862021e-01 -1.52252197e+00 -5.54134488e-01
-2.20183834e-01 3.31039697e-01 6.88474357e-01 1.09068120e+00
-1.69906288e-01 5.57782769e-01 3.12209547e-01 -1.67994690e+00
-1.76879555e-01 -1.43808174e+00 -1.28349710e+00 2.00269863e-01
-1.32475615e-01 -1.12374723e+00 -6.00903511e-01 -1.84333369e-01]
|
[6.412940979003906, 3.3438875675201416]
|
d200b7a9-b718-4541-921c-8911fa895a08
|
unsupervised-lifelong-person-re
|
2203.06468
| null |
https://arxiv.org/abs/2203.06468v1
|
https://arxiv.org/pdf/2203.06468v1.pdf
|
Unsupervised Lifelong Person Re-identification via Contrastive Rehearsal
|
Existing unsupervised person re-identification (ReID) methods focus on adapting a model trained on a source domain to a fixed target domain. However, an adapted ReID model usually only works well on a certain target domain, but can hardly memorize the source domain knowledge and generalize to upcoming unseen data. In this paper, we propose unsupervised lifelong person ReID, which focuses on continuously conducting unsupervised domain adaptation on new domains without forgetting the knowledge learnt from old domains. To tackle unsupervised lifelong ReID, we conduct a contrastive rehearsal on a small number of stored old samples while sequentially adapting to new domains. We further set an image-to-image similarity constraint between old and new models to regularize the model updates in a way that suits old knowledge. We sequentially train our model on several large-scale datasets in an unsupervised manner and test it on all seen domains as well as several unseen domains to validate the generalizability of our method. Our proposed unsupervised lifelong method achieves strong generalizability, which significantly outperforms previous lifelong methods on both seen and unseen domains. Code will be made available at https://github.com/chenhao2345/UCR.
|
['Francois Bremond', 'Benoit Lagadec', 'Hao Chen']
|
2022-03-12
| null | null | null | null |
['unsupervised-person-re-identification']
|
['computer-vision']
|
[ 3.23213823e-02 -3.46877426e-01 2.95563471e-02 -6.80509806e-01
-4.04512048e-01 -6.21919990e-01 6.96862280e-01 1.28571481e-01
-9.28608239e-01 9.01259124e-01 2.60307997e-01 5.02767026e-01
5.70347868e-02 -6.52373314e-01 -6.52070105e-01 -5.13938546e-01
1.85690522e-01 9.52025414e-01 2.99748063e-01 -7.80375898e-02
-1.66844036e-02 2.50644445e-01 -1.47738135e+00 -1.10176988e-01
9.29300249e-01 2.24235907e-01 2.61768818e-01 4.74647045e-01
1.69514507e-01 1.75613001e-01 -6.79549515e-01 -6.90372288e-01
3.10051173e-01 -4.06077921e-01 -8.98942053e-01 1.55669883e-01
6.66846573e-01 -4.64445025e-01 -6.99137926e-01 1.01402879e+00
7.01670885e-01 6.05441570e-01 5.82037747e-01 -1.03210557e+00
-1.07988751e+00 3.48249048e-01 -5.01642168e-01 5.44373155e-01
5.15392959e-01 1.94191292e-01 3.07262301e-01 -9.43017900e-01
6.72096848e-01 1.03276014e+00 6.63745105e-01 1.16524518e+00
-1.25825071e+00 -1.00513470e+00 6.32784128e-01 3.44016403e-01
-1.58471882e+00 -7.28505909e-01 6.87943697e-01 -1.33596897e-01
6.62832737e-01 -2.21689045e-01 5.55526435e-01 1.51447380e+00
-2.16261134e-01 7.24325776e-01 9.97228861e-01 -3.04343373e-01
3.52718145e-01 3.29783469e-01 4.75979298e-01 2.03665644e-01
5.04759431e-01 1.96638092e-01 -7.57080674e-01 -5.95364794e-02
4.51356113e-01 1.83861405e-01 -4.78115007e-02 -2.86388487e-01
-1.19124818e+00 3.97814989e-01 2.57553577e-01 1.67068020e-01
-1.77956253e-01 -2.56321192e-01 3.22180301e-01 4.98227417e-01
4.93710220e-01 2.12704122e-01 -3.67006838e-01 1.80758964e-02
-6.89932823e-01 2.70123154e-01 6.04469180e-01 1.16327131e+00
1.09226024e+00 -2.21086025e-01 5.52680157e-02 1.14973938e+00
-8.10082257e-02 7.08354473e-01 1.01322007e+00 -5.56691408e-01
1.65645733e-01 3.53607088e-01 1.78015858e-01 -5.47192991e-01
-3.04859638e-01 -5.30665994e-01 -8.33679676e-01 -2.65480220e-01
4.12708104e-01 -2.19887435e-01 -1.10572612e+00 2.03441286e+00
5.30356407e-01 4.83503342e-01 2.35628158e-01 6.80990100e-01
8.73184443e-01 4.91177648e-01 3.77666622e-01 -1.08215839e-01
1.06639504e+00 -9.86666441e-01 -4.19605732e-01 -7.43765891e-01
4.96659964e-01 -2.65423089e-01 1.00844264e+00 3.02163750e-01
-7.99242675e-01 -9.32766438e-01 -9.17700052e-01 -6.40201122e-02
-4.86344755e-01 -1.01638250e-01 1.67148188e-01 5.19491613e-01
-1.08049238e+00 3.00496340e-01 -5.83537817e-01 -9.34028983e-01
2.73622245e-01 5.42939842e-01 -7.22784162e-01 -4.10576195e-01
-1.26881707e+00 6.83650315e-01 7.51166344e-01 -9.13177729e-02
-1.02118456e+00 -4.72575098e-01 -8.85738373e-01 -2.55553722e-01
2.62609303e-01 -8.37141275e-01 1.30888450e+00 -1.02686155e+00
-1.28599906e+00 1.25691950e+00 -5.49408555e-01 -5.38388848e-01
4.79968458e-01 -4.56879318e-01 -8.13630879e-01 -5.06066009e-02
4.05059576e-01 8.97352457e-01 1.00398517e+00 -1.31986248e+00
-6.73797727e-01 -6.27891362e-01 -9.11845639e-02 4.11753327e-01
-8.00849497e-01 -3.20155829e-01 -6.51991129e-01 -8.32526147e-01
-1.06928863e-01 -1.04821503e+00 4.64896299e-02 -3.51312876e-01
-6.27310127e-02 -2.47531056e-01 6.62235916e-01 -5.32683074e-01
1.08398366e+00 -2.25456548e+00 2.87406310e-03 4.27998006e-02
1.23099826e-01 4.32788223e-01 -4.49513793e-01 2.63451546e-01
-1.26950964e-01 -3.29092443e-01 -3.27064604e-01 -6.46919131e-01
-2.40392536e-01 2.96307057e-01 -8.67145583e-02 4.61009383e-01
-1.93326205e-01 7.99171686e-01 -1.05903447e+00 -3.95698398e-01
-8.34059417e-02 2.21759692e-01 -4.04358119e-01 4.15458173e-01
1.76294625e-01 8.38813007e-01 -2.54359096e-01 2.73689598e-01
9.38026011e-01 -1.88139617e-01 1.51024222e-01 2.30498195e-01
2.19640642e-01 -1.00211382e-01 -1.16162121e+00 1.89366913e+00
-2.34427184e-01 3.19858342e-01 -3.59011710e-01 -1.01714289e+00
1.07866955e+00 2.90135611e-02 1.90289333e-01 -8.96049023e-01
-1.23688415e-01 -4.42704968e-02 -3.15119565e-01 -2.35938340e-01
5.06332457e-01 -3.01974297e-01 -1.96360558e-01 5.74360132e-01
2.60672301e-01 4.45691049e-01 2.16723666e-01 2.60906607e-01
8.54351223e-01 -3.47170867e-02 3.16529125e-01 -1.57455906e-01
7.21315086e-01 4.99306954e-02 8.41316462e-01 1.10586047e+00
-5.14271379e-01 6.65853024e-01 -3.63275588e-01 -6.25775635e-01
-1.04270422e+00 -1.37212467e+00 -9.56956223e-02 1.39713204e+00
5.68941653e-01 -2.26727545e-01 -7.32185721e-01 -1.00186992e+00
1.88703015e-01 7.39216149e-01 -7.03028858e-01 -5.04526377e-01
-6.66256905e-01 -7.41695583e-01 4.57934320e-01 5.22745728e-01
9.41600204e-01 -9.88120079e-01 -4.56528030e-02 2.16449097e-01
-1.24988049e-01 -1.06091154e+00 -7.96016037e-01 -2.24875703e-01
-7.03290820e-01 -8.40573609e-01 -1.18379450e+00 -1.18085361e+00
1.05845678e+00 5.31372428e-01 9.17129219e-01 3.61834932e-03
3.65458727e-02 7.67428279e-01 -4.56922829e-01 -3.40764642e-01
-4.19031709e-01 3.09322208e-01 7.73820281e-01 1.47616833e-01
9.67082918e-01 -6.00838959e-01 -5.46167612e-01 5.30679524e-01
-8.26902807e-01 -2.33839564e-02 4.50912207e-01 9.21194315e-01
5.50410151e-01 1.62794039e-01 9.52583790e-01 -1.13281441e+00
5.86633027e-01 -6.22143686e-01 -3.44300479e-01 3.56182069e-01
-7.33130932e-01 5.92663176e-02 6.38824701e-01 -9.18998480e-01
-1.48613584e+00 1.71222985e-02 -2.92072427e-02 -1.63814276e-01
-5.13832211e-01 1.31107196e-01 -2.63228416e-01 1.49513677e-01
8.76176238e-01 6.10996485e-01 -2.65728980e-01 -6.78889990e-01
2.40364656e-01 7.20572174e-01 1.04151237e+00 -8.20779681e-01
1.23903382e+00 5.78140318e-01 -7.24717259e-01 -5.45662045e-01
-8.77983809e-01 -6.64643764e-01 -1.09950137e+00 -3.93577516e-02
5.71010411e-01 -1.10249627e+00 -3.30939978e-01 9.32370543e-01
-7.84077466e-01 -6.09043896e-01 -4.53746736e-01 3.73462468e-01
-2.38559976e-01 6.13885224e-01 -3.89058203e-01 -4.41825032e-01
-2.28987902e-01 -4.99953717e-01 7.39623010e-01 5.42320907e-01
-3.78201753e-01 -1.05468202e+00 4.06985492e-01 2.57361174e-01
2.20574737e-01 -1.52796969e-01 3.29015851e-01 -9.97105956e-01
6.45050313e-03 -2.81528383e-01 -1.37067139e-01 3.11245620e-01
3.18416536e-01 -7.80697823e-01 -9.95109081e-01 -8.67890894e-01
-9.77348164e-02 -3.46436918e-01 1.05304492e+00 -1.19812384e-01
8.78020227e-01 -3.75977039e-01 -5.43322086e-01 6.31917775e-01
1.08216977e+00 1.26700342e-01 3.89490545e-01 6.43279970e-01
6.22401059e-01 5.08050025e-01 5.18902779e-01 4.44933653e-01
6.81385994e-01 5.30361354e-01 -2.55375117e-01 1.23285525e-01
-2.54617661e-01 -6.16266727e-01 3.83839756e-01 5.55139542e-01
7.17557073e-02 -1.76817536e-01 -1.08551204e+00 9.94544625e-01
-1.75814211e+00 -9.39208746e-01 4.90153402e-01 2.51366568e+00
1.05162621e+00 2.36683637e-01 4.01330441e-01 -3.58088434e-01
1.08414841e+00 -1.62140742e-01 -1.07636964e+00 4.92480062e-02
-2.36869708e-01 6.60150051e-02 3.76702636e-01 4.74088430e-01
-1.08924639e+00 1.13386631e+00 5.82466030e+00 4.71296012e-01
-8.68544698e-01 3.36225092e-01 3.76464874e-01 -1.73727304e-01
-7.69110769e-02 2.31645145e-02 -1.04593194e+00 5.64932466e-01
9.67931986e-01 -7.41326809e-01 3.86335492e-01 9.29050684e-01
-1.32993579e-01 -1.24654837e-01 -1.28127253e+00 1.23981762e+00
2.51585484e-01 -8.10488164e-01 2.31606632e-01 -1.60550892e-01
8.36191356e-01 -3.06883175e-02 1.48436069e-01 5.37661433e-01
4.25619692e-01 -6.41054749e-01 4.47485030e-01 5.92815280e-01
7.91237235e-01 -7.14197218e-01 6.46534383e-01 5.32037139e-01
-1.00890684e+00 -1.69338152e-01 -5.36755204e-01 -3.56887281e-02
-3.98300886e-02 2.56918281e-01 -7.68835127e-01 2.79170275e-01
9.91229832e-01 1.01515806e+00 -9.72837210e-01 1.01124835e+00
-2.92632371e-01 3.65632236e-01 -2.56673336e-01 4.29164827e-01
-3.00976783e-01 1.69592589e-01 5.18740058e-01 1.33579123e+00
2.43203461e-01 2.40899280e-01 1.27882168e-01 5.09162486e-01
-1.99711636e-01 -9.82020944e-02 -5.30106366e-01 3.02397549e-01
7.37010896e-01 8.28361988e-01 -3.10456216e-01 -7.04330921e-01
-3.47064197e-01 1.62004399e+00 4.90669280e-01 7.45806038e-01
-5.42495191e-01 -1.64106399e-01 5.47849953e-01 3.25913399e-01
6.73140213e-02 -3.04068357e-01 1.81910530e-01 -1.64004135e+00
9.51866582e-02 -7.44211435e-01 9.87344801e-01 -5.79486191e-01
-1.88839090e+00 6.27870142e-01 3.54730517e-01 -1.24486578e+00
-3.16266865e-01 -1.14496917e-01 -5.67589939e-01 7.35284328e-01
-1.49584079e+00 -1.05282760e+00 -6.89446926e-01 9.95127916e-01
6.19889081e-01 -4.26250935e-01 7.93651283e-01 2.87694097e-01
-6.77227736e-01 1.10816276e+00 3.59248638e-01 3.03968430e-01
1.46632373e+00 -1.02927518e+00 7.71464944e-01 1.02537334e+00
-7.58706406e-02 9.45144296e-01 7.03277349e-01 -8.92777264e-01
-1.02172625e+00 -1.18870497e+00 7.89799154e-01 -5.02460659e-01
2.95628905e-01 -4.88249123e-01 -1.30584145e+00 1.02428257e+00
-1.32428613e-02 4.47061798e-03 6.33689165e-01 1.59164160e-01
-5.25361717e-01 -4.30889219e-01 -1.32515764e+00 4.54534620e-01
1.55488622e+00 -3.75612557e-01 -8.48487973e-01 1.71341747e-01
3.86515766e-01 -3.33713412e-01 -6.29651666e-01 2.23409787e-01
4.30401862e-01 -8.07762623e-01 9.67331052e-01 -5.62237501e-01
-3.26624840e-01 -3.57155710e-01 2.34228864e-01 -1.46703780e+00
-6.06660128e-01 -5.00240445e-01 -1.37005582e-01 1.43429673e+00
6.24609888e-02 -1.13058257e+00 9.75351393e-01 8.74020934e-01
2.04079568e-01 1.37875497e-01 -8.29579771e-01 -1.25922954e+00
1.95172116e-01 -1.27692923e-01 8.27451408e-01 1.06192303e+00
-1.22450128e-01 3.44727546e-01 -4.18740034e-01 4.08562571e-01
1.06143844e+00 -1.87193170e-01 1.20565999e+00 -1.32582259e+00
-1.56817004e-01 9.26083177e-02 -4.26637352e-01 -1.24736071e+00
4.22198445e-01 -8.86595428e-01 -1.82925940e-01 -1.24294329e+00
6.45697653e-01 -4.85042393e-01 -6.07072592e-01 5.56934953e-01
-4.33192879e-01 2.92086452e-01 4.87774312e-02 6.08806133e-01
-8.73488426e-01 6.24626935e-01 9.70559597e-01 -3.54974896e-01
-4.09594417e-01 -2.62483060e-02 -9.37733889e-01 5.32628715e-01
8.74261796e-01 -5.56059659e-01 -6.12328589e-01 -7.01885104e-01
-2.66797692e-01 -5.42329133e-01 4.15587187e-01 -1.34653986e+00
6.14556611e-01 -6.89856112e-02 8.46149325e-01 -2.50989228e-01
1.88934267e-01 -5.47575295e-01 2.17050195e-01 1.55221581e-01
-3.50766540e-01 3.57526867e-03 3.29487443e-01 6.96615279e-01
-5.54458536e-02 -2.16902807e-01 9.42866087e-01 -1.27881244e-01
-1.38964093e+00 6.21154785e-01 1.01889984e-03 2.24802822e-01
1.03846717e+00 -5.01865208e-01 -2.63930261e-01 -4.41205531e-01
-1.10279739e+00 4.69159156e-01 1.00737035e+00 7.13453233e-01
7.54873693e-01 -1.36713719e+00 -9.19045150e-01 4.99784559e-01
5.75226009e-01 -8.73235893e-03 5.59405744e-01 1.97565272e-01
4.39645723e-02 -1.36589454e-02 -3.74964297e-01 -5.07588804e-01
-1.12663591e+00 9.80481446e-01 3.13549131e-01 -3.57365571e-02
-7.94839025e-01 7.95759201e-01 5.65837502e-01 -4.71133232e-01
1.91626817e-01 4.66558158e-01 -1.76073447e-01 -1.78051099e-01
8.99136066e-01 2.42358536e-01 -2.69805580e-01 -9.01638627e-01
-4.87713784e-01 6.44928336e-01 -7.25972950e-01 -1.38861477e-01
1.21499527e+00 -5.96004784e-01 2.12330371e-01 3.46776485e-01
1.10861540e+00 -2.34868601e-01 -1.41685617e+00 -8.39689851e-01
3.91017366e-03 -5.66332161e-01 -5.02750695e-01 -7.47395515e-01
-6.47257566e-01 5.00269592e-01 8.42819512e-01 -3.77928466e-01
1.32017863e+00 1.97754532e-01 1.06551862e+00 5.75555027e-01
5.44199109e-01 -1.35402262e+00 2.35523820e-01 4.76207882e-01
5.93802929e-01 -1.37027383e+00 9.07481909e-02 8.46280158e-02
-6.67793989e-01 8.27972531e-01 9.27406669e-01 -7.94356987e-02
5.06552279e-01 -2.31420115e-01 1.25327725e-02 1.37985140e-01
-3.62541854e-01 -2.90464938e-01 1.28643751e-01 1.05036771e+00
-1.11415023e-02 -5.78366779e-02 3.84240435e-03 6.84021890e-01
-2.49225557e-01 2.64299840e-01 4.27261621e-01 8.98833334e-01
-3.17454994e-01 -1.37562406e+00 -4.43707645e-01 1.90766498e-01
2.91967746e-02 9.62428376e-02 -4.16588306e-01 7.01172471e-01
2.70935565e-01 8.36647630e-01 8.72073099e-02 -3.22340339e-01
4.94924545e-01 1.18658677e-01 4.12301332e-01 -8.96667778e-01
-1.78876042e-01 -2.93721497e-01 -1.31667122e-01 -9.70352963e-02
-3.80401582e-01 -9.37676013e-01 -1.16770566e+00 -4.84745502e-01
9.80153009e-02 2.25746349e-01 5.72464988e-02 8.76932025e-01
4.49536085e-01 -9.90587473e-02 5.22738099e-01 -6.13218784e-01
-3.79554898e-01 -9.09777939e-01 -5.68305492e-01 9.44102645e-01
4.32131648e-01 -6.90286458e-01 -1.56167805e-01 4.19162959e-01]
|
[14.741759300231934, 1.123348593711853]
|
6307a7f3-28b2-48bf-a5be-445e99e6c2ff
|
segmenting-moving-objects-via-an-object
|
2207.02206
| null |
https://arxiv.org/abs/2207.02206v2
|
https://arxiv.org/pdf/2207.02206v2.pdf
|
Segmenting Moving Objects via an Object-Centric Layered Representation
|
The objective of this paper is a model that is able to discover, track and segment multiple moving objects in a video. We make four contributions: First, we introduce an object-centric segmentation model with a depth-ordered layer representation. This is implemented using a variant of the transformer architecture that ingests optical flow, where each query vector specifies an object and its layer for the entire video. The model can effectively discover multiple moving objects and handle mutual occlusions; Second, we introduce a scalable pipeline for generating multi-object synthetic training data via layer compositions, that is used to train the proposed model, significantly reducing the requirements for labour-intensive annotations, and supporting Sim2Real generalisation; Third, we conduct thorough ablation studies, showing that the model is able to learn object permanence and temporal shape consistency, and is able to predict amodal segmentation masks; Fourth, we evaluate our model, trained only on synthetic data, on standard video segmentation benchmarks, DAVIS, MoCA, SegTrack, FBMS-59, and achieve state-of-the-art performance among existing methods that do not rely on any manual annotations. With test-time adaptation, we observe further performance boosts.
|
['Andrew Zisserman', 'Weidi Xie', 'Junyu Xie']
|
2022-07-05
| null | null | null | null |
['unsupervised-object-segmentation', 'motion-segmentation']
|
['computer-vision', 'computer-vision']
|
[ 2.57367045e-01 5.38871512e-02 -1.13975003e-01 -7.70508870e-02
-7.51419425e-01 -8.10701847e-01 6.28082514e-01 -2.89215714e-01
-3.32848996e-01 5.68140507e-01 -1.56103969e-01 -1.42491415e-01
6.95963800e-02 -4.20961618e-01 -1.14963806e+00 -5.00169754e-01
-1.92656577e-01 7.79642522e-01 1.05450726e+00 -2.06895601e-02
7.28289410e-02 5.67537248e-01 -1.71186686e+00 5.51734865e-01
8.94253314e-01 1.19129550e+00 2.19924062e-01 1.10660696e+00
6.24058433e-02 1.09625125e+00 -5.08916080e-01 -4.14317876e-01
3.51150066e-01 -4.58707064e-01 -1.31803548e+00 4.44911301e-01
1.11822212e+00 -4.72259820e-01 -2.07884982e-01 6.06977105e-01
2.40419626e-01 1.65346429e-01 6.44151866e-01 -1.36653018e+00
-3.27158660e-01 5.94406784e-01 -4.39579725e-01 3.19312990e-01
2.78594792e-01 6.24465585e-01 8.48991692e-01 -8.50647867e-01
1.18923199e+00 1.33119845e+00 6.24629021e-01 8.00183833e-01
-1.64924896e+00 -2.74492472e-01 4.67671514e-01 2.98910856e-01
-1.00266492e+00 -5.95915318e-01 3.66601437e-01 -6.04028523e-01
8.83686006e-01 2.93823957e-01 9.05060291e-01 1.42317164e+00
-4.91077453e-03 1.45097971e+00 8.84454608e-01 -1.07721902e-01
2.90771246e-01 -9.68685299e-02 -1.19535867e-02 7.62370348e-01
-4.93636876e-02 2.50026919e-02 -4.30700123e-01 3.26121718e-01
6.80584371e-01 -2.94979542e-01 -3.53251427e-01 -6.71700120e-01
-1.52223945e+00 2.71257877e-01 3.28439623e-01 1.79804757e-01
-3.68594915e-01 4.70152050e-01 5.76445878e-01 5.86132072e-02
3.50447893e-01 3.17272335e-01 -4.06657785e-01 -2.09845230e-01
-1.51790392e+00 6.49501562e-01 8.21759105e-01 1.28732204e+00
7.19332278e-01 1.25453889e-01 -6.91486537e-01 3.43914896e-01
7.17495531e-02 4.63704437e-01 5.29550612e-01 -1.55020595e+00
3.84684652e-01 3.87191892e-01 1.48992687e-01 -6.19356275e-01
-4.87538099e-01 -1.72530532e-01 -4.92999583e-01 2.89670259e-01
4.36973125e-01 -2.05319282e-03 -1.40967798e+00 1.58223855e+00
3.66659015e-01 8.21022153e-01 3.06398608e-02 9.74860072e-01
8.54857028e-01 4.53544974e-01 1.82949901e-01 8.03745687e-02
1.16939998e+00 -1.59326100e+00 -4.81677860e-01 -2.93972999e-01
3.87532413e-01 -4.99385059e-01 1.03862488e+00 3.68645996e-01
-1.56554019e+00 -7.57752120e-01 -8.95198345e-01 -8.20962805e-03
-1.99754104e-01 -6.20522127e-02 8.02758813e-01 3.21694046e-01
-1.35341930e+00 7.89221644e-01 -1.27237082e+00 -3.02607715e-01
6.97156727e-01 2.37205923e-01 -3.11374366e-01 -3.43474187e-02
-7.94676840e-01 5.06670713e-01 7.30779231e-01 9.18225795e-02
-1.41427422e+00 -9.10088599e-01 -9.59810078e-01 -4.19243984e-02
2.95528889e-01 -1.00243008e+00 1.30624604e+00 -1.15314162e+00
-1.35561776e+00 8.75054419e-01 -2.67890483e-01 -6.83297634e-01
1.02516997e+00 -1.99014872e-01 -2.12956190e-01 5.28532922e-01
2.59886414e-01 1.41287804e+00 9.27833617e-01 -1.40262413e+00
-7.68039346e-01 9.36411545e-02 1.09584071e-01 -4.42680791e-02
9.91998091e-02 -1.67108905e-02 -1.03312850e+00 -8.23888242e-01
-1.01208441e-01 -9.79812980e-01 -3.46488386e-01 -1.17835812e-02
-7.43021548e-01 -1.02700694e-02 9.67445433e-01 -5.31018853e-01
9.37564790e-01 -1.99559903e+00 5.08604944e-01 -8.76411572e-02
5.45536615e-02 3.67703438e-01 -4.39114898e-01 5.94339101e-04
6.22255355e-02 6.21994883e-02 -7.10722506e-01 -8.02961946e-01
-5.18833101e-02 2.68583804e-01 -3.94054562e-01 3.36767793e-01
5.85455060e-01 1.29385674e+00 -9.07724321e-01 -7.20878005e-01
4.21784699e-01 2.28478804e-01 -7.16949105e-01 4.14469540e-01
-7.01435328e-01 6.57999098e-01 -3.11544705e-02 8.04745317e-01
6.58083975e-01 -4.72786337e-01 -1.48804948e-01 -2.06818148e-01
2.46216618e-02 -9.83790606e-02 -1.34264755e+00 2.10808396e+00
-1.44546852e-01 6.81550145e-01 -6.16414584e-02 -8.14856410e-01
2.18237698e-01 3.42568785e-01 6.50093257e-01 -6.25651658e-01
-2.49799900e-02 2.46804580e-01 -2.91757762e-01 -7.49274373e-01
2.97738642e-01 4.87145543e-01 2.24504411e-01 3.37604314e-01
4.34016466e-01 -1.89981788e-01 9.57392633e-01 2.86567211e-01
1.00123334e+00 7.52099812e-01 -3.59318286e-01 -1.40434027e-01
4.82558608e-01 1.80416271e-01 4.60037470e-01 8.28287601e-01
-2.04888776e-01 9.56033885e-01 4.92578059e-01 -5.85501671e-01
-9.59057987e-01 -1.12936819e+00 -1.06103405e-01 1.12461567e+00
4.20245141e-01 -1.58197358e-01 -9.67592299e-01 -8.79275858e-01
-1.85536772e-01 6.26430869e-01 -6.61036849e-01 1.17477655e-01
-8.13434303e-01 -6.40430927e-01 6.72998190e-01 7.43108332e-01
4.75341588e-01 -1.51336849e+00 -1.12445652e+00 2.68474877e-01
-3.95877063e-01 -1.56024241e+00 -4.73551244e-01 -1.60222426e-01
-9.10164177e-01 -1.33439016e+00 -9.00374711e-01 -8.74036789e-01
4.29974824e-01 -3.35104391e-02 1.44718814e+00 -4.39475179e-02
-4.66884971e-01 4.97970521e-01 -1.90754816e-01 2.18023639e-02
-4.20274466e-01 1.99771523e-01 -2.04899564e-01 1.87211215e-01
-1.06473282e-01 -4.87413198e-01 -7.99732685e-01 3.69835556e-01
-1.11385119e+00 3.90663773e-01 4.05257732e-01 4.80386466e-01
6.97448909e-01 -5.57775736e-01 3.07589859e-01 -9.25282657e-01
1.34716287e-01 -2.23376676e-01 -6.10163391e-01 2.64739990e-01
-2.19166189e-01 8.33056942e-02 4.69136775e-01 -5.36917210e-01
-9.27995920e-01 3.65848690e-01 -1.27097577e-01 -7.38451123e-01
-4.35349196e-01 -1.43457606e-01 2.33584326e-02 -7.44822994e-02
6.28582239e-01 3.04459751e-01 -2.29544431e-01 -2.84812093e-01
7.52260923e-01 6.52326643e-02 9.54963803e-01 -5.84898829e-01
5.76226532e-01 7.15881228e-01 -7.34428987e-02 -5.48787177e-01
-7.59582758e-01 -4.58006412e-01 -9.47761834e-01 -3.04817826e-01
1.17801189e+00 -8.67406070e-01 -7.50327468e-01 6.19149327e-01
-1.28432643e+00 -9.07762527e-01 -4.15855467e-01 3.17332596e-01
-1.02234006e+00 2.13861227e-01 -7.66776621e-01 -4.49386626e-01
-1.95235968e-01 -1.36302400e+00 1.27237117e+00 2.86743790e-01
-2.09563941e-01 -1.08895516e+00 3.22532505e-02 2.95447409e-01
4.23120469e-01 6.08202100e-01 4.09689993e-01 -5.84773123e-01
-1.27064598e+00 1.72416151e-01 -2.79871106e-01 2.87295163e-01
-2.50563741e-01 2.93026447e-01 -9.91033614e-01 -4.77797508e-01
-5.41319013e-01 -6.88745439e-01 1.09491777e+00 5.14856279e-01
1.35467076e+00 -3.00082684e-01 -5.06294310e-01 1.05105400e+00
1.11128259e+00 -1.29983306e-01 6.44746006e-01 3.06789994e-01
9.36147630e-01 5.27748525e-01 4.56257880e-01 8.09857696e-02
5.62776923e-01 8.06094587e-01 7.34746754e-01 -4.25693929e-01
-5.54157734e-01 -8.92946646e-02 3.19068044e-01 3.18949908e-01
-2.51192570e-01 -3.05850834e-01 -7.11158335e-01 8.82363200e-01
-2.07820773e+00 -9.61922109e-01 -2.11473912e-01 1.82096148e+00
6.29616678e-01 2.63809025e-01 3.82536203e-01 -1.66733816e-01
3.73388380e-01 2.13399500e-01 -7.68125057e-01 2.29759254e-02
-1.85935751e-01 2.64205098e-01 4.32221532e-01 5.51387310e-01
-1.38306606e+00 1.43332767e+00 7.07551622e+00 4.68812913e-01
-1.12851536e+00 6.11877330e-02 5.93172967e-01 -1.87835693e-01
-1.84381738e-01 -1.16854846e-01 -6.57844484e-01 3.58721554e-01
8.05803537e-01 2.88724720e-01 5.37914217e-01 5.78012347e-01
2.39163890e-01 1.18386671e-01 -1.37016511e+00 8.02034557e-01
1.98971927e-01 -1.63616765e+00 1.76542088e-01 -2.73912758e-01
9.57900226e-01 2.47662678e-01 -2.18200758e-02 2.07668811e-01
3.67149025e-01 -9.94227231e-01 1.18668878e+00 6.43131196e-01
7.68789887e-01 -3.84335041e-01 4.39631969e-01 1.12032622e-01
-1.15662956e+00 4.05969024e-02 4.18638997e-02 4.23765242e-01
5.44779599e-01 9.67202038e-02 -7.17560470e-01 4.64074671e-01
9.05533195e-01 9.40436542e-01 -8.44051540e-01 1.29862380e+00
-2.48256996e-01 6.33072793e-01 -1.82208955e-01 4.62462187e-01
4.62782234e-01 6.32580519e-02 6.11932278e-01 1.60297084e+00
-3.23970206e-02 -3.80689800e-01 4.84132469e-01 1.17539537e+00
2.44960506e-02 -2.19946653e-01 -1.62296042e-01 1.41615912e-01
2.25042492e-01 1.22409260e+00 -1.02726543e+00 -5.78460217e-01
-1.44750610e-01 1.39550424e+00 1.97192982e-01 7.22162127e-01
-1.13513207e+00 -2.43638515e-01 4.68809009e-01 1.39705107e-01
6.50106132e-01 -4.32636030e-02 2.38328408e-02 -1.25387073e+00
5.25222272e-02 -8.34035277e-01 3.46219331e-01 -7.51760423e-01
-8.50393414e-01 7.94508994e-01 7.38679692e-02 -1.04477918e+00
-5.90614915e-01 -5.87619007e-01 -5.32492995e-01 5.44624090e-01
-1.46613812e+00 -1.29957056e+00 -4.02399123e-01 6.00196719e-01
8.59186113e-01 3.07320170e-02 5.27233362e-01 3.79782677e-01
-6.34227157e-01 4.61055845e-01 -3.51716489e-01 2.17165053e-01
5.14067769e-01 -1.42825043e+00 8.83644521e-01 1.05704474e+00
3.12744826e-01 2.21124336e-01 6.19346440e-01 -3.87812704e-01
-9.71684813e-01 -1.42698634e+00 5.31591594e-01 -7.57602811e-01
4.76544529e-01 -5.61778426e-01 -9.44096208e-01 1.06545973e+00
1.84648499e-01 5.54567218e-01 1.72809571e-01 -4.48287338e-01
-1.48185641e-01 7.39328563e-02 -1.01046729e+00 5.92955887e-01
1.41497898e+00 -6.49931878e-02 -4.13779527e-01 4.65626121e-01
1.01662767e+00 -8.71191859e-01 -6.99967682e-01 4.16677386e-01
4.31556374e-01 -1.32005036e+00 1.12747622e+00 -8.69152844e-01
5.55426538e-01 -4.99146968e-01 1.77369103e-01 -9.94589686e-01
-1.96214437e-01 -7.55299568e-01 -6.59218550e-01 1.06073499e+00
4.66793120e-01 -1.46233752e-01 8.51460814e-01 4.91848081e-01
-3.43707681e-01 -7.43279099e-01 -8.71629298e-01 -6.75574541e-01
-3.41461778e-01 -5.01503050e-01 3.98742050e-01 7.41121948e-01
-6.16373003e-01 1.48337841e-01 -3.14616680e-01 3.37655507e-02
5.83141088e-01 1.85239732e-01 9.03288126e-01 -9.02313411e-01
-5.41314125e-01 -6.17373884e-01 -4.54448551e-01 -1.48189151e+00
2.17932776e-01 -8.07347715e-01 1.12597212e-01 -1.66433954e+00
-2.35243216e-02 -3.39001000e-01 -1.22917965e-01 6.13653123e-01
-1.29924804e-01 7.11247325e-01 3.86653364e-01 2.00925946e-01
-1.18379629e+00 3.86513710e-01 1.56446874e+00 -2.83590019e-01
-3.49407852e-01 1.06952824e-01 -2.23250687e-01 7.19125926e-01
1.66057885e-01 -2.90089637e-01 -4.64502931e-01 -6.02144182e-01
-1.96724042e-01 -1.42115876e-02 7.38002181e-01 -1.24993753e+00
1.49881065e-01 -1.38687223e-01 4.42821592e-01 -5.51364124e-01
2.50759572e-01 -6.76363528e-01 1.39125735e-01 5.05005836e-01
-2.60645568e-01 6.79385960e-02 4.07721221e-01 4.85055029e-01
-1.63780317e-01 -3.46103609e-02 7.89817095e-01 -2.01888412e-01
-1.09339976e+00 5.88133872e-01 -1.49306521e-01 4.03173327e-01
1.21716666e+00 -2.84852773e-01 -5.71082458e-02 2.34586447e-02
-8.47734451e-01 5.10004580e-01 5.90275526e-01 7.08607852e-01
4.39528495e-01 -1.16053462e+00 -7.43645430e-01 1.80218190e-01
8.72401446e-02 4.37442750e-01 3.13668311e-01 9.34304118e-01
-9.49761987e-01 3.40146333e-01 -2.08223224e-01 -1.23977685e+00
-8.58029068e-01 6.04854643e-01 5.60282826e-01 -8.00145864e-02
-9.22065616e-01 9.44031417e-01 2.49554679e-01 -2.98845023e-01
3.13934207e-01 -6.38319016e-01 -5.93116172e-02 -9.38582234e-03
6.23541713e-01 4.11487281e-01 -9.74699017e-03 -6.34890497e-01
-4.83268082e-01 6.81573331e-01 -8.25998932e-02 -2.64028400e-01
1.06695330e+00 -1.06086414e-02 9.04049277e-02 3.75582695e-01
8.72179925e-01 -3.48234832e-01 -1.95976496e+00 8.59347284e-02
1.24440119e-02 -3.73252809e-01 -4.19947982e-01 -7.56154478e-01
-1.19394195e+00 7.13756084e-01 5.61400235e-01 1.84467688e-01
1.03205287e+00 1.22295536e-01 8.76002789e-01 1.74943492e-01
1.12218998e-01 -8.26339245e-01 2.36394614e-01 4.70810324e-01
6.63648725e-01 -1.19593370e+00 -4.15203810e-01 -2.70335227e-01
-7.86858737e-01 9.76968348e-01 7.80226231e-01 -1.06513269e-01
1.08014017e-01 1.32963970e-01 1.30827531e-01 -1.50048777e-01
-7.48765171e-01 -2.80807018e-01 4.41634625e-01 6.13062859e-01
1.50809586e-01 -3.04998279e-01 2.12504536e-01 6.03792332e-02
-3.70018221e-02 4.79295999e-02 3.75492573e-01 7.17711568e-01
-6.48997352e-02 -8.43312740e-01 -3.90965007e-02 1.28094971e-01
-3.53435606e-01 2.94396803e-02 -1.02935173e-01 8.79405618e-01
2.29497269e-01 5.94130635e-01 3.57493579e-01 -7.02440590e-02
5.18577278e-01 -3.77162248e-02 5.33435941e-01 -4.49546993e-01
-4.16415334e-01 -7.20488131e-02 -1.65019214e-01 -1.03216958e+00
-7.72503734e-01 -5.97825527e-01 -1.34739494e+00 5.56609742e-02
1.60164684e-02 -1.76521659e-01 3.88525337e-01 1.05966258e+00
5.10225475e-01 8.75479221e-01 1.40976802e-01 -1.37419724e+00
-1.02121122e-01 -7.39907265e-01 -7.68341646e-02 7.44000375e-01
6.55690610e-01 -5.05197108e-01 -1.78602979e-01 6.41358733e-01]
|
[9.129650115966797, -0.09626523405313492]
|
efee579b-8e03-433d-b161-318e54858c65
|
selective-query-processing-a-risk-sensitive
|
2305.18311
| null |
https://arxiv.org/abs/2305.18311v1
|
https://arxiv.org/pdf/2305.18311v1.pdf
|
Selective Query Processing: a Risk-Sensitive Selection of System Configurations
|
In information retrieval systems, search parameters are optimized to ensure high effectiveness based on a set of past searches and these optimized parameters are then used as the system configuration for all subsequent queries. A better approach, however, would be to adapt the parameters to fit the query at hand. Selective query expansion is one such an approach, in which the system decides automatically whether or not to expand the query, resulting in two possible system configurations. This approach was extended recently to include many other parameters, leading to many possible system configurations where the system automatically selects the best configuration on a per-query basis. To determine the ideal configurations to use on a per-query basis in real-world systems we developed a method in which a restricted number of possible configurations is pre-selected and then used in a meta-search engine that decides the best search configuration on a per query basis. We define a risk-sensitive approach for configuration pre-selection that considers the risk-reward trade-off between the number of configurations kept, and system effectiveness. For final configuration selection, the decision is based on query feature similarities. We find that a relatively small number of configurations (20) selected by our risk-sensitive model is sufficient to increase effectiveness by about 15% according(P@10, nDCG@10) when compared to traditional grid search using a single configuration and by about 20% when compared to learning to rank documents. Our risk-sensitive approach works for both diversity- and ad hoc-oriented searches. Moreover, the similarity-based selection method outperforms the more sophisticated approaches. Thus, we demonstrate the feasibility of developing per-query information retrieval systems, which will guide future research in this direction.
|
['Md Zia Ullah', 'Josiane Mothe']
|
2023-05-17
| null | null | null | null |
['information-retrieval']
|
['natural-language-processing']
|
[ 3.58738266e-02 -4.18288767e-01 -4.77610797e-01 -2.28992134e-01
-1.03993952e+00 -8.23738873e-01 5.51512659e-01 5.69516480e-01
-8.25099528e-01 5.47119141e-01 2.27547064e-02 -4.08589810e-01
-6.83165252e-01 -9.13357258e-01 -9.78012756e-02 -5.20505250e-01
1.00856721e-02 9.90464509e-01 7.01851487e-01 -3.74448657e-01
8.35483432e-01 6.95965707e-01 -1.92834723e+00 1.86310202e-01
8.47298205e-01 7.23539531e-01 4.53185499e-01 7.03811705e-01
-1.84718281e-01 9.73962992e-02 -7.23975778e-01 1.84628088e-03
3.16380501e-01 -4.80630070e-01 -9.56209183e-01 -3.59471589e-01
-1.54596657e-01 -1.79023951e-01 2.53954738e-01 7.45927155e-01
8.72492313e-01 3.93160492e-01 5.67144632e-01 -6.78742349e-01
4.46388051e-02 5.58908880e-01 -1.06430382e-01 4.56622481e-01
8.12343359e-01 -1.08725868e-01 1.16376209e+00 -6.37740135e-01
5.15234947e-01 1.12265193e+00 7.96835944e-02 -1.28684685e-01
-1.28415155e+00 -4.15395230e-01 7.07640275e-02 2.15500370e-01
-1.85143971e+00 -3.03745002e-01 4.55948472e-01 -1.70071334e-01
1.21058178e+00 6.70086622e-01 7.05410182e-01 1.37752786e-01
2.61739135e-01 3.62924814e-01 7.04619169e-01 -1.11756837e+00
5.98491073e-01 5.00153840e-01 2.05367684e-01 2.71967947e-01
3.48655224e-01 2.25679308e-01 -1.40099034e-01 -7.22081780e-01
3.20474923e-01 -2.61626184e-01 -2.22176567e-01 -3.69616151e-01
-7.93883741e-01 9.86921608e-01 2.85764188e-01 7.87594378e-01
-6.33556783e-01 -2.37923622e-01 1.69985190e-01 6.11774683e-01
-1.11451419e-02 1.16672814e+00 -4.61451054e-01 -5.97932860e-02
-1.12288821e+00 4.75074798e-01 9.82007325e-01 5.32108068e-01
7.75271833e-01 -6.01617813e-01 -6.65493727e-01 9.64259863e-01
2.60323197e-01 3.20531249e-01 6.90105975e-01 -8.19379687e-01
1.12445854e-01 7.91517377e-01 2.78415948e-01 -8.65686953e-01
-3.28479588e-01 -5.00218093e-01 -7.64115974e-02 4.32304218e-02
4.12301160e-02 1.17354713e-01 -9.26135480e-01 1.61444449e+00
3.20121020e-01 -7.64964819e-01 -8.86285827e-02 6.69055998e-01
3.01428121e-02 6.01172864e-01 -6.09610565e-02 -8.73548925e-01
1.24798656e+00 -4.01236862e-01 -4.53762233e-01 3.35566998e-01
9.05607402e-01 -1.10011137e+00 1.14377654e+00 5.65137506e-01
-1.06019890e+00 -3.21418375e-01 -1.01726925e+00 7.06517041e-01
-4.35810655e-01 -5.95645346e-02 1.64535969e-01 8.97117257e-01
-1.27689803e+00 4.10678029e-01 -4.61763829e-01 -5.44579208e-01
-4.52402890e-01 7.59751797e-01 2.97200114e-01 3.15524936e-02
-1.34249067e+00 1.07951081e+00 7.96251535e-01 -3.97803694e-01
-4.75420684e-01 -1.93149373e-01 -2.21853420e-01 2.91700423e-01
7.59284079e-01 -8.25080335e-01 1.24776077e+00 -4.52208400e-01
-1.30309021e+00 2.87378103e-01 -1.62687913e-01 -2.45050609e-01
3.25850815e-01 1.51990160e-01 -3.73217911e-01 1.34156808e-01
-1.93546303e-02 3.32436055e-01 4.76789534e-01 -1.23976254e+00
-8.64946425e-01 -2.67963886e-01 2.20256597e-01 6.76527143e-01
-4.31106061e-01 3.18142116e-01 -1.04074430e+00 -1.76442504e-01
1.47333099e-02 -1.14280474e+00 -5.19688427e-01 -7.60451138e-01
-1.50587469e-01 -4.23337311e-01 3.34755361e-01 -2.49738932e-01
2.06895232e+00 -1.68306231e+00 -6.54633418e-02 9.53826368e-01
-4.36597556e-01 3.88374656e-01 -1.89618558e-01 8.45147669e-01
1.73836127e-01 3.81854862e-01 2.26591960e-01 3.27792168e-01
-2.18155086e-01 -1.06420949e-01 2.57371347e-02 -1.58842608e-01
-4.35212255e-01 4.22612220e-01 -7.43900657e-01 -6.92544401e-01
1.50094330e-01 2.42413983e-01 -7.90791750e-01 2.43388012e-01
-2.77685642e-01 -7.60344192e-02 -8.37220252e-01 4.42363888e-01
2.41614699e-01 -2.10195020e-01 3.99210304e-01 -2.89587583e-02
-1.28465086e-01 3.04676414e-01 -1.36313093e+00 1.06371188e+00
-7.42984235e-01 -2.79787742e-03 -4.89238918e-01 -7.05173016e-01
9.00298774e-01 4.31526244e-01 6.68858349e-01 -9.22177255e-01
-8.07203166e-03 4.53265578e-01 1.93112835e-01 -3.53331655e-01
7.09433258e-01 2.37422571e-01 -1.10592373e-01 7.51719594e-01
-4.54171926e-01 -1.48735836e-01 5.74991763e-01 2.34704286e-01
1.11373806e+00 -3.24976355e-01 6.98094666e-01 -1.61644354e-01
6.94623888e-01 2.08273366e-01 2.12340042e-01 1.13183224e+00
1.73614934e-01 2.97515810e-01 2.31993660e-01 -3.02632451e-01
-8.79480362e-01 -4.80737507e-01 -3.01971376e-01 1.34320235e+00
2.28353366e-01 -5.35017610e-01 -7.20192134e-01 -6.42138839e-01
-1.89471450e-02 1.09513319e+00 -2.75825649e-01 -2.25680619e-01
-5.14417648e-01 -7.04348087e-01 8.94758999e-02 -1.30200326e-01
1.79413334e-01 -1.23346651e+00 -8.16132963e-01 3.55244279e-01
-3.57621312e-02 -2.31961146e-01 -5.43755531e-01 2.62144417e-01
-8.54627490e-01 -8.82405877e-01 -6.45580411e-01 -3.96247238e-01
5.69867253e-01 2.12348461e-01 1.06795120e+00 4.87823039e-01
-1.73233986e-01 4.06685680e-01 -5.79990089e-01 -9.27876458e-02
-4.99923766e-01 4.22216952e-01 7.87002146e-02 -4.73434776e-01
3.55672777e-01 -2.94395119e-01 -8.77361834e-01 6.44672334e-01
-9.84182835e-01 -6.79447412e-01 9.22976434e-01 7.85743356e-01
7.96875119e-01 5.13345897e-01 4.61741447e-01 -8.21944058e-01
1.28079319e+00 -3.60159278e-01 -7.53114223e-01 9.57091391e-01
-1.51666427e+00 4.78531569e-01 4.01195943e-01 -6.02014124e-01
-9.02920961e-01 3.07386015e-02 -6.26876354e-02 -2.59116711e-03
6.92138076e-02 8.55795562e-01 9.91308410e-03 -1.65436432e-01
1.01071787e+00 2.10663572e-01 -2.93663323e-01 -4.05747235e-01
2.07357049e-01 8.13750327e-01 -8.19037482e-02 -4.17974442e-01
4.12080586e-01 -3.49476546e-01 -2.12350368e-01 -5.59913576e-01
-1.79068178e-01 -7.95925438e-01 -3.69441986e-01 -4.27564681e-02
2.21230984e-01 -4.23478395e-01 -6.82435036e-01 -2.78983831e-01
-7.16152132e-01 8.16383958e-03 -1.86369732e-01 4.63274896e-01
-3.64163458e-01 3.57439429e-01 7.98816308e-02 -1.01482844e+00
-4.46128130e-01 -1.26567519e+00 7.74858356e-01 2.59712428e-01
-5.74852645e-01 -6.84115887e-01 3.52437764e-01 6.71984926e-02
5.44162989e-01 -4.10246521e-01 1.13198519e+00 -1.10065150e+00
-5.54834366e-01 -5.84534168e-01 1.83687508e-01 -1.41890794e-01
1.75234511e-01 -1.45043463e-01 -2.65766680e-01 -7.25472629e-01
-5.01349121e-02 8.74478277e-03 7.18168974e-01 4.72462267e-01
9.04359519e-01 -3.74252945e-01 -8.03672910e-01 1.55099332e-01
1.60966158e+00 9.56049263e-01 6.28209531e-01 6.40985489e-01
-1.59445610e-02 3.66783172e-01 1.10563660e+00 5.82335413e-01
-2.62101907e-02 1.22310400e+00 -5.43232635e-03 1.79188177e-01
2.47493863e-01 -3.81507650e-02 -3.79517935e-02 3.08936566e-01
1.95137225e-02 -7.24530220e-01 -9.36733007e-01 4.53169793e-01
-1.65706921e+00 -8.38235557e-01 6.70559645e-01 2.86333847e+00
8.09079647e-01 2.97955483e-01 4.67248023e-01 1.79624364e-01
5.94563842e-01 -3.32645267e-01 -4.09771204e-01 -6.09258652e-01
3.04800510e-01 9.05653387e-02 4.57945228e-01 6.78630888e-01
-5.94154537e-01 7.09228992e-01 6.92253876e+00 9.97218072e-01
-1.08407724e+00 -4.15892184e-01 5.35030484e-01 -2.02727437e-01
-5.41483104e-01 3.79158556e-01 -1.04501438e+00 5.27238846e-01
1.17070746e+00 -6.94604456e-01 4.14856642e-01 7.09331393e-01
1.89015925e-01 -6.04271710e-01 -8.63227725e-01 7.29781449e-01
-5.89236096e-02 -9.35340941e-01 3.77228022e-01 3.01394910e-01
5.36799014e-01 -3.71121794e-01 -1.43957555e-01 1.94515169e-01
2.27085233e-01 -6.90144062e-01 1.87663659e-01 4.37627286e-01
6.84537053e-01 -9.95452881e-01 8.19487870e-01 5.92656851e-01
-1.01936376e+00 -3.56917262e-01 -2.88981348e-01 3.60217363e-01
1.05475567e-01 3.65734428e-01 -1.17166638e+00 6.21068299e-01
6.43962324e-01 -3.66338789e-01 -5.14983296e-01 1.49988806e+00
2.89606482e-01 4.57652092e-01 -6.68477654e-01 -5.51129460e-01
1.15591977e-02 4.59187180e-02 5.53214610e-01 1.11292708e+00
6.09808207e-01 1.94296271e-01 2.46953249e-01 4.01051134e-01
6.39817178e-01 6.93240106e-01 -3.39281529e-01 2.00507119e-01
1.01838613e+00 8.66160214e-01 -8.34889293e-01 -2.88995981e-01
1.15933657e-01 5.42657077e-01 9.79702082e-03 4.07304347e-01
-2.98688054e-01 -6.40220225e-01 -1.18975304e-01 2.51538932e-01
4.86918569e-01 1.86961874e-01 1.08885214e-01 -5.37321329e-01
6.81541488e-02 -1.01567566e+00 8.60481799e-01 -2.63830811e-01
-7.40168273e-01 9.48330760e-01 5.83171785e-01 -1.31614232e+00
-9.79464293e-01 -3.11454013e-02 -2.86285639e-01 1.05568838e+00
-1.21918690e+00 -4.10670340e-01 6.73906282e-02 4.12969708e-01
3.32041353e-01 -3.02534014e-01 9.95563626e-01 4.66938913e-02
-2.60212988e-01 8.24450731e-01 2.27596253e-01 -6.34464622e-01
6.79649293e-01 -9.77613211e-01 -2.51011759e-01 4.96089309e-01
1.59978583e-01 9.03238475e-01 7.94739485e-01 -6.56603694e-01
-1.09390664e+00 -3.85239363e-01 1.24987316e+00 -1.76607117e-01
5.57905808e-02 2.41067126e-01 -9.91012096e-01 -2.48452332e-02
4.18505222e-02 -5.29240191e-01 6.90490186e-01 3.17995012e-01
3.99141721e-02 -3.23683619e-01 -1.24312747e+00 6.45816505e-01
5.61960042e-01 -1.97124630e-01 -4.88066077e-01 2.26404414e-01
6.11162424e-01 -1.24411628e-01 -9.44069982e-01 5.08222580e-01
6.76875889e-01 -1.06786728e+00 9.60183442e-01 -4.87000309e-02
-4.99794066e-01 -2.96256065e-01 -9.45543125e-02 -1.28255033e+00
-6.13040507e-01 -8.53647113e-01 2.01083720e-01 9.54168379e-01
8.13793719e-01 -6.19795263e-01 5.58796763e-01 7.84464419e-01
3.81679535e-01 -1.03374541e+00 -7.14648664e-01 -7.87225723e-01
-2.21714556e-01 1.64457299e-02 8.19498837e-01 4.14930254e-01
4.51341756e-02 2.85064876e-01 6.38254806e-02 -7.50796720e-02
1.11814424e-01 3.61619860e-01 5.40082157e-01 -1.16304576e+00
-5.40634573e-01 -6.74138129e-01 -1.42805308e-01 -9.03531313e-01
-4.37243462e-01 -5.26141703e-01 3.46520022e-02 -1.34460723e+00
3.19899023e-01 -8.79779398e-01 -7.24241912e-01 3.63429427e-01
-3.35145772e-01 -1.67923689e-01 -3.02066971e-02 5.84124625e-01
-5.31462729e-01 7.48350769e-02 7.72930384e-01 1.18958518e-01
-1.10156357e+00 5.07774830e-01 -7.87134707e-01 1.29455462e-01
6.39269352e-01 -5.25824010e-01 -6.58967614e-01 1.40955076e-01
3.79732937e-01 3.24873477e-01 -3.57235223e-01 -8.61798823e-01
4.60112333e-01 -3.95772249e-01 2.11065322e-01 -7.34492004e-01
2.31106989e-02 -8.38286340e-01 4.42393959e-01 6.70613170e-01
-6.02207661e-01 2.43298218e-01 3.16483751e-02 4.48096663e-01
-2.42462203e-01 -6.63577497e-01 4.41670209e-01 -4.90547866e-02
-5.76929808e-01 1.98072232e-02 -2.87780225e-01 -3.23732138e-01
1.00693202e+00 -4.55345184e-01 1.87385812e-01 -5.40486217e-01
-5.94264925e-01 3.61929327e-01 4.76108074e-01 1.35159478e-01
3.71671617e-01 -1.08283448e+00 -5.18818498e-01 1.28301485e-02
3.98200005e-01 -4.87681776e-01 -1.19425356e-01 4.20829296e-01
-3.66964042e-01 8.56533170e-01 1.11856118e-01 -5.58707595e-01
-1.49598801e+00 7.23974407e-01 2.02593252e-01 -7.42038965e-01
-1.24281526e-01 5.91144860e-01 -4.07076329e-01 5.36793731e-02
3.84202987e-01 9.41894129e-02 -6.16996527e-01 1.45114094e-01
5.14436781e-01 1.46147087e-01 4.55484211e-01 -2.36578330e-01
-4.65808719e-01 7.04997182e-01 -4.11835223e-01 -5.55563569e-01
9.59711730e-01 -2.12740213e-01 -5.63745461e-02 1.28703907e-01
9.62467194e-01 1.65666997e-01 -4.56371278e-01 -3.55138004e-01
2.62938976e-01 -6.44661307e-01 2.22113028e-01 -1.02830994e+00
-6.01356983e-01 4.33941111e-02 7.70474136e-01 6.29905462e-01
1.65977752e+00 -6.67257085e-02 2.33126223e-01 8.72209251e-01
6.30718589e-01 -1.17760110e+00 -2.93837078e-02 2.33533055e-01
8.30634058e-01 -9.15119767e-01 3.57097536e-01 -1.18989885e-01
-4.76151198e-01 9.47983921e-01 4.41985935e-01 2.41271719e-01
7.15120852e-01 -6.65285811e-02 -4.65410873e-02 -1.54426262e-01
-9.43043768e-01 -3.07600349e-01 6.01268888e-01 1.17655404e-01
4.65537518e-01 -8.17650706e-02 -1.00830960e+00 1.07036240e-01
-2.28531376e-01 -1.55585324e-02 -8.34422782e-02 1.06631446e+00
-7.30611384e-01 -1.74016392e+00 -5.15269220e-01 6.56478763e-01
-2.71311313e-01 -3.27713676e-02 -4.34615314e-01 6.13198996e-01
-1.36964560e-01 1.19663012e+00 -8.19239542e-02 -5.29439330e-01
5.88178098e-01 1.38886631e-01 3.83477718e-01 -6.13936722e-01
-8.30634117e-01 4.00118947e-01 2.01667294e-01 -4.58818465e-01
-1.52597591e-01 -5.73677063e-01 -9.65223372e-01 2.75377501e-02
-9.62567151e-01 7.74982214e-01 6.06900990e-01 7.94061244e-01
5.67147374e-01 2.20035478e-01 1.02445364e+00 -4.95638490e-01
-9.19072807e-01 -8.27149987e-01 -3.36010665e-01 1.75373748e-01
-9.95144844e-02 -7.33679652e-01 -4.07764912e-01 -5.36312580e-01]
|
[11.491726875305176, 7.53472900390625]
|
2c9af49b-d266-429e-ad3e-c68d29597d28
|
using-program-induction-to-interpret
|
1708.00376
| null |
http://arxiv.org/abs/1708.00376v1
|
http://arxiv.org/pdf/1708.00376v1.pdf
|
Using Program Induction to Interpret Transition System Dynamics
|
Explaining and reasoning about processes which underlie observed black-box
phenomena enables the discovery of causal mechanisms, derivation of suitable
abstract representations and the formulation of more robust predictions. We
propose to learn high level functional programs in order to represent abstract
models which capture the invariant structure in the observed data. We introduce
the $\pi$-machine (program-induction machine) -- an architecture able to induce
interpretable LISP-like programs from observed data traces. We propose an
optimisation procedure for program learning based on backpropagation, gradient
descent and A* search. We apply the proposed method to two problems: system
identification of dynamical systems and explaining the behaviour of a DQN
agent. Our results show that the $\pi$-machine can efficiently induce
interpretable programs from individual data traces.
|
['Svetlin Penkov', 'Subramanian Ramamoorthy']
|
2017-07-26
| null | null | null | null |
['program-induction']
|
['computer-code']
|
[ 4.11644071e-01 4.57976818e-01 -2.80211926e-01 -5.75557768e-01
-5.58420680e-02 -3.69864367e-02 8.58326852e-01 4.39337134e-01
1.13747269e-01 6.48021579e-01 -9.55212042e-02 -9.95655715e-01
-4.63061601e-01 -8.90379190e-01 -1.13028550e+00 -3.04234326e-01
-7.04367220e-01 5.81336319e-01 4.17730547e-02 -1.59635425e-01
6.00082815e-01 5.73330462e-01 -1.80105519e+00 4.90747482e-01
8.35985184e-01 3.19990128e-01 2.39084899e-01 1.03698254e+00
-1.24758959e-01 1.45854247e+00 -2.88158447e-01 -7.09751099e-02
-2.18276709e-01 -7.33590543e-01 -9.04502392e-01 -8.76483545e-02
4.61211381e-03 1.08263254e-01 -1.89374819e-01 1.08191192e+00
-4.25554782e-01 -2.85106242e-01 7.05853105e-01 -1.42273474e+00
-7.57421076e-01 8.23988497e-01 5.90111949e-02 1.41780615e-01
5.46969831e-01 3.65278363e-01 1.05100262e+00 -3.65045756e-01
5.62955499e-01 1.47554076e+00 4.94708180e-01 6.80119872e-01
-2.10814667e+00 -2.07127914e-01 -2.14508295e-01 5.68396039e-02
-8.79096329e-01 -1.20394729e-01 5.40050685e-01 -8.69114161e-01
1.24590206e+00 2.70709902e-01 6.26259327e-01 8.78663957e-01
6.25789881e-01 5.96138120e-01 1.18528533e+00 -5.38842738e-01
3.98075491e-01 2.82263786e-01 7.57604778e-01 1.48014235e+00
3.35329056e-01 8.19778323e-01 -5.22149563e-01 -6.54490173e-01
8.68312478e-01 -1.26325667e-01 -3.75825316e-02 -3.34949046e-01
-8.95957053e-01 9.18256164e-01 1.29585013e-01 4.37640548e-02
-3.95948738e-01 9.55277264e-01 4.54970539e-01 6.03222132e-01
-7.80481100e-02 5.77725112e-01 -5.85228622e-01 -5.84055856e-02
-4.13569778e-01 5.49108148e-01 1.32059586e+00 8.15646410e-01
9.21740949e-01 4.43335235e-01 2.07770824e-01 -1.53250722e-02
6.89855754e-01 4.42924500e-01 3.70818764e-01 -9.36161697e-01
2.01427862e-01 1.09215152e+00 -2.03823652e-02 -7.47791469e-01
-2.48698175e-01 -2.20091548e-02 -4.06507313e-01 5.80500603e-01
1.16655543e-01 -1.64291218e-01 -6.69938445e-01 1.67424142e+00
-1.83206677e-01 7.26496950e-02 1.63614303e-01 2.98229694e-01
2.96448946e-01 9.93640780e-01 2.95655936e-01 -2.62148887e-01
9.55331147e-01 -4.29481179e-01 -3.26482266e-01 -1.35355115e-01
8.08951676e-01 1.96559310e-01 1.04443896e+00 2.24266872e-01
-9.96974051e-01 -7.72610545e-01 -1.03235757e+00 4.93643612e-01
-1.56624451e-01 2.44492088e-02 1.03215885e+00 5.02418756e-01
-1.09509254e+00 1.09929883e+00 -9.43969846e-01 -1.82224244e-01
1.60008371e-01 8.05218637e-01 -1.89546928e-01 6.12420201e-01
-7.65103400e-01 8.41766715e-01 8.32943797e-01 8.54536518e-03
-1.59182966e+00 -5.82691848e-01 -9.47061300e-01 2.66416490e-01
1.64101459e-02 -6.83688819e-01 1.17451048e+00 -1.24893010e+00
-1.50702584e+00 9.00948644e-01 -3.73165786e-01 -9.15837348e-01
-9.61424410e-02 3.14571947e-01 -3.91229600e-01 -1.43204227e-01
-3.46154630e-01 2.14765906e-01 9.40730810e-01 -1.13397467e+00
-3.38038951e-01 -2.92905658e-01 -8.39684531e-02 -5.99964499e-01
4.12798911e-01 -3.07368208e-02 6.04147255e-01 -2.20040917e-01
-6.59214482e-02 -9.76653874e-01 -4.78068739e-01 -4.40939903e-01
-4.04341072e-01 -3.83957982e-01 5.35801589e-01 -5.21234214e-01
1.13145697e+00 -1.94143796e+00 5.05292952e-01 5.14234185e-01
2.50777036e-01 -1.11078598e-01 7.75843859e-02 6.01756692e-01
-5.08398533e-01 1.70606077e-01 -4.24717188e-01 2.67008781e-01
3.52652550e-01 6.15210295e-01 -7.50495017e-01 2.27632090e-01
3.62135798e-01 9.26185966e-01 -5.21960676e-01 -2.11630374e-01
1.95416734e-01 -2.02295914e-01 -7.46394217e-01 5.88290036e-01
-1.10619903e+00 3.59218806e-01 -6.74709439e-01 7.69355744e-02
-3.24903652e-02 -1.67691439e-01 3.45961183e-01 5.27399600e-01
-2.09123448e-01 3.87036592e-01 -9.33013141e-01 1.20422506e+00
-4.35021549e-01 7.38955915e-01 -4.37752217e-01 -1.41727829e+00
1.26958323e+00 3.35014015e-01 -7.36452714e-02 -4.88185734e-01
3.08324210e-02 2.20159024e-01 1.94951102e-01 -9.21567142e-01
-1.08625762e-01 -2.26499245e-01 -3.22610408e-01 6.70411706e-01
2.35976696e-01 -2.82239735e-01 1.63133621e-01 -2.10930333e-02
1.31109965e+00 3.51788253e-01 4.33881342e-01 -7.46635854e-01
9.92511332e-01 3.33321452e-01 3.74622017e-01 1.09253466e+00
2.28688300e-01 -2.69795746e-01 1.20243847e+00 -1.02938628e+00
-1.08639300e+00 -9.98203039e-01 1.70697629e-01 1.11396444e+00
-1.70061842e-01 -3.09147507e-01 -7.91628301e-01 -3.19934130e-01
-1.54270949e-02 1.22319877e+00 -7.09822714e-01 -3.11603010e-01
-7.99432099e-01 -8.09430897e-01 6.46399558e-01 2.53502429e-01
1.33996695e-01 -1.57209826e+00 -8.10995102e-01 3.83398384e-01
4.52276975e-01 -3.78306240e-01 3.37877691e-01 8.36656570e-01
-1.38289309e+00 -1.05690444e+00 4.57224220e-01 -8.39961648e-01
9.26748931e-01 -8.30732584e-01 1.15138853e+00 2.99503088e-01
-4.56512123e-01 2.10173890e-01 1.84305847e-01 -3.85776579e-01
-1.30671263e+00 -2.87106782e-01 -1.40778631e-01 -2.75509953e-01
3.52391660e-01 -8.09745908e-01 2.38750339e-01 -3.11962962e-01
-8.51899743e-01 1.43611133e-01 4.76918846e-01 8.78114700e-01
2.22045541e-01 1.92071021e-01 2.37767279e-01 -1.13338375e+00
8.78548324e-01 -3.84090275e-01 -1.18835580e+00 3.24443638e-01
-8.40232193e-01 1.22400558e+00 9.24026132e-01 -2.95233041e-01
-1.21966326e+00 2.82361031e-01 7.21535608e-02 1.54581100e-01
-6.12647235e-01 6.04323745e-01 -9.55217704e-02 -5.55241033e-02
1.08265007e+00 5.62945187e-01 -1.67120337e-01 -1.77478492e-01
2.74158806e-01 2.55256593e-01 4.94721025e-01 -1.12876272e+00
6.49230301e-01 9.64113101e-02 3.28095883e-01 -5.75035393e-01
-2.64793575e-01 2.19835818e-01 -5.00914931e-01 1.32662982e-01
5.91916144e-01 -5.32790363e-01 -1.06107378e+00 3.34423222e-02
-1.46676660e+00 -6.89476967e-01 -3.36997271e-01 1.64170802e-01
-1.11783922e+00 -1.98095843e-01 -7.48756289e-01 -8.16423059e-01
-1.71379566e-01 -1.14126050e+00 7.23962545e-01 9.96443853e-02
-6.49013638e-01 -1.23506379e+00 7.78891861e-01 -1.98632121e-01
-1.17616326e-01 2.63500124e-01 1.75704241e+00 -6.73590660e-01
-8.44210088e-01 -1.56674176e-01 1.50539652e-01 4.22530398e-02
-4.37889278e-01 3.45517904e-01 -7.52442360e-01 2.47974113e-01
1.34752214e-01 5.53785544e-03 3.39767218e-01 2.33702481e-01
1.22445464e+00 -6.30916059e-01 -3.35574687e-01 4.31030571e-01
1.57409298e+00 4.66069102e-01 5.68636954e-01 1.60035565e-01
3.09279710e-01 6.40850782e-01 -1.27761304e-01 2.72761375e-01
2.04338296e-03 2.71091372e-01 4.19441283e-01 5.13907194e-01
4.35701638e-01 -4.63614643e-01 8.86396408e-01 5.05045533e-01
-2.29528204e-01 4.00726229e-01 -1.24974751e+00 4.33881432e-01
-2.15030980e+00 -1.24251699e+00 -5.81609190e-01 1.93737948e+00
8.41318905e-01 5.14680743e-01 -7.91245978e-03 2.80618970e-03
6.61249757e-01 -2.02950656e-01 -3.33455294e-01 -1.27616858e+00
3.08045715e-01 7.15396821e-01 1.08599000e-01 8.22766840e-01
-5.47733724e-01 6.01311922e-01 6.93444204e+00 2.61852086e-01
-7.95334458e-01 -2.47361083e-02 2.43964761e-01 6.34808064e-01
-4.35556769e-01 4.23021138e-01 -6.80981696e-01 2.17860937e-01
1.81903231e+00 -4.86005813e-01 7.85215020e-01 9.50167418e-01
2.82596141e-01 4.89373393e-02 -1.89861119e+00 5.39123893e-01
-3.22231114e-01 -1.59747350e+00 2.13091642e-01 -6.77703917e-02
5.84191680e-01 -1.08922914e-01 -3.00604433e-01 4.57675606e-01
8.35257828e-01 -1.28121114e+00 8.79786611e-01 7.55811810e-01
-3.83239146e-03 -6.38549984e-01 2.57480800e-01 7.59389877e-01
-8.40519190e-01 -4.17415082e-01 -2.65441328e-01 -5.75732827e-01
-3.63575429e-01 4.71993059e-01 -7.39638150e-01 1.32857695e-01
1.06777385e-01 3.38607699e-01 -5.08519292e-01 5.64970732e-01
-6.22546375e-01 7.71777451e-01 -7.92126078e-03 -4.91579622e-01
3.39687355e-02 -3.04599077e-01 4.36109900e-01 1.32467556e+00
1.43838972e-01 7.39521161e-02 -1.53302506e-01 1.81670892e+00
3.17760915e-01 -4.15077925e-01 -1.01854241e+00 -2.94263810e-01
-9.23631191e-02 7.04773307e-01 -5.11972547e-01 -3.64693552e-01
-2.35140771e-01 5.08732736e-01 3.89626801e-01 2.13065401e-01
-7.25537241e-01 -3.08344156e-01 4.43141609e-01 -7.25842721e-04
1.14002928e-01 -3.28982919e-01 -4.20721114e-01 -1.14237785e+00
-3.74390334e-01 -1.07638323e+00 2.40958974e-01 -8.06796372e-01
-7.30399191e-01 5.44451535e-01 1.43792003e-01 -3.50053757e-01
-9.05212879e-01 -8.17491293e-01 -9.78088379e-01 8.03906977e-01
-7.96245217e-01 -6.44456625e-01 7.98525661e-02 6.06368840e-01
2.37643719e-01 -3.07348907e-01 1.18461800e+00 -4.06569034e-01
-5.23772180e-01 -2.70543098e-01 -7.72867203e-02 -1.40601201e-02
-3.67128968e-01 -1.45694590e+00 3.99168402e-01 7.42780387e-01
3.46611738e-01 8.26970935e-01 1.29355168e+00 -5.93680203e-01
-1.72717214e+00 -9.46123362e-01 8.70078385e-01 -5.60530245e-01
1.00247419e+00 -3.03092182e-01 -9.08575594e-01 1.16596854e+00
1.76579133e-01 -2.54365087e-01 4.41484183e-01 9.07064881e-03
-3.32456946e-01 -2.08520383e-01 -7.56450295e-01 6.71813786e-01
7.45403409e-01 -9.28302526e-01 -1.18159330e+00 8.76376629e-02
6.03707194e-01 5.57297766e-02 -7.46123910e-01 -5.43880835e-03
1.85327724e-01 -9.02760923e-01 7.20049202e-01 -1.04804659e+00
7.53183842e-01 -2.97741771e-01 1.14609197e-01 -1.15789247e+00
-8.55671093e-02 -8.51983547e-01 -4.64634567e-01 7.50781298e-01
5.11641443e-01 -6.68573916e-01 6.40381753e-01 8.55763674e-01
1.76393420e-01 -2.04541028e-01 -5.17674625e-01 -4.35522050e-01
-1.06291205e-01 -6.33390963e-01 4.82193947e-01 4.90121156e-01
6.05745256e-01 3.16825032e-01 6.45101741e-02 4.43276405e-01
9.32071924e-01 6.32630527e-01 5.45393050e-01 -1.30708098e+00
-8.73061061e-01 -4.18609321e-01 -4.89046782e-01 -5.49249947e-01
7.78345883e-01 -9.94858265e-01 1.70797348e-01 -7.60995209e-01
1.38483003e-01 -2.01015741e-01 6.81340182e-03 3.44209790e-01
1.94152355e-01 -6.57424629e-01 -8.19551945e-02 2.05609351e-01
-2.78634459e-01 3.18471879e-01 5.68302453e-01 -7.67119899e-02
-1.95644647e-01 9.05722007e-02 -3.84804338e-01 9.93274570e-01
9.22539234e-01 -8.95292640e-01 -4.49583322e-01 -1.44115612e-01
7.01219618e-01 7.97812164e-01 1.04495001e+00 -8.66731644e-01
2.34538481e-01 -5.13272524e-01 -1.50889549e-02 -1.33900538e-01
-4.17496502e-01 -7.84788668e-01 5.41766763e-01 1.17619944e+00
-9.04145062e-01 2.29514688e-01 1.77764028e-01 4.98132259e-01
-2.06549481e-01 -9.11669970e-01 4.09752905e-01 -2.84766883e-01
-8.81292880e-01 -7.50218406e-02 -8.59772027e-01 -2.11317182e-01
9.00438428e-01 3.60253930e-01 -2.48178065e-01 -1.06398039e-01
-9.08894956e-01 -8.22302252e-02 4.73002046e-02 3.39029878e-02
5.13509929e-01 -9.47129667e-01 -4.28803056e-01 4.84439790e-01
-1.04755886e-01 -4.86757159e-01 -3.74148011e-01 4.91021067e-01
-1.01949704e+00 6.55218899e-01 -3.95365238e-01 -4.31123912e-01
-1.02176893e+00 5.42960227e-01 7.04710245e-01 -3.63359332e-01
-3.77780348e-01 3.00412536e-01 9.71105620e-02 -7.65290678e-01
-7.06712529e-02 -6.84544086e-01 -3.81386019e-02 -8.48606467e-01
4.00819212e-01 1.72701553e-01 -4.34357643e-01 -2.75489129e-03
-5.72900772e-02 8.98925439e-02 2.96719909e-01 -1.47231624e-01
1.58733904e+00 4.11130488e-01 -8.26320231e-01 6.68517590e-01
8.75715613e-01 -6.43810987e-01 -1.14299214e+00 7.89699182e-02
7.79702425e-01 8.66740122e-02 -2.21418813e-01 -4.47872728e-01
-3.44031036e-01 1.06054246e+00 4.13078010e-01 8.24245393e-01
7.91777134e-01 1.02782264e-01 -7.34701753e-02 9.70858812e-01
4.62237388e-01 -5.81822753e-01 -8.48653316e-02 6.01433933e-01
7.23646879e-01 -8.55825305e-01 -2.47004285e-01 -4.45417687e-02
-7.98562616e-02 1.67301381e+00 4.73764539e-01 -6.12755179e-01
4.76731300e-01 6.12498641e-01 -5.94189465e-01 -6.14806712e-01
-1.27173102e+00 5.66267371e-02 4.56092544e-02 5.50468266e-01
2.16473296e-01 1.62438616e-01 -4.31645513e-02 5.03746510e-01
-7.18258694e-02 3.50859672e-01 7.40072191e-01 8.30005884e-01
-6.01169705e-01 -1.05843306e+00 -4.98298973e-01 4.34525788e-01
-1.82439074e-01 1.29076047e-03 -2.13978633e-01 8.12009990e-01
1.26436278e-01 4.39626366e-01 3.86815630e-02 -2.84986287e-01
2.68990576e-01 4.02818561e-01 7.42905557e-01 -6.49588943e-01
-5.69616079e-01 -5.26229978e-01 1.74133345e-01 -5.84233701e-01
-4.69519407e-01 -7.14377105e-01 -1.55866265e+00 -6.14653289e-01
2.09861025e-01 2.78029382e-01 5.33756018e-01 1.07406640e+00
1.53353810e-01 6.18491471e-01 4.05372053e-01 -5.55744708e-01
-6.22579634e-01 -4.88980651e-01 -5.53112864e-01 2.66886294e-01
4.25240904e-01 -7.05346763e-02 -3.32414955e-01 6.35023177e-01]
|
[8.423381805419922, 7.255739212036133]
|
a8cf2194-e51e-4604-b251-af95de1344e3
|
sharcs-shared-concept-space-for-explainable
|
2307.00316
| null |
https://arxiv.org/abs/2307.00316v1
|
https://arxiv.org/pdf/2307.00316v1.pdf
|
SHARCS: Shared Concept Space for Explainable Multimodal Learning
|
Multimodal learning is an essential paradigm for addressing complex real-world problems, where individual data modalities are typically insufficient to accurately solve a given modelling task. While various deep learning approaches have successfully addressed these challenges, their reasoning process is often opaque; limiting the capabilities for a principled explainable cross-modal analysis and any domain-expert intervention. In this paper, we introduce SHARCS (SHARed Concept Space) -- a novel concept-based approach for explainable multimodal learning. SHARCS learns and maps interpretable concepts from different heterogeneous modalities into a single unified concept-manifold, which leads to an intuitive projection of semantically similar cross-modal concepts. We demonstrate that such an approach can lead to inherently explainable task predictions while also improving downstream predictive performance. Moreover, we show that SHARCS can operate and significantly outperform other approaches in practically significant scenarios, such as retrieval of missing modalities and cross-modal explanations. Our approach is model-agnostic and easily applicable to different types (and number) of modalities, thus advancing the development of effective, interpretable, and trustworthy multimodal approaches.
|
['Nikola Simidjievski', 'Pietro Liò', 'Lucie Charlotte Magister', 'Pietro Barbiero', 'Gabriele Dominici']
|
2023-07-01
| null | null | null | null |
['retrieval']
|
['methodology']
|
[ 3.00722033e-01 3.86158377e-01 -3.64308834e-01 -5.15597820e-01
-1.16200852e+00 -6.55503929e-01 7.25277841e-01 2.60344237e-01
1.12156853e-01 6.10630572e-01 4.16475415e-01 -3.40095162e-01
-4.31556165e-01 -3.98127884e-01 -6.94257081e-01 -4.86055702e-01
1.61696076e-01 7.81555176e-01 -3.99811655e-01 -1.59245238e-01
-1.51345134e-01 1.27089277e-01 -1.61356425e+00 8.99759293e-01
8.40944111e-01 8.85899305e-01 -6.52301982e-02 5.20385683e-01
-3.52260768e-01 6.89886808e-01 -1.09766357e-01 -7.93828130e-01
-1.33719176e-01 -5.27451038e-01 -1.14214230e+00 -8.26955959e-02
5.27670622e-01 -4.21144404e-02 9.08183455e-02 8.53145242e-01
2.50313040e-02 3.15163806e-02 8.06612492e-01 -1.66132104e+00
-8.80683959e-01 7.23969758e-01 -2.70886809e-01 -4.24774021e-01
5.38417876e-01 -1.02894269e-01 1.33027589e+00 -8.92931461e-01
5.27783394e-01 1.59964538e+00 4.75619853e-01 9.65822577e-01
-1.43689239e+00 -3.83334726e-01 2.79164732e-01 2.82124728e-01
-1.02617586e+00 -3.26345265e-01 5.72187960e-01 -4.19404894e-01
5.36442697e-01 6.67400658e-01 4.41835314e-01 1.27912998e+00
-1.04449980e-01 9.81309831e-01 9.55691993e-01 -4.76756990e-01
2.15750203e-01 2.03347445e-01 1.01106793e-01 6.84070706e-01
5.63399568e-02 -1.42376229e-01 -9.99868453e-01 -3.00393607e-02
3.88112843e-01 3.97000581e-01 -2.95635104e-01 -6.35170043e-01
-1.76471972e+00 8.26020241e-01 6.12443209e-01 1.92253739e-01
-2.48446196e-01 2.21886545e-01 2.15953723e-01 2.24112287e-01
3.83519500e-01 5.90645969e-01 -6.28179967e-01 1.15911551e-01
-6.85600340e-01 1.29116446e-01 6.76083744e-01 9.24889743e-01
7.51437664e-01 -1.54966891e-01 7.15980083e-02 4.34505880e-01
4.69025433e-01 6.96259439e-01 1.47826210e-01 -1.13042450e+00
5.47051609e-01 9.41637516e-01 -2.69870032e-02 -9.83654320e-01
-5.56319892e-01 -5.19939214e-02 -9.76050854e-01 -1.63674485e-02
3.76066387e-01 -3.25356759e-02 -7.60099947e-01 1.93170249e+00
2.80977756e-01 -8.67034867e-02 4.02505428e-01 1.11038291e+00
9.24887061e-01 4.20752287e-01 5.24014533e-01 2.19091550e-01
1.48172343e+00 -8.26308012e-01 -6.48401678e-01 -3.67137164e-01
5.96622407e-01 -4.85795915e-01 1.26970553e+00 2.13492513e-01
-9.03596640e-01 -3.15085173e-01 -6.79361701e-01 -3.50052863e-01
-5.11879861e-01 -1.76018596e-01 1.00005472e+00 3.41806293e-01
-8.95678163e-01 2.69824833e-01 -7.16177404e-01 -5.10331333e-01
5.58970034e-01 3.41862947e-01 -9.96598601e-01 -3.59658748e-01
-9.77566302e-01 1.02554655e+00 5.91075718e-01 9.77365002e-02
-7.92542219e-01 -8.73239696e-01 -9.29757774e-01 3.30380917e-01
4.50100183e-01 -1.21076870e+00 1.03183639e+00 -1.27088523e+00
-9.90815461e-01 8.29148591e-01 -4.77366239e-01 -2.97944456e-01
3.55338126e-01 -3.35054815e-01 -3.52928519e-01 2.79970437e-01
-2.07053106e-02 1.01776254e+00 9.92774248e-01 -1.72546113e+00
-4.04594809e-01 -4.59665716e-01 2.98334241e-01 4.33317870e-01
-4.95172977e-01 -2.99240977e-01 -3.35175306e-01 -2.35014677e-01
3.46161753e-01 -9.89187896e-01 -7.52746612e-02 1.77310005e-01
-7.71279335e-01 7.69402385e-02 6.57270253e-01 -5.33254683e-01
7.13930309e-01 -2.06960726e+00 9.21552896e-01 1.27118975e-01
7.34670162e-01 -1.12433113e-01 -2.35166281e-01 4.37965453e-01
-2.02190056e-01 4.36695695e-01 -5.26047945e-01 -7.39719987e-01
3.59376431e-01 4.04178500e-01 -4.58473355e-01 8.42708945e-02
4.28017288e-01 1.20866978e+00 -9.12468314e-01 -4.12072957e-01
3.74640942e-01 7.42050707e-01 -4.41768229e-01 2.28379384e-01
-4.13200051e-01 8.03459167e-01 -4.81376588e-01 7.82614768e-01
3.34247261e-01 -5.09645283e-01 3.94765735e-01 -3.30950826e-01
5.36514401e-01 -3.47339392e-01 -8.09264719e-01 1.97401106e+00
-6.18900657e-01 8.10823917e-01 -2.25497484e-01 -1.01868045e+00
6.30960464e-01 4.09440339e-01 3.51762682e-01 -4.75377351e-01
-6.98058456e-02 1.14726499e-01 -3.05227757e-01 -6.15831733e-01
3.15446883e-01 -6.61283731e-01 -1.23140790e-01 5.14667571e-01
2.26508290e-01 -2.13816501e-02 -3.35014194e-01 4.97889876e-01
6.43493712e-01 6.45735040e-02 2.38962099e-01 -3.92122529e-02
5.04315376e-01 8.69723707e-02 1.12367541e-01 6.61171198e-01
6.71691298e-02 8.05338979e-01 4.61211115e-01 -5.75178385e-01
-8.29079926e-01 -1.29140902e+00 6.57136440e-02 1.16123736e+00
3.83356363e-01 -3.97839457e-01 -6.53137505e-01 -7.52560258e-01
2.11939543e-01 7.76022673e-01 -7.56314337e-01 -2.42161900e-01
2.52937544e-02 -3.44437450e-01 4.03553188e-01 4.97080535e-01
9.08181593e-02 -8.17166030e-01 -2.16568574e-01 -2.54484594e-01
-7.82264292e-01 -1.19762146e+00 9.68719646e-03 -3.48600633e-02
-8.93053949e-01 -1.29562426e+00 -4.95502681e-01 -3.26371700e-01
8.35222423e-01 2.85249352e-01 1.33277917e+00 2.91292131e-01
-5.76369688e-02 9.97498393e-01 -2.87448198e-01 -4.65469837e-01
-4.88938004e-01 4.86916788e-02 -4.17452380e-02 3.29602897e-01
3.56925339e-01 -2.74424016e-01 -5.25710523e-01 4.52475622e-02
-1.13139760e+00 4.30779487e-01 4.56604481e-01 8.79942715e-01
5.72844565e-01 -3.19969565e-01 6.72792912e-01 -9.12448406e-01
4.52105522e-01 -6.35328948e-01 -9.11080241e-02 6.14630580e-01
-5.30639887e-01 3.51392269e-01 4.99970555e-01 -2.69704163e-01
-1.24798930e+00 6.78059235e-02 3.09850782e-01 -5.09189367e-01
-4.49764907e-01 8.00224602e-01 -3.73075366e-01 1.34530857e-01
6.13966823e-01 3.14450935e-02 1.90795466e-01 -5.06796241e-01
1.00328600e+00 3.15505832e-01 5.65510571e-01 -5.71451664e-01
9.61434662e-01 7.40671277e-01 2.53181636e-01 -4.28594053e-01
-1.04856527e+00 -2.81673670e-01 -8.91942859e-01 -2.31890649e-01
1.09484553e+00 -1.08678830e+00 -8.87271285e-01 -1.56719595e-01
-1.23065078e+00 4.91061062e-02 -7.02400655e-02 3.46612602e-01
-5.93862474e-01 2.35386699e-01 -3.42699617e-01 -7.31606781e-01
-2.15423241e-01 -9.71950710e-01 1.49360359e+00 9.28903744e-02
-5.77385426e-01 -1.50977194e+00 -1.45759076e-01 9.60738897e-01
2.08997354e-01 4.13317591e-01 1.31460786e+00 -8.08550775e-01
-7.56053567e-01 -3.44352543e-01 -2.72378385e-01 -3.82178389e-02
1.89225394e-02 -1.66407526e-01 -1.28217673e+00 -2.92096846e-02
-4.82244730e-01 -5.80046892e-01 7.87974656e-01 1.78722978e-01
1.16948688e+00 -2.64258921e-01 -3.71621221e-01 2.73548603e-01
1.28632259e+00 -3.69602174e-01 3.91900688e-01 1.39363796e-01
9.11949217e-01 9.91135597e-01 4.32200849e-01 7.22323433e-02
7.39927351e-01 4.89109218e-01 7.60217130e-01 -4.89235193e-01
1.02938794e-01 -2.50667304e-01 7.33931586e-02 5.00065506e-01
-1.12569295e-01 -3.59361142e-01 -1.04024088e+00 4.50319052e-01
-2.29235148e+00 -1.14496648e+00 -1.43670604e-01 1.97790945e+00
5.65409958e-01 -4.68170851e-01 -2.17527404e-01 -1.36197746e-01
5.35222650e-01 -2.74855375e-01 -5.78494728e-01 -2.72805303e-01
-3.29511881e-01 -1.60021737e-01 -2.70823896e-01 5.17908752e-01
-9.66926217e-01 7.63844192e-01 6.62429619e+00 4.27186251e-01
-7.51019657e-01 2.43533567e-01 5.11178613e-01 -1.27317548e-01
-1.01214826e+00 9.03687179e-02 -1.49997577e-01 7.29838908e-02
8.80194366e-01 6.55504987e-02 5.15624762e-01 6.60922408e-01
-6.40298873e-02 1.90140922e-02 -1.75354719e+00 1.11920667e+00
2.39224121e-01 -1.54310596e+00 5.20369411e-01 -8.55726153e-02
6.23844385e-01 -3.53681207e-01 2.60578692e-01 2.47512981e-01
9.38950107e-02 -1.42944872e+00 8.29860568e-01 8.61168146e-01
6.40980601e-01 -6.30444705e-01 6.58758938e-01 2.89045691e-01
-9.22426760e-01 -2.23563135e-01 -1.03272028e-01 1.46806285e-01
1.24433361e-01 3.72747242e-01 -5.49945116e-01 9.68659639e-01
5.76481342e-01 7.40197957e-01 -6.11933172e-01 5.94728231e-01
-3.64128165e-02 2.70942003e-01 -4.16501388e-02 3.18831921e-01
1.80561412e-02 9.29241553e-02 3.98396552e-01 1.04455769e+00
2.57383496e-01 1.76752603e-03 -1.50002211e-01 1.06562793e+00
-1.69812843e-01 -1.33221641e-01 -7.50262916e-01 -1.30621925e-01
3.57884228e-01 1.25141001e+00 -4.56681907e-01 -2.72792846e-01
-5.45363247e-01 1.00073659e+00 4.67441320e-01 6.04127109e-01
-6.98230147e-01 2.17774168e-01 7.67766595e-01 -2.82933444e-01
-1.82913691e-01 -1.99639332e-02 -5.42926192e-01 -1.56431115e+00
-4.38653044e-02 -9.69559014e-01 6.85444415e-01 -1.04065216e+00
-1.60621881e+00 5.52942634e-01 1.20740257e-01 -1.35747540e+00
-2.79023707e-01 -6.71210706e-01 -2.98972011e-01 8.95662904e-01
-1.52396560e+00 -1.79861987e+00 -4.22173411e-01 8.54819596e-01
3.86853129e-01 -2.45856538e-01 1.25922334e+00 -2.07020249e-02
-2.90010214e-01 2.79226750e-01 4.10712548e-02 -3.10147405e-01
8.14206183e-01 -1.40616560e+00 -2.27070317e-01 5.71980596e-01
3.87271106e-01 8.45824838e-01 7.34939873e-01 -3.06388080e-01
-1.76020038e+00 -9.17827666e-01 8.05322230e-01 -1.04999888e+00
7.61431515e-01 -2.88450688e-01 -1.01042008e+00 9.04067755e-01
2.80399591e-01 -3.40342037e-02 1.25357592e+00 6.78089023e-01
-7.90970981e-01 -8.32719505e-02 -9.23245907e-01 7.72513151e-01
9.03797090e-01 -9.95567322e-01 -6.14708722e-01 4.73971099e-01
8.14159036e-01 -2.67266184e-01 -1.01226509e+00 1.59779057e-01
6.26551926e-01 -1.01631033e+00 1.06794202e+00 -1.37815213e+00
8.91972244e-01 -2.44110614e-01 -4.81990337e-01 -1.24578857e+00
-1.98104996e-02 -2.73985416e-01 -3.20178241e-01 1.12290144e+00
7.60889471e-01 -4.37974453e-01 5.35864949e-01 1.32955694e+00
-5.97798713e-02 -4.95521486e-01 -8.73300612e-01 -4.11295831e-01
1.35858208e-01 -7.43087292e-01 7.52175093e-01 1.21632564e+00
4.17269468e-01 3.12339664e-01 -4.90570456e-01 4.31514502e-01
7.68998146e-01 3.50588888e-01 7.45723784e-01 -1.33700728e+00
-1.15574799e-01 -3.04882467e-01 -5.05159557e-01 -5.45541167e-01
5.13740003e-01 -1.06862354e+00 -8.58094469e-02 -1.61882532e+00
6.02565408e-01 -2.52129763e-01 -2.66449124e-01 8.23929906e-01
-3.03805351e-01 2.36787856e-01 5.79372346e-01 3.89876544e-01
-8.80063057e-01 6.10724211e-01 1.18956339e+00 -3.72255057e-01
1.42875612e-01 -2.88199157e-01 -9.87265050e-01 6.61575556e-01
4.95279670e-01 -2.72779763e-01 -6.97627068e-01 -6.99536860e-01
4.21944439e-01 1.29354060e-01 1.04432249e+00 -5.61170876e-01
4.46946993e-02 -3.29531461e-01 3.84288937e-01 -2.03793973e-01
5.28646469e-01 -1.21365905e+00 4.22120392e-01 7.60145783e-02
-7.35921443e-01 -1.64528459e-01 3.81779283e-01 9.77731943e-01
-3.96949321e-01 1.55065656e-01 2.15351179e-01 5.37367202e-02
-8.15980732e-01 2.12143749e-01 -6.53160959e-02 -1.35405257e-01
9.71078455e-01 -6.72193542e-02 -5.32878518e-01 -6.48436129e-01
-1.09153771e+00 3.27787548e-01 2.36158714e-01 6.26467526e-01
7.65560746e-01 -1.49913394e+00 -5.84695399e-01 -1.61874756e-01
5.38198113e-01 -2.31457539e-02 5.94984472e-01 7.47595549e-01
-8.91865045e-02 5.72290421e-01 -1.34921446e-01 -9.21730280e-01
-1.22836804e+00 5.71152151e-01 4.27886546e-01 1.43219486e-01
-5.16704738e-01 5.18842220e-01 5.87669432e-01 -9.13374782e-01
7.91066214e-02 -9.05704871e-02 -9.19569880e-02 -6.10196702e-02
4.88460422e-01 -1.80286970e-02 -2.08901674e-01 -7.41666555e-01
-4.51301992e-01 4.91950363e-01 2.53854483e-01 -1.26618385e-01
1.15424120e+00 -5.19534886e-01 -1.61235288e-01 7.45224297e-01
9.61754560e-01 -4.33288574e-01 -1.15797806e+00 -2.59714633e-01
1.38865039e-02 -3.12575728e-01 -2.51889139e-01 -1.12945187e+00
-9.33065236e-01 1.20745838e+00 5.12730598e-01 2.53132492e-01
1.13745666e+00 5.01849532e-01 4.51948881e-01 4.86811519e-01
2.03878328e-01 -5.02902508e-01 1.27377838e-01 1.06843702e-01
1.03978455e+00 -1.86242580e+00 -1.45721778e-01 -4.81606364e-01
-1.08775878e+00 1.22766387e+00 4.98699427e-01 8.36160123e-01
3.54155838e-01 -3.70855451e-01 2.26299763e-01 -4.02943462e-01
-8.01483691e-01 -6.59647062e-02 8.71911466e-01 8.41762781e-01
4.12410617e-01 2.53899068e-01 5.65772533e-01 7.66818404e-01
7.84249902e-02 -3.05579990e-01 1.35076106e-01 4.73247945e-01
-1.12326331e-01 -9.11869586e-01 -4.98657554e-01 1.71376750e-01
2.13876143e-02 -1.85176536e-01 -6.05196774e-01 9.72737610e-01
2.73907790e-03 1.01909220e+00 6.95033325e-03 -4.02050197e-01
7.24577755e-02 3.52882653e-01 2.99205035e-01 -4.52213317e-01
-1.74939245e-01 -2.46849403e-01 5.02674654e-02 -6.48556054e-01
-7.26719558e-01 -5.54212511e-01 -1.40818286e+00 -5.12789011e-01
-9.55783427e-02 -7.75043443e-02 6.01046801e-01 1.52008057e+00
6.03603244e-01 3.31390321e-01 2.02544987e-01 -8.14159572e-01
-1.52623966e-01 -4.13592696e-01 -3.82642120e-01 7.97190368e-01
5.88982701e-01 -7.13828802e-01 -3.10993940e-01 4.91859913e-01]
|
[10.760170936584473, 1.7241144180297852]
|
72b36da5-1067-4f1b-9360-dbfaee02dc5b
|
new-wrapper-method-based-on-normalized-mutual
|
2210.14346
| null |
https://arxiv.org/abs/2210.14346v1
|
https://arxiv.org/pdf/2210.14346v1.pdf
|
New wrapper method based on normalized mutual information for dimension reduction and classification of hyperspectral images
|
Feature selection is one of the most important problems in hyperspectral images classification. It consists to choose the most informative bands from the entire set of input datasets and discard the noisy, redundant and irrelevant ones. In this context, we propose a new wrapper method based on normalized mutual information (NMI) and error probability (PE) using support vector machine (SVM) to reduce the dimensionality of the used hyperspectral images and increase the classification efficiency. The experiments have been performed on two challenging hyperspectral benchmarks datasets captured by the NASA's Airborne Visible/Infrared Imaging Spectrometer Sensor (AVIRIS). Several metrics had been calculated to evaluate the performance of the proposed algorithm. The obtained results prove that our method can increase the classification performance and provide an accurate thematic map in comparison with other reproduced algorithms. This method may be improved for more classification efficiency. Keywords-Feature selection, hyperspectral images, classification, wrapper, normalized mutual information, support vector machine.
|
['Ahmed Hammouch', 'Elkebir Sarhrouni', 'Asma Elmaizi', 'Hasna Nhaila']
|
2022-10-25
| null | null | null | null |
['classification-of-hyperspectral-images']
|
['computer-vision']
|
[ 7.45979249e-01 -6.85033321e-01 1.43901378e-01 -4.30858761e-01
-2.42794231e-01 -6.13691211e-01 3.51313591e-01 1.21478178e-01
-2.33634815e-01 9.27936792e-01 -1.82858825e-01 -3.75821106e-02
-9.40687001e-01 -9.15614784e-01 5.27060926e-02 -1.01621735e+00
-1.32365003e-01 2.73780767e-02 -1.46171838e-01 -7.41895661e-02
5.47332108e-01 6.20908856e-01 -2.27724719e+00 2.44844154e-01
1.21657336e+00 1.19911790e+00 4.68816549e-01 3.41251045e-01
3.78962904e-02 2.09330693e-01 -7.70418271e-02 4.93525356e-01
4.88907725e-01 -3.25233966e-01 -8.26941490e-01 2.90680408e-01
-4.04496193e-02 1.80436969e-01 1.74159810e-01 1.42227232e+00
4.28587914e-01 5.03477275e-01 8.06994677e-01 -1.13910341e+00
-5.00550270e-02 6.53855920e-01 -7.52791882e-01 2.33364895e-01
-1.33885965e-01 -1.60040140e-01 8.48074734e-01 -7.40795374e-01
3.32639903e-01 7.84543991e-01 4.06391650e-01 -1.22330867e-01
-1.01673257e+00 -7.39761829e-01 -3.16173196e-01 5.53232074e-01
-1.35127258e+00 -5.33770472e-02 6.93387508e-01 -6.14288628e-01
5.59950769e-01 9.38241303e-01 6.55958831e-01 7.76123479e-02
2.99307052e-02 3.03469747e-01 1.42037618e+00 -5.93280137e-01
2.34607264e-01 3.44555140e-01 7.86567211e-01 5.16463459e-01
6.39805257e-01 2.32695460e-01 -1.67226508e-01 -3.86261374e-01
-1.75325751e-01 1.55583009e-01 -6.52844489e-01 -1.94422439e-01
-8.98832023e-01 8.62733901e-01 3.82152766e-01 4.81684923e-01
-4.96243596e-01 -7.75125980e-01 3.92951488e-01 4.25011873e-01
3.99154276e-01 5.71793437e-01 -6.47908270e-01 3.46081883e-01
-8.66125762e-01 -1.36630580e-01 4.65388179e-01 3.11910868e-01
1.12126398e+00 -1.44404247e-01 -3.93569432e-02 9.04692173e-01
4.12220716e-01 7.51651824e-01 8.09262514e-01 -2.58995503e-01
1.32286415e-01 1.04247010e+00 1.83943227e-01 -1.30359209e+00
-6.60989463e-01 -6.30126894e-01 -6.19171917e-01 5.69865525e-01
-3.86961460e-01 -3.31234902e-01 -7.90068924e-01 9.21160519e-01
2.71312237e-01 -1.10262312e-01 2.24668071e-01 1.03408003e+00
7.77669907e-01 9.13180649e-01 -8.51383712e-03 -3.85452420e-01
9.89346206e-01 -6.24194741e-01 -4.56231922e-01 -5.69406636e-02
4.26750183e-01 -9.87020671e-01 5.43305099e-01 5.28787911e-01
-1.33620366e-01 -3.48050833e-01 -1.44403172e+00 8.10369253e-01
-6.73596442e-01 7.29562759e-01 8.17628860e-01 5.94273388e-01
-2.32410550e-01 9.38340425e-01 -6.05229199e-01 -3.54202479e-01
3.01542491e-01 4.12802815e-01 -6.73021793e-01 2.00174632e-03
-9.90427256e-01 7.01287210e-01 1.11359727e+00 1.23682737e-01
-4.94987778e-02 -1.72277257e-01 -5.72216809e-01 -6.03203364e-02
2.73962524e-02 -3.17380205e-02 4.67238277e-01 -1.43082249e+00
-1.15644002e+00 5.20502388e-01 8.02502409e-02 -1.88950554e-01
-6.34127930e-02 -3.62345092e-02 -4.91711468e-01 5.68797998e-02
-1.06418133e-01 5.54985786e-03 6.24381363e-01 -1.02480090e+00
-9.01081085e-01 -8.83579791e-01 -6.38383150e-01 3.65276664e-01
-5.64976394e-01 -9.17521119e-02 6.26092494e-01 -3.19417924e-01
6.82745218e-01 -1.08119023e+00 -4.52958420e-03 -5.66553831e-01
-4.58376914e-01 1.09467328e-01 1.27067125e+00 -7.02037334e-01
9.51444447e-01 -2.27415919e+00 1.56180635e-01 7.61843204e-01
-3.91556114e-01 5.25546253e-01 4.60841954e-02 4.14889187e-01
-4.18524861e-01 8.42008516e-02 -5.19977510e-01 4.76886272e-01
-5.21579802e-01 -1.47173777e-01 3.62201274e-04 6.82833493e-01
-9.91795864e-03 8.83814320e-02 -5.00318468e-01 -9.53191891e-02
5.32692969e-01 3.32263887e-01 1.93437934e-01 -5.16595431e-02
1.64220836e-02 3.00874949e-01 -4.74142462e-01 6.46283925e-01
1.03239703e+00 -5.43098114e-02 -4.23964560e-02 -6.44949734e-01
-4.73961353e-01 -4.53724772e-01 -1.52551401e+00 1.14696574e+00
-8.67849886e-02 6.31607711e-01 -1.85698748e-01 -1.05276239e+00
1.08692122e+00 1.26937389e-01 4.81023461e-01 -2.18287766e-01
4.19260651e-01 3.28996539e-01 -1.60224605e-02 -7.48802722e-01
2.77086794e-01 1.60841327e-02 7.00773060e-01 2.25170568e-01
-1.94491550e-01 -1.84456015e-03 1.98034599e-01 -5.67588449e-01
4.26359802e-01 -1.49335861e-01 7.04222500e-01 -4.97653127e-01
8.95976722e-01 4.05812413e-01 3.53733063e-01 2.12413773e-01
5.60281659e-03 1.77298293e-01 -1.45959809e-01 -3.51731896e-01
-7.29043663e-01 -3.13652903e-01 -5.98465919e-01 6.93658054e-01
1.03373609e-01 3.58514160e-01 -3.75849605e-01 -4.34645027e-01
5.67198806e-02 7.41234958e-01 -3.96758318e-01 -3.52110751e-02
3.15658242e-01 -1.25732994e+00 5.11350669e-02 -2.08112478e-01
7.44840384e-01 -7.26303756e-01 -7.06828356e-01 1.01050384e-01
5.60099296e-02 -5.33616722e-01 3.01256388e-01 3.65527719e-01
-1.06939244e+00 -1.25171268e+00 -2.01307371e-01 -5.09768546e-01
5.18035233e-01 7.23884404e-01 2.37316802e-01 -2.34139010e-01
-6.06763303e-01 -2.52298445e-01 -9.38268721e-01 -6.53419971e-01
-1.66771129e-01 1.65971573e-02 -1.45871654e-01 2.96261102e-01
4.21975851e-01 -4.66360122e-01 -4.27032381e-01 3.48985344e-01
-1.06952775e+00 -1.19200788e-01 6.95451260e-01 9.22977269e-01
5.68775177e-01 9.40853894e-01 4.15789455e-01 -7.07122147e-01
3.80945325e-01 -4.70072269e-01 -9.01678741e-01 2.19150916e-01
-7.74752617e-01 1.61725029e-01 5.60818613e-01 6.54130355e-02
-9.20798123e-01 3.16030234e-01 4.23194319e-01 1.19794667e-01
-2.12504700e-01 8.61799717e-01 -5.96232079e-02 -6.17789745e-01
9.04832363e-01 5.32075167e-01 1.84798792e-01 -4.68760699e-01
-1.82195112e-01 1.16002142e+00 7.59459361e-02 9.66724083e-02
6.62520885e-01 3.76117617e-01 3.21222931e-01 -1.12590468e+00
-7.65970945e-01 -8.73746455e-01 -4.78425413e-01 -9.45242029e-03
5.08087575e-01 -7.48908997e-01 -4.62776333e-01 2.88747698e-01
-5.33331454e-01 4.94332194e-01 2.17066407e-01 1.01424944e+00
-9.95710120e-02 4.59499478e-01 1.37004375e-01 -1.02336061e+00
-8.09272110e-01 -1.10287964e+00 5.62206209e-01 5.28011441e-01
2.69529372e-01 -5.46616614e-01 -7.21017495e-02 3.22536200e-01
4.74461794e-01 5.40549815e-01 9.14967597e-01 -9.54541206e-01
-5.44115789e-02 -3.48398536e-01 -4.38804150e-01 7.44392097e-01
5.78158021e-01 4.39530283e-01 -8.83318722e-01 -1.67469740e-01
8.09860155e-02 9.46827512e-03 1.08755064e+00 3.36058050e-01
1.12957668e+00 -2.91012079e-01 -4.35612202e-01 6.10466242e-01
2.01062131e+00 6.84498787e-01 5.90364516e-01 6.93943024e-01
2.31990173e-01 5.76423824e-01 1.12089360e+00 8.05116653e-01
-4.97414917e-01 3.55128385e-02 6.93623245e-01 4.79749776e-02
5.39685428e-01 4.60233182e-01 -8.85914192e-02 4.48955208e-01
-2.29244828e-01 -2.53824499e-02 -7.44728565e-01 3.85161825e-02
-1.75670969e+00 -1.17069387e+00 -3.95663887e-01 2.40197515e+00
4.60441977e-01 -4.48484570e-01 -2.82621771e-01 8.75619829e-01
9.57302272e-01 6.79086149e-03 -3.35925817e-01 -7.71146035e-03
-4.62772638e-01 3.15061003e-01 7.77864158e-01 4.00690794e-01
-1.67351186e+00 4.59512353e-01 4.90599489e+00 6.35133564e-01
-1.53943825e+00 -1.55342534e-01 5.56963861e-01 1.73575148e-01
2.41634343e-02 -1.19426744e-02 -4.31502730e-01 3.50751013e-01
5.70658028e-01 -3.40557098e-01 4.81214732e-01 8.09923649e-01
2.75486737e-01 -4.19953436e-01 -3.54425639e-01 1.08544099e+00
4.43771519e-02 -9.43962812e-01 1.79180309e-01 -1.78250328e-01
8.57278824e-01 1.23430908e-01 -1.06476426e-01 -2.70332783e-01
-2.63250083e-01 -9.07296658e-01 1.16728663e-01 6.73714459e-01
2.12790534e-01 -1.17924643e+00 1.12712479e+00 2.21855104e-01
-1.08760130e+00 -4.76139873e-01 -5.23781717e-01 3.82438153e-02
-5.90316355e-01 9.54615593e-01 -7.04811871e-01 9.41771507e-01
5.80730975e-01 6.04133904e-01 -5.56472778e-01 1.47343218e+00
3.33287567e-01 3.96207809e-01 -3.60196739e-01 -3.86099935e-01
2.65355378e-01 -6.09304726e-01 6.97864950e-01 9.00463581e-01
6.04999483e-01 3.38419437e-01 1.85106352e-01 3.62673253e-01
4.29591805e-01 9.06254768e-01 -6.08482122e-01 -2.30410576e-01
4.18663591e-01 1.30216074e+00 -6.68432653e-01 -1.69466779e-01
-8.20445195e-02 7.67114878e-01 -4.74119216e-01 2.77354685e-03
-5.97600758e-01 -8.41264009e-01 4.38069463e-01 -2.60811657e-01
6.09359182e-02 -4.74400632e-03 -2.34680966e-01 -8.58364344e-01
-8.27802867e-02 -7.40333080e-01 5.17721236e-01 -6.99986219e-01
-1.09966815e+00 8.03442001e-01 -1.88200906e-01 -1.60894263e+00
4.18779939e-01 -7.94892967e-01 -1.65577233e-01 9.22548950e-01
-1.62212336e+00 -8.05968583e-01 -7.44548500e-01 4.59008813e-01
3.96125406e-01 -3.91538620e-01 9.48226154e-01 1.43819645e-01
-5.48207760e-01 -1.78153917e-01 7.72407234e-01 -2.74863690e-01
4.08016443e-01 -8.57712567e-01 -7.67246544e-01 7.85849750e-01
-2.43747070e-01 1.73169091e-01 7.89932430e-01 -6.12979472e-01
-1.37048018e+00 -1.20085049e+00 3.18897516e-01 6.24369681e-01
3.43152374e-01 7.26999819e-01 -6.08313978e-01 2.64476418e-01
4.62215245e-02 -1.01098120e-01 1.13368928e+00 -2.90690422e-01
6.68434724e-02 -4.15300488e-01 -1.40350866e+00 1.10203892e-01
1.84420586e-01 -1.34898340e-02 -2.99982995e-01 8.22246373e-01
2.34409690e-01 2.17039227e-01 -9.44818735e-01 6.61997020e-01
6.79803193e-01 -1.03309953e+00 6.45757020e-01 -4.81449693e-01
1.53163508e-01 -5.49345315e-01 -4.31898028e-01 -1.57334173e+00
-4.60580319e-01 -4.96755280e-02 7.28479385e-01 8.95078480e-01
6.29094839e-01 -8.64545286e-01 4.81583118e-01 3.72791529e-01
2.53431946e-01 -2.89058477e-01 -5.90074301e-01 -7.28562534e-01
-6.75956666e-01 1.55079747e-02 7.06022918e-01 1.20938766e+00
3.77393328e-02 1.01858325e-01 -3.12556446e-01 5.74159920e-01
8.50257695e-01 5.06125867e-01 1.78443953e-01 -1.92242301e+00
-4.70885374e-02 -3.71412247e-01 -8.19783747e-01 3.73202115e-01
-4.39941585e-02 -1.02040923e+00 -2.19742775e-01 -1.37278080e+00
2.02720806e-01 -3.99354070e-01 -5.34090459e-01 5.87874830e-01
-1.40223458e-01 1.61955077e-02 -2.98098922e-02 1.99069142e-01
1.82178453e-01 5.38503885e-01 8.44594538e-01 -3.70422155e-01
-5.15472353e-01 2.48957589e-01 -4.94120091e-01 5.98356307e-01
1.09442055e+00 -5.56722641e-01 -4.49127644e-01 -3.47424783e-02
-6.81385174e-02 -9.70318690e-02 3.47844255e-03 -1.32856524e+00
7.20689371e-02 -6.11220300e-01 6.38777137e-01 -5.73012292e-01
5.97334392e-02 -1.20277488e+00 6.00484312e-01 6.98095620e-01
-1.36036158e-01 -4.23233896e-01 2.59291619e-01 2.26965144e-01
-3.51676226e-01 -6.49110675e-01 1.03189647e+00 -1.28273115e-01
-9.88947690e-01 2.77868301e-01 -9.56892371e-02 -9.31643665e-01
1.39649355e+00 -2.03060150e-01 -3.72063518e-01 2.64603108e-01
-7.65360594e-01 -1.87842131e-01 1.45397782e-01 7.61873201e-02
6.28550947e-01 -1.04179406e+00 -9.89308119e-01 3.38361382e-01
5.26248276e-01 -7.26822376e-01 2.21237823e-01 5.77830613e-01
-9.17280853e-01 3.25009674e-01 -6.28918648e-01 -4.81744766e-01
-1.78210127e+00 1.20500647e-01 3.24745953e-01 1.11433648e-01
-9.53451023e-02 6.18116200e-01 -6.30393803e-01 -3.39605570e-01
-2.71535426e-01 -2.56884843e-02 -9.54698265e-01 3.45659256e-01
6.33099079e-01 6.09387338e-01 4.25906569e-01 -7.13607132e-01
-4.46011037e-01 6.87589169e-01 3.04607779e-01 1.16016977e-01
1.61076295e+00 8.10617432e-02 -6.65983737e-01 2.77094305e-01
1.23990822e+00 -3.80772464e-02 -3.65703076e-01 -1.76905412e-02
-4.93069775e-02 -8.33625317e-01 5.63237846e-01 -9.60666239e-01
-9.14932668e-01 4.24048692e-01 1.40648592e+00 3.16607207e-01
1.36504209e+00 -7.93810368e-01 1.24183662e-01 8.56236637e-01
2.43545458e-01 -1.16396642e+00 -8.40596497e-01 1.88713655e-01
7.63061047e-01 -1.50775528e+00 3.62775236e-01 -5.55332959e-01
-5.37294447e-01 1.61722791e+00 1.55432865e-01 1.01004839e-01
1.03323877e+00 -1.01426132e-01 -1.44459933e-01 -5.49076684e-02
-2.69925684e-01 -4.46692467e-01 2.75088996e-01 3.51302654e-01
5.37079513e-01 4.52994913e-01 -7.13683069e-01 2.60841966e-01
-4.05295230e-02 7.68400207e-02 2.56031275e-01 9.73161042e-01
-1.16262627e+00 -8.36624622e-01 -6.27126992e-01 1.00146830e+00
-2.51547933e-01 -7.73578957e-02 -3.17828655e-01 4.45002079e-01
9.16016996e-02 1.21869242e+00 -4.37110573e-01 -7.10917711e-01
1.32303014e-01 2.18903586e-01 1.59248084e-01 -3.25453252e-01
-3.25574905e-01 -3.12895983e-01 4.28830385e-02 -9.30911079e-02
-6.87527478e-01 -4.99690145e-01 -1.00081539e+00 3.83283943e-02
-9.21667755e-01 5.13787031e-01 1.32668066e+00 9.71683979e-01
3.82641077e-01 2.82471895e-01 1.17347467e+00 -7.89034426e-01
-7.56025493e-01 -1.11501276e+00 -1.00477219e+00 2.44427964e-01
9.39204916e-02 -6.77569687e-01 -6.66574895e-01 -1.28436312e-01]
|
[9.781513214111328, -1.8349343538284302]
|
a5685750-f508-4a02-b77c-4e17ef03d65c
|
world-models
|
1803.10122
| null |
http://arxiv.org/abs/1803.10122v4
|
http://arxiv.org/pdf/1803.10122v4.pdf
|
World Models
|
We explore building generative neural network models of popular reinforcement
learning environments. Our world model can be trained quickly in an
unsupervised manner to learn a compressed spatial and temporal representation
of the environment. By using features extracted from the world model as inputs
to an agent, we can train a very compact and simple policy that can solve the
required task. We can even train our agent entirely inside of its own
hallucinated dream generated by its world model, and transfer this policy back
into the actual environment.
An interactive version of this paper is available at
https://worldmodels.github.io/
|
['Jürgen Schmidhuber', 'David Ha']
|
2018-03-27
| null | null | null | null |
['carracing-v0']
|
['playing-games']
|
[-4.92166787e-01 5.21217167e-01 -1.03107490e-01 -1.60971448e-01
-1.89546987e-01 -5.76264560e-01 7.53364205e-01 -4.64960039e-01
-4.92984325e-01 1.01547730e+00 4.77397978e-01 2.45520566e-03
9.63309631e-02 -1.18631315e+00 -1.00011599e+00 -7.37528741e-01
-2.70662874e-01 8.19398701e-01 -1.64199173e-01 -3.63306612e-01
3.65722924e-03 3.18354338e-01 -1.29886007e+00 -1.05841652e-01
5.92454553e-01 4.30348843e-01 9.68989909e-01 7.67571032e-01
2.65921265e-01 1.42961192e+00 -4.82728630e-01 2.05935061e-01
2.84930408e-01 -6.51199579e-01 -8.59906077e-01 9.16612968e-02
-6.30561590e-01 -7.27441728e-01 -1.19343781e+00 1.02810085e+00
1.83242738e-01 5.47662795e-01 1.92649871e-01 -1.07666588e+00
-9.65938985e-01 6.55037284e-01 2.00992167e-01 1.25003278e-01
4.06961501e-01 6.55669451e-01 4.99656379e-01 -6.57416165e-01
9.12178457e-01 8.48668933e-01 -1.19298942e-01 8.05349231e-01
-1.11919153e+00 -6.09201789e-01 5.39545203e-04 3.27523917e-01
-1.24842322e+00 -5.84778607e-01 8.18564475e-01 8.14193189e-02
1.27701259e+00 -1.74202263e-01 1.25196719e+00 1.54198718e+00
4.35018480e-01 7.24052072e-01 1.10700905e+00 -2.02512011e-01
8.40493977e-01 5.91571704e-02 -5.08290350e-01 8.16289961e-01
-2.67895520e-01 7.52611995e-01 -6.59137487e-01 1.16099454e-01
1.08070219e+00 4.82793339e-02 -1.47580847e-01 -5.47773600e-01
-1.07280684e+00 1.03675044e+00 8.64778280e-01 2.47901574e-01
-5.78298211e-01 5.62108994e-01 -1.41177610e-01 3.40892792e-01
2.12919325e-01 7.81153977e-01 -6.01815060e-02 -4.69167709e-01
-5.90882361e-01 3.03977221e-01 8.89921069e-01 7.66106904e-01
9.63086665e-01 4.37855601e-01 3.19387376e-01 3.66088688e-01
3.30539703e-01 6.13424182e-01 9.69085515e-01 -1.51527011e+00
-1.02720670e-01 1.02683753e-01 2.57529348e-01 -4.13452983e-01
-3.62290323e-01 -3.29322755e-01 -5.85210919e-01 3.17045778e-01
-2.96588749e-01 -5.15999675e-01 -1.08849394e+00 1.92282629e+00
1.40312150e-01 5.15141010e-01 4.74735558e-01 7.63853431e-01
6.52980685e-01 1.06519997e+00 -1.14488497e-01 -9.68062356e-02
7.09343314e-01 -1.04508400e+00 -5.66891372e-01 -8.30319464e-01
5.23358762e-01 -9.45648253e-02 7.57646501e-01 3.08843166e-01
-1.40521669e+00 -3.77351105e-01 -1.08273137e+00 3.76102999e-02
-4.25359458e-01 -2.42599711e-01 9.86711740e-01 -1.46647424e-01
-1.68574393e+00 6.97168887e-01 -1.33073914e+00 -4.76421535e-01
4.73344803e-01 4.77519065e-01 -5.01141131e-01 8.03806260e-02
-1.17887390e+00 1.31569695e+00 9.05515909e-01 -8.61076936e-02
-1.80902624e+00 5.76814152e-02 -1.06896770e+00 -9.48742554e-02
3.47343773e-01 -8.32215607e-01 1.68947995e+00 -7.87735224e-01
-1.88396502e+00 4.86217618e-01 -2.17518836e-01 -6.79393411e-01
-4.72641848e-02 -2.50894815e-01 -2.00145513e-01 1.66029483e-01
2.19303910e-02 7.14091241e-01 7.07935095e-01 -1.17496800e+00
-1.96028024e-01 -2.22219691e-01 2.24503294e-01 5.08957326e-01
-2.85817590e-02 -2.61562556e-01 -3.19238782e-01 -2.91133881e-01
3.32474262e-02 -1.07021379e+00 -6.01774395e-01 -2.57163793e-01
-1.47981301e-01 -1.27248988e-01 6.35902047e-01 -3.73124927e-01
5.93827963e-01 -2.00200438e+00 5.35602450e-01 -4.38240692e-02
1.58336952e-01 1.53694585e-01 -2.89325356e-01 7.43625283e-01
-2.54970901e-02 -2.22794086e-01 -6.14536414e-03 -3.45678151e-01
-5.89777790e-02 7.37069190e-01 -7.14990437e-01 5.13074815e-01
-6.56994879e-02 1.23929286e+00 -1.31732047e+00 -1.09195068e-01
4.18784052e-01 3.85206312e-01 -6.25122726e-01 7.29064822e-01
-5.05313456e-01 7.56745160e-01 -5.15962899e-01 6.32818937e-02
2.63180315e-01 -4.06871229e-01 3.77398610e-01 4.84408826e-01
8.83593857e-02 5.27991891e-01 -8.56641531e-01 2.21480393e+00
-6.40804827e-01 6.20261490e-01 -1.24788038e-01 -9.96263862e-01
7.10361421e-01 5.19038260e-01 3.42590541e-01 -9.68627095e-01
1.63799554e-01 -1.55651659e-01 -4.00220044e-02 -5.77882171e-01
3.60048354e-01 -3.60455930e-01 -1.10485077e-01 7.69374311e-01
7.07249224e-01 -3.67522448e-01 6.57780841e-02 3.77501428e-01
1.24459088e+00 6.07945323e-01 3.67889911e-01 5.60231805e-02
-2.23581314e-01 -2.09607221e-02 3.73814404e-01 8.84445190e-01
-4.98489104e-02 1.27684027e-01 8.39799568e-02 -7.34917223e-01
-1.01855588e+00 -1.38112533e+00 2.87817329e-01 9.35146093e-01
1.06769390e-01 -5.39011180e-01 -5.29613793e-01 -2.93907940e-01
-4.13072675e-01 1.04917359e+00 -8.87754202e-01 -5.40974915e-01
-6.77619874e-01 -4.70713556e-01 2.73997575e-01 6.16954386e-01
7.23382056e-01 -1.84379065e+00 -1.03788841e+00 2.80239046e-01
-5.40186241e-02 -9.15159583e-01 1.55528158e-01 6.48777664e-01
-7.76753008e-01 -6.84736371e-01 -2.68454820e-01 -8.66934061e-01
5.38485825e-01 7.52127171e-03 1.14597380e+00 -6.20394526e-03
-3.96604761e-02 3.27237129e-01 -1.95393816e-01 -2.35234022e-01
-5.26575506e-01 -6.07695505e-02 2.79470801e-01 -5.04539192e-01
1.31049603e-01 -1.03389907e+00 -5.28954446e-01 -2.53139704e-01
-7.98452497e-01 4.48017150e-01 4.32626456e-01 6.32595837e-01
3.93453032e-01 5.02858721e-02 4.16771501e-01 -5.96815646e-01
6.92052722e-01 -9.33808625e-01 -5.70706069e-01 -1.66123584e-01
-4.12603557e-01 4.29704338e-01 7.34068573e-01 -4.28448409e-01
-1.18468344e+00 6.89320415e-02 -3.42667073e-01 -4.52554971e-01
-4.24799591e-01 4.73834455e-01 1.16878018e-01 2.66023815e-01
6.62519455e-01 8.03188443e-01 -2.14183524e-01 -3.58058989e-01
6.73077643e-01 2.82779068e-01 6.62792742e-01 -5.93528807e-01
1.06099808e+00 4.75640237e-01 -4.35006410e-01 -5.50305068e-01
-7.68577456e-01 1.15641229e-01 -5.08979678e-01 -8.78174454e-02
8.69350970e-01 -1.00644505e+00 -5.75665176e-01 3.12939465e-01
-9.92607236e-01 -1.17619562e+00 -7.34930396e-01 5.23278892e-01
-1.23012233e+00 -3.49400073e-01 -6.23399079e-01 -6.01972938e-01
1.05429143e-01 -8.14351976e-01 5.84713578e-01 5.95108449e-01
-2.06228644e-01 -1.09043860e+00 8.01678956e-01 -2.33658150e-01
5.61190486e-01 3.07437703e-02 4.15870339e-01 -4.27409202e-01
-5.52708030e-01 1.62446156e-01 3.25888008e-01 -6.54754229e-03
1.00443467e-01 -3.37831140e-01 -1.07093036e+00 -2.88397729e-01
4.14829940e-01 -8.30606699e-01 8.58989120e-01 3.00326467e-01
1.24136674e+00 -7.93866456e-01 -4.47167456e-01 9.28031027e-01
1.42563140e+00 3.92891914e-01 8.24004173e-01 5.66034138e-01
2.08132520e-01 1.64681345e-01 2.09504902e-01 6.29235566e-01
5.81290364e-01 3.67190540e-01 7.66637683e-01 1.47327766e-01
7.56798908e-02 -7.31081367e-01 6.21706009e-01 6.48806393e-01
-1.56443402e-01 -9.53868851e-02 -8.39991629e-01 5.68218231e-01
-1.91845644e+00 -1.34847867e+00 8.49174261e-01 1.64737558e+00
9.16345477e-01 -1.21325828e-01 -6.12747110e-02 -4.31752890e-01
1.17021494e-01 3.94912779e-01 -8.74018312e-01 -4.77567166e-01
9.52280834e-02 4.31657702e-01 7.64481202e-02 5.08985579e-01
-8.85864377e-01 1.20226824e+00 7.08674717e+00 3.20909321e-01
-1.16548383e+00 1.82810917e-01 3.29142183e-01 -4.80102509e-01
-2.85626829e-01 2.54520118e-01 -2.38525361e-01 4.42841411e-01
1.50059295e+00 -6.14304662e-01 1.37090266e+00 1.02244449e+00
2.35433251e-01 -1.21763967e-01 -1.03437781e+00 7.95946836e-01
-1.14423417e-01 -1.57301319e+00 -3.83492023e-01 2.93212891e-01
6.41683757e-01 6.73107386e-01 3.25987697e-01 7.36839414e-01
1.14134026e+00 -1.29008126e+00 7.41177619e-01 4.67391580e-01
4.15494293e-01 -6.56833649e-01 1.99875832e-01 9.81069684e-01
-6.72704160e-01 -3.04756761e-01 -5.04061401e-01 -5.31421602e-01
2.10765563e-02 -5.27251959e-02 -9.83356595e-01 2.06393795e-03
5.93391538e-01 7.49711335e-01 -4.73901689e-01 9.41344798e-01
-6.07406735e-01 5.38830698e-01 -2.70061463e-01 1.81561662e-03
3.66636604e-01 -1.79643318e-01 4.09772366e-01 6.69184268e-01
4.60314125e-01 3.74234051e-01 2.19142094e-01 1.20297885e+00
-2.93655396e-01 -4.74475712e-01 -1.20311010e+00 -1.12188846e-01
4.08923477e-01 1.13426995e+00 -4.67491716e-01 -4.63562191e-01
-4.47068829e-03 1.10995281e+00 8.42038572e-01 6.59917712e-01
-9.08822060e-01 6.93131611e-02 4.42120254e-01 -1.75662234e-01
2.99989700e-01 -4.56667006e-01 2.60091811e-01 -1.47088802e+00
-4.28915352e-01 -7.79383540e-01 -2.66870204e-02 -1.28682399e+00
-6.18547082e-01 7.81489253e-01 2.58810278e-02 -9.75238025e-01
-9.67142642e-01 -2.42236704e-01 -9.39474821e-01 5.76132596e-01
-1.19021261e+00 -9.58035231e-01 -1.32476658e-01 1.02489913e+00
5.64625025e-01 -2.50471115e-01 1.51840127e+00 -5.09008765e-01
-4.66708869e-01 -1.71378329e-01 2.47508019e-01 2.08777979e-01
1.34299323e-01 -1.27782750e+00 4.37418044e-01 8.20830941e-01
5.32939970e-01 6.99512541e-01 9.61698472e-01 -5.07823646e-01
-1.54866576e+00 -1.15625429e+00 3.08801025e-01 -4.46335703e-01
7.47120440e-01 -3.08658779e-01 -7.69148171e-01 1.54412687e+00
8.99668097e-01 1.21994071e-01 5.20576537e-01 -2.74835881e-02
-9.57036689e-02 2.05629587e-01 -1.10326982e+00 6.69600666e-01
9.70484436e-01 -6.17738307e-01 -8.04397702e-01 6.15635931e-01
5.93628824e-01 -5.87369740e-01 -5.86748183e-01 -2.70468533e-01
1.42685607e-01 -8.88072193e-01 8.11195672e-01 -6.86975181e-01
4.11168665e-01 -8.98300260e-02 -8.02756399e-02 -1.96599245e+00
-5.71890235e-01 -8.59710574e-01 -8.08485866e-01 4.98555958e-01
1.48512796e-01 -8.07397962e-01 7.84814239e-01 4.78076160e-01
-7.40244836e-02 -5.67275703e-01 -1.03060186e+00 -6.29088521e-01
2.42349684e-01 -2.98737675e-01 6.77685678e-01 5.70978999e-01
3.08356225e-01 1.99822739e-01 -4.91345286e-01 1.90309361e-01
5.40913284e-01 3.26652348e-01 5.64375460e-01 -6.55053854e-01
-6.89266860e-01 7.40109533e-02 -8.29992741e-02 -9.44044769e-01
4.72678274e-01 -9.68768418e-01 2.93286383e-01 -1.78609467e+00
3.66087049e-01 -3.03514451e-01 -3.88907403e-01 8.88399243e-01
3.51162463e-01 2.82637239e-01 3.88642162e-01 2.79197693e-01
-8.57401967e-01 9.82653320e-01 1.31281233e+00 6.96231350e-02
-2.83716202e-01 -2.45619744e-01 -8.10598671e-01 9.76634622e-01
1.48057258e+00 -6.34787977e-01 -6.97610736e-01 -7.55546689e-01
3.82377565e-01 3.64481300e-01 5.26168823e-01 -1.22393394e+00
3.33749712e-01 -6.48608088e-01 8.82370889e-01 -6.57519209e-04
8.34892929e-01 -4.73855883e-01 2.90384173e-01 6.42857611e-01
-3.20898622e-01 5.64716123e-02 1.88473746e-01 4.05143440e-01
-5.58158732e-04 -1.27048820e-01 6.76039100e-01 -8.03232729e-01
-9.85099435e-01 3.19621593e-01 -6.00141525e-01 -4.93974462e-02
1.17719960e+00 1.13025062e-01 -5.70231795e-01 -8.51416767e-01
-1.21362281e+00 1.53898820e-01 8.92952621e-01 4.57922548e-01
9.21094775e-01 -1.40075254e+00 -4.72791284e-01 4.30597931e-01
-3.14117432e-01 -1.03111453e-01 5.08682728e-02 2.42512867e-01
-3.47000390e-01 3.60143214e-01 -6.44680798e-01 -1.11095205e-01
-3.11243355e-01 8.61620188e-01 6.68137133e-01 -1.92224339e-01
-8.92144918e-01 6.83442295e-01 2.81171709e-01 -4.20977890e-01
-1.61330640e-01 -1.22489203e-02 -5.97949810e-02 -5.62461495e-01
7.23401189e-01 -9.04297680e-02 -4.20594841e-01 -5.36622107e-01
-2.50567108e-01 -4.82331142e-02 2.04582348e-01 -6.14996195e-01
1.83326185e+00 -2.31978465e-02 1.40442969e-02 4.35560167e-01
1.11697078e+00 -3.36121053e-01 -1.52501678e+00 -3.28432061e-02
-5.51342130e-01 -2.11807162e-01 1.91908196e-01 -7.69583046e-01
-8.91677976e-01 6.22484744e-01 4.12055761e-01 2.87210792e-01
1.13598096e+00 3.80774379e-01 5.78141987e-01 6.77093506e-01
7.27579355e-01 -1.12895179e+00 3.70685130e-01 7.53984988e-01
1.09799886e+00 -9.51455116e-01 -2.38311827e-01 6.63154483e-01
-7.85214782e-01 8.32450449e-01 6.54211164e-01 -8.95011246e-01
8.02107930e-01 5.10245979e-01 -2.47602209e-01 -2.87065119e-01
-1.21970320e+00 -2.20480055e-01 -3.68519217e-01 8.98757398e-01
-9.39420834e-02 1.82991549e-01 5.17317474e-01 2.85772204e-01
-6.74132645e-01 2.90551305e-01 7.81870067e-01 9.16518271e-01
-7.05090463e-01 -1.08363724e+00 -1.05195669e-02 2.28433430e-01
1.62007660e-03 4.54549454e-02 -3.28286774e-02 5.89731991e-01
1.03532396e-01 7.01942503e-01 1.36646941e-01 -4.23720539e-01
-1.52135536e-01 5.92097566e-02 7.12106168e-01 -9.16696787e-01
-1.09920092e-01 -1.74486235e-01 -1.57091826e-01 -1.20040703e+00
-2.34743938e-01 -4.00354028e-01 -1.66170073e+00 -4.04543757e-01
3.50828558e-01 2.60227770e-01 4.72408623e-01 9.29782569e-01
2.41486013e-01 4.71752733e-01 7.35038519e-01 -1.25685728e+00
-4.04186785e-01 -9.78289008e-01 -5.69776535e-01 1.26073107e-01
6.05404556e-01 -5.73966742e-01 -7.20051154e-02 6.87549561e-02]
|
[4.21512508392334, 1.4773885011672974]
|
6e901e39-465d-4b5c-9ad2-ddf5b889e6f9
|
a-discrete-cvae-for-response-generation-on-1
|
1911.09845
| null |
https://arxiv.org/abs/1911.09845v1
|
https://arxiv.org/pdf/1911.09845v1.pdf
|
A Discrete CVAE for Response Generation on Short-Text Conversation
|
Neural conversation models such as encoder-decoder models are easy to generate bland and generic responses. Some researchers propose to use the conditional variational autoencoder(CVAE) which maximizes the lower bound on the conditional log-likelihood on a continuous latent variable. With different sampled la-tent variables, the model is expected to generate diverse responses. Although the CVAE-based models have shown tremendous potential, their improvement of generating high-quality responses is still unsatisfactory. In this paper, we introduce a discrete latent variable with an explicit semantic meaning to improve the CVAE on short-text conversation. A major advantage of our model is that we can exploit the semantic distance between the latent variables to maintain good diversity between the sampled latent variables. Accordingly, we pro-pose a two-stage sampling approach to enable efficient diverse variable selection from a large latent space assumed in the short-text conversation task. Experimental results indicate that our model outperforms various kinds of generation models under both automatic and human evaluations and generates more diverse and in-formative responses.
|
['Xiaojiang Liu', 'Jun Gao', 'Shuming Shi', 'Junhui Li', 'Guodong Zhou', 'Wei Bi']
|
2019-11-22
|
a-discrete-cvae-for-response-generation-on
|
https://aclanthology.org/D19-1198
|
https://aclanthology.org/D19-1198.pdf
|
ijcnlp-2019-11
|
['short-text-conversation']
|
['natural-language-processing']
|
[ 2.57138256e-02 2.69727081e-01 -9.23721120e-02 -6.97117686e-01
-1.11858642e+00 -3.53089809e-01 6.70021653e-01 -5.33264279e-01
-2.72395853e-02 1.12038779e+00 7.66063511e-01 1.34962142e-01
2.19102800e-01 -8.12704384e-01 -3.81912917e-01 -7.44416654e-01
6.09328806e-01 8.43228936e-01 -3.39977205e-01 -2.88257420e-01
1.42196670e-01 -2.38871068e-01 -1.62898123e+00 5.30803502e-01
1.11730945e+00 6.03405058e-01 5.76667070e-01 7.23697484e-01
-5.08635879e-01 6.60283387e-01 -1.05734003e+00 -6.64014518e-01
-2.17333585e-01 -1.07848001e+00 -7.40297139e-01 1.67854890e-01
-8.03140849e-02 -2.91554600e-01 -6.00408539e-02 9.16982889e-01
6.77099109e-01 3.23150784e-01 9.05913413e-01 -1.26869977e+00
-8.17419231e-01 9.64800179e-01 -3.81315537e-02 -3.50797296e-01
6.16231382e-01 -4.05983850e-02 1.20180392e+00 -8.20197344e-01
7.20261812e-01 1.52448094e+00 3.87599558e-01 1.11211479e+00
-1.31191969e+00 -6.54624879e-01 1.18770547e-01 1.63055118e-02
-1.17481363e+00 -3.85149390e-01 9.31731403e-01 -3.14830333e-01
7.28529155e-01 3.09483200e-01 4.36987936e-01 1.79267263e+00
2.19347268e-01 1.14444017e+00 1.13720572e+00 -4.32166189e-01
4.57472950e-01 7.28898287e-01 -8.54119360e-02 1.89109489e-01
-3.30444932e-01 -1.59607366e-01 -8.38871837e-01 -4.84721422e-01
6.14515483e-01 -9.52420160e-02 -3.59149098e-01 -7.85183161e-02
-1.01285207e+00 1.39341629e+00 -2.83635482e-02 7.71934763e-02
-6.62660480e-01 -9.78924334e-02 2.79868782e-01 3.89774799e-01
6.50102913e-01 5.68560362e-01 -3.06295335e-01 -4.69249457e-01
-9.39063430e-01 5.00958681e-01 1.14380836e+00 1.02836621e+00
5.99228859e-01 2.47178257e-01 -4.58858132e-01 1.29210055e+00
3.76402199e-01 3.97259057e-01 6.90348148e-01 -1.04249501e+00
3.97264510e-01 3.23018879e-01 2.10969612e-01 -8.57788026e-01
2.91035831e-01 -3.23636621e-01 -8.65396142e-01 -1.00281522e-01
8.18462893e-02 -4.71326083e-01 -5.77778161e-01 1.96550167e+00
1.05952643e-01 -2.86357164e-01 4.19192880e-01 8.33320498e-01
7.00417876e-01 1.02294934e+00 5.54727279e-02 -4.18307722e-01
9.56540644e-01 -1.12611628e+00 -1.18315053e+00 -1.58916637e-01
2.37693399e-01 -6.09531224e-01 1.18243718e+00 3.90478671e-01
-1.13866460e+00 -6.36021435e-01 -8.10377419e-01 1.24933749e-01
1.42724335e-01 -4.88866903e-02 7.61419833e-01 6.30631506e-01
-1.04059303e+00 2.82004029e-01 -4.45465446e-01 -3.55821103e-02
-1.59004286e-01 1.48369998e-01 -1.28666505e-01 1.92251541e-02
-1.58707392e+00 7.35485375e-01 6.76215962e-02 -1.11246062e-02
-1.04847884e+00 -1.44866869e-01 -8.45161676e-01 1.65050656e-01
8.25121924e-02 -9.18670535e-01 1.47334433e+00 -1.35156429e+00
-2.12653494e+00 3.45071048e-01 -5.54242969e-01 -2.21026465e-01
6.03040636e-01 -3.81468586e-03 -2.01093495e-01 3.49111818e-02
3.15787643e-02 8.85547876e-01 9.44693804e-01 -1.30775571e+00
-2.98317432e-01 7.95601159e-02 2.40727626e-02 4.14348274e-01
-4.51618165e-01 -1.17971294e-01 -2.30150998e-01 -6.20244622e-01
-7.21770152e-02 -9.73998129e-01 -2.28527248e-01 -5.02491951e-01
-1.98689863e-01 -6.48423135e-01 3.91759306e-01 -5.37888229e-01
1.24319613e+00 -1.88304090e+00 4.56151605e-01 -1.12605914e-01
1.03844233e-01 -2.04521090e-01 -1.41348645e-01 7.05170691e-01
2.47074679e-01 -5.98632097e-02 -1.21782355e-01 -6.05284572e-01
1.65255532e-01 2.19056442e-01 -5.91614485e-01 -3.44108343e-02
4.40473780e-02 7.65608728e-01 -8.11717272e-01 -4.80711728e-01
-8.66554445e-04 6.08969092e-01 -8.82423401e-01 8.24072003e-01
-5.89709520e-01 3.62590790e-01 -6.38682604e-01 2.99247235e-01
4.23212826e-01 -3.10372204e-01 3.28292787e-01 3.77334446e-01
2.53793865e-01 4.43633169e-01 -9.08441722e-01 1.54243553e+00
-5.98712802e-01 7.80773640e-01 5.69398254e-02 -5.98625839e-01
1.31943154e+00 5.83689928e-01 1.45255879e-01 -2.80578464e-01
1.09127045e-01 5.65625913e-02 -2.34641343e-01 -5.30780196e-01
8.57219517e-01 -4.38575208e-01 -3.56891900e-01 5.66807866e-01
9.45785046e-02 -3.31537157e-01 -4.36805263e-02 3.66669863e-01
3.71753514e-01 8.31714049e-02 8.07577372e-02 -1.44517675e-01
4.61553603e-01 -1.91115484e-01 5.41555941e-01 8.66638899e-01
-1.03947423e-01 6.79958880e-01 6.88060343e-01 3.54684070e-02
-9.28636968e-01 -1.00927424e+00 9.23127159e-02 1.25742829e+00
1.63005255e-02 -3.21591407e-01 -9.97188330e-01 -3.61189365e-01
-3.33377481e-01 1.21649837e+00 -4.50261295e-01 -7.61086121e-02
-2.79757887e-01 -5.06974399e-01 4.19978708e-01 3.86758000e-01
2.05316931e-01 -1.29284716e+00 -2.48514369e-01 4.37660605e-01
-8.65888953e-01 -5.97497761e-01 -4.39810663e-01 3.14489789e-02
-6.16649806e-01 -3.42688173e-01 -1.02960253e+00 -8.47500384e-01
4.58419681e-01 1.23738550e-01 1.15220773e+00 -3.98898482e-01
3.27804446e-01 2.14660496e-01 -5.31657338e-01 -3.31241220e-01
-7.89223731e-01 1.29559696e-01 3.05971373e-02 7.35347942e-02
7.00931489e-01 -3.90411019e-01 -3.90759110e-01 2.37631246e-01
-7.12185383e-01 2.45034158e-01 3.83672029e-01 1.28004467e+00
4.15456742e-01 -2.18639567e-01 1.03375304e+00 -8.08553696e-01
1.34321380e+00 -6.96393132e-01 -8.71140137e-02 4.09356266e-01
-5.79658389e-01 3.66090834e-01 7.92510390e-01 -7.26374865e-01
-1.48871386e+00 -4.34108675e-01 -3.03180009e-01 -3.32248151e-01
-1.28662243e-01 5.14996767e-01 -1.99530616e-01 7.48115182e-01
5.54220736e-01 4.53377873e-01 5.97500131e-02 -3.66681635e-01
4.23657417e-01 1.16036642e+00 1.69178352e-01 -7.44117975e-01
1.31751865e-01 1.08605810e-01 -7.62713909e-01 -6.74473047e-01
-6.96927726e-01 -2.20001042e-01 -1.29517831e-03 -3.80472541e-01
9.50023115e-01 -9.60696399e-01 -6.43825293e-01 3.76105279e-01
-1.34803665e+00 -2.86873937e-01 -8.43360871e-02 6.08819723e-01
-7.66852021e-01 3.11417520e-01 -6.88382685e-01 -1.15639973e+00
-4.52448815e-01 -1.37508047e+00 1.11450267e+00 3.30506563e-01
-6.95328474e-01 -1.04779863e+00 3.01622003e-01 4.20514792e-01
5.55419981e-01 -7.17274547e-02 7.97490656e-01 -6.22348487e-01
-5.29883087e-01 -1.93749636e-01 2.60988593e-01 3.78815800e-01
-7.57000893e-02 -9.78990737e-03 -1.00031388e+00 -1.68430910e-01
3.15515637e-01 -8.07493925e-01 7.89151609e-01 4.18355972e-01
1.04171824e+00 -3.38529855e-01 -7.58911520e-02 4.01829541e-01
9.92485762e-01 1.40128806e-01 5.60942829e-01 -5.50175980e-02
2.71999836e-01 9.83477652e-01 5.85654140e-01 5.98985553e-01
4.99943763e-01 4.76423651e-01 -5.09665534e-02 2.88655549e-01
3.04119915e-01 -5.83045602e-01 6.91576004e-01 1.45499194e+00
3.00360799e-01 -7.11246789e-01 -4.28538233e-01 6.18467927e-01
-1.81530881e+00 -1.22424400e+00 5.16913794e-02 1.75481796e+00
1.19576848e+00 -6.50558844e-02 -6.39992356e-02 -4.32406336e-01
8.97861421e-01 3.49637866e-01 -4.33682382e-01 -7.13103652e-01
-1.32324815e-01 4.92269956e-02 -1.76827386e-01 6.95425570e-01
-4.21782970e-01 8.58874083e-01 6.96850920e+00 8.59227121e-01
-8.76607537e-01 6.80707395e-02 6.18918478e-01 -1.41188830e-01
-9.71662164e-01 -7.33928010e-02 -9.51004803e-01 6.43186569e-01
1.11737537e+00 -4.17454898e-01 4.09759879e-01 1.18077946e+00
2.07001925e-01 1.36684194e-01 -1.01990986e+00 8.41116905e-01
3.64220262e-01 -1.17885447e+00 2.01332867e-01 -1.94127448e-02
9.70227122e-01 -4.55118299e-01 9.16432664e-02 8.06881011e-01
5.25828600e-01 -1.07666671e+00 6.32346392e-01 6.34865403e-01
7.60096312e-01 -8.36963177e-01 7.35470593e-01 6.99947119e-01
-6.40047729e-01 1.43777668e-01 -5.70438981e-01 -2.00890121e-03
4.24742699e-01 2.17138037e-01 -9.07468081e-01 2.78631330e-01
2.37097338e-01 2.60640085e-01 -2.88347825e-02 4.26471263e-01
-3.82014543e-01 6.92943871e-01 1.40019447e-01 -6.96093440e-01
2.44159475e-01 -3.66610348e-01 5.27358711e-01 1.15376198e+00
3.11659634e-01 8.21405947e-02 1.30890921e-01 1.09077704e+00
2.94638723e-02 6.64916262e-02 -3.90263647e-01 -1.98555738e-01
6.74749672e-01 8.43836486e-01 -8.47177058e-02 -4.95665699e-01
-9.44207162e-02 1.19434536e+00 3.45561713e-01 5.75408697e-01
-8.36644948e-01 -3.54672372e-01 6.26460493e-01 -3.61117601e-01
1.84962034e-01 -9.33885761e-03 -3.55629772e-01 -1.31843174e+00
-9.58562046e-02 -1.15524650e+00 6.37676716e-02 -8.22192550e-01
-1.45025873e+00 1.00201166e+00 9.42080170e-02 -9.92565930e-01
-1.18144977e+00 -1.99758723e-01 -4.63918924e-01 1.34946871e+00
-1.05835211e+00 -8.26316714e-01 -1.40377298e-01 5.41860223e-01
1.14587879e+00 -4.65602189e-01 1.12542379e+00 -6.02599047e-02
-3.71857703e-01 8.02411914e-01 3.04364890e-01 -1.48569658e-01
7.71202624e-01 -1.20513797e+00 2.81744391e-01 4.20707375e-01
-1.66048810e-01 9.77994084e-01 1.04578841e+00 -7.00662076e-01
-1.09608078e+00 -6.59126222e-01 1.28596783e+00 -4.69015509e-01
2.25437462e-01 -5.71255147e-01 -8.77736926e-01 5.80894887e-01
5.55475116e-01 -1.02909124e+00 9.99882340e-01 2.48994052e-01
-1.35233685e-01 3.91054690e-01 -8.77657712e-01 7.38385797e-01
5.24228990e-01 -7.27670014e-01 -8.08537781e-01 2.10955590e-01
9.72791195e-01 -1.57549903e-01 -5.88375807e-01 -4.22769152e-02
7.11574674e-01 -9.60878193e-01 6.12874329e-01 -5.48033476e-01
7.82779932e-01 1.76852152e-01 -2.31242657e-01 -1.58142221e+00
-3.10907483e-01 -8.39013815e-01 -1.73767909e-01 1.23389995e+00
4.17713165e-01 -4.85376894e-01 8.25741887e-01 8.24627101e-01
6.10954650e-02 -6.77110493e-01 -5.73724627e-01 -4.28525716e-01
9.61123630e-02 -2.79500306e-01 6.54609978e-01 9.05550838e-01
8.41242224e-02 6.82565749e-01 -9.13096786e-01 -3.91210198e-01
4.90114987e-01 2.39960954e-01 9.72178698e-01 -9.75725472e-01
-6.81991458e-01 -2.39893883e-01 2.23742947e-01 -1.67662036e+00
5.51170647e-01 -6.15027010e-01 3.65740687e-01 -1.48585248e+00
3.19206119e-01 -1.87572882e-01 2.53689336e-03 -9.99204218e-02
-5.81846774e-01 -3.61377686e-01 -4.93057109e-02 1.87401429e-01
-4.12437290e-01 1.16364026e+00 1.30151689e+00 1.54762939e-02
-4.03298378e-01 1.53365850e-01 -8.05511713e-01 2.80926377e-01
6.89237297e-01 -5.76952815e-01 -7.32273042e-01 -3.78387898e-01
6.29529580e-02 7.39897370e-01 -9.36632305e-02 -2.61461854e-01
1.78118721e-01 -4.36842412e-01 2.17465609e-01 -7.43955135e-01
6.32756770e-01 -2.98820227e-01 2.65611380e-01 3.44436876e-02
-9.54352260e-01 -4.59555387e-02 -3.45492125e-01 6.86564922e-01
-5.05655766e-01 -3.15988839e-01 5.59696078e-01 -3.41792077e-01
-3.98336560e-01 -5.01347147e-02 -6.60628557e-01 -4.42152768e-02
7.20983863e-01 -2.54196286e-01 5.25965020e-02 -1.15472901e+00
-5.29210746e-01 2.22162440e-01 3.25306475e-01 6.86021805e-01
7.88289845e-01 -1.36465693e+00 -1.03741801e+00 2.41083130e-01
1.12954438e-01 -2.64475405e-01 4.73856419e-01 2.30290771e-01
-2.66074359e-01 4.36743349e-01 -1.91221818e-01 -5.24897993e-01
-1.02676260e+00 2.26894036e-01 1.72554418e-01 -2.77755797e-01
-3.34410906e-01 1.15064657e+00 2.86355197e-01 -5.90219140e-01
3.77077043e-01 2.14362413e-01 -2.74183184e-01 1.75464079e-01
5.08931637e-01 1.25307024e-01 -4.85911220e-01 -4.78539884e-01
3.96389887e-02 -2.12816581e-01 -1.74884573e-01 -6.91063881e-01
1.13390350e+00 -3.61339241e-01 1.93220507e-02 6.83912873e-01
1.30205083e+00 6.77151838e-03 -1.41610956e+00 -1.90624312e-01
-2.84245640e-01 -5.82831204e-01 -1.75132573e-01 -5.98626852e-01
-5.27473569e-01 9.20227468e-01 2.83963997e-02 3.68611097e-01
8.03184569e-01 -4.49360013e-02 9.18471038e-01 2.45918334e-01
3.57319683e-01 -1.29187095e+00 3.33804101e-01 5.84162831e-01
9.78306949e-01 -1.09829080e+00 -3.89284492e-01 -5.27406074e-02
-1.42701471e+00 9.69957769e-01 8.01373601e-01 -1.43796410e-02
1.13169551e-01 1.39471120e-03 3.94406199e-01 2.83851177e-02
-1.40423512e+00 1.71614423e-01 1.43220916e-01 5.85486770e-01
7.98830748e-01 3.53620261e-01 -4.04461086e-01 7.27469087e-01
-5.92942715e-01 -2.84957170e-01 6.09875858e-01 4.17888641e-01
-5.15625656e-01 -1.23886359e+00 -2.85954148e-01 2.18973994e-01
-3.48100632e-01 -5.13160601e-02 -6.18591666e-01 3.41177434e-01
-4.35160697e-01 1.34724641e+00 -3.11880317e-02 -4.29614365e-01
-7.00649992e-02 4.66457486e-01 1.58471912e-01 -5.72544336e-01
-5.12501299e-01 2.93498397e-01 2.47995079e-01 -1.81967199e-01
-2.68181533e-01 -6.83734775e-01 -7.58307040e-01 -3.05039495e-01
-6.46306515e-01 6.22938693e-01 7.38079369e-01 8.10796082e-01
1.84318766e-01 4.95462388e-01 9.34782505e-01 -6.01018667e-01
-1.13103342e+00 -1.25156629e+00 -5.11076927e-01 5.12931168e-01
1.61035314e-01 -4.49768841e-01 -5.48963547e-01 -3.97719033e-02]
|
[12.559650421142578, 8.376569747924805]
|
7c880f49-1505-45c9-a5cb-3b113cd49609
|
physics-informed-neural-networks-for-pathloss
|
2211.12986
| null |
https://arxiv.org/abs/2211.12986v1
|
https://arxiv.org/pdf/2211.12986v1.pdf
|
Physics-informed neural networks for pathloss prediction
|
This paper introduces a physics-informed machine learning approach for pathloss prediction. This is achieved by including in the training phase simultaneously (i) physical dependencies between spatial loss field and (ii) measured pathloss values in the field. It is shown that the solution to a proposed learning problem improves generalization and prediction quality with a small number of neural network layers and parameters. The latter leads to fast inference times which are favorable for downstream tasks such as localization. Moreover, the physics-informed formulation allows training and prediction with small amount of training data which makes it appealing for a wide range of practical pathloss prediction scenarios.
|
['Nicola Michailow', 'Alberto Martinez Alba', 'Steffen Limmer']
|
2022-11-23
| null | null | null | null |
['physics-informed-machine-learning']
|
['graphs']
|
[ 7.11874589e-02 1.43739164e-01 -5.19440055e-01 -5.87104380e-01
-6.72093987e-01 -7.66485408e-02 1.79800212e-01 5.30359924e-01
-3.20768595e-01 1.06965733e+00 -4.43414599e-01 -9.03840423e-01
-8.93159509e-01 -9.02682245e-01 -6.28150344e-01 -8.27522993e-01
-6.73533022e-01 3.90203744e-01 5.40392816e-01 -1.29554227e-01
1.13641664e-01 1.06641531e+00 -9.35878754e-01 -3.96254987e-01
7.77019382e-01 1.54174066e+00 2.47873977e-01 7.59131789e-01
3.15043569e-01 8.46037090e-01 -2.80183345e-01 -2.22556368e-01
5.81250042e-02 6.71178326e-02 -6.79512024e-01 -6.01269543e-01
1.69822693e-01 -3.74948502e-01 -7.51748502e-01 4.59880978e-01
5.79788566e-01 3.69209319e-01 6.95958316e-01 -1.10695040e+00
3.60821873e-01 4.79002416e-01 -2.99226791e-01 4.47647393e-01
-2.59914696e-01 -1.60879359e-01 8.89915705e-01 -9.93086472e-02
2.19204709e-01 7.64003515e-01 1.16421092e+00 -9.40121859e-02
-1.10465777e+00 -5.23873627e-01 -7.61871487e-02 4.51085985e-01
-1.30470622e+00 -2.24857852e-01 8.50827277e-01 -1.79563358e-01
9.46802795e-01 1.42606482e-01 4.15011287e-01 5.71583927e-01
4.85451102e-01 3.49339485e-01 6.74633503e-01 -6.26061141e-01
3.93006057e-01 5.43860972e-01 1.63559839e-01 6.06153190e-01
7.51892999e-02 6.37319505e-01 -1.38991535e-01 -7.65157342e-02
8.14789176e-01 -3.92487615e-01 -1.79647759e-01 -3.21034104e-01
-5.16562581e-01 7.00653851e-01 1.20317447e+00 3.60409683e-03
-2.37340227e-01 4.51794773e-01 1.86065391e-01 3.18305552e-01
2.31550187e-01 4.66641337e-01 -4.17638332e-01 1.40998811e-01
-9.60001588e-01 2.39665154e-02 7.31068432e-01 9.90076303e-01
8.57291043e-01 1.08510412e-01 6.78308979e-02 7.14142978e-01
4.29780424e-01 8.34873438e-01 -3.19975555e-01 -1.06117833e+00
4.80893373e-01 -2.32505977e-01 3.84046108e-01 -1.16785717e+00
-1.02405548e+00 -9.82465446e-01 -7.55269706e-01 3.13093603e-01
4.91999865e-01 -3.69696498e-01 -5.64362705e-01 1.72392666e+00
2.87008226e-01 3.05053473e-01 -2.90889516e-02 6.19872689e-01
5.26987135e-01 9.51144695e-01 1.95965931e-01 -1.48603871e-01
5.62774479e-01 -8.55322301e-01 -1.68549731e-01 -2.52013743e-01
9.71769154e-01 -4.27309453e-01 4.28804129e-01 4.24834222e-01
-9.16052461e-01 -3.96711648e-01 -1.13023758e+00 7.69086257e-02
-1.95426807e-01 9.05809850e-02 9.06317174e-01 8.99813771e-01
-1.04708493e+00 1.06228769e+00 -7.70068228e-01 -4.72948968e-01
4.15062755e-01 7.01544940e-01 2.37572253e-01 7.43175149e-02
-1.03856492e+00 7.32859731e-01 3.10037673e-01 1.96022555e-01
-6.00179493e-01 -1.00651288e+00 -4.65494841e-01 2.57399768e-01
-4.16880213e-02 -3.97410363e-01 1.21114933e+00 -4.18841898e-01
-1.55450439e+00 1.26875043e-01 -1.40748248e-01 -7.33580410e-01
4.05511528e-01 -1.75707564e-01 -7.08333552e-01 4.02841568e-01
-1.33767307e-01 3.79345447e-01 6.73866391e-01 -1.23453724e+00
-6.79349899e-01 -7.51913115e-02 4.48985286e-02 -1.09994933e-01
-2.27146521e-01 -3.60917091e-01 -3.48565370e-01 -1.57781005e-01
1.01180613e-01 -6.84305668e-01 -6.83028460e-01 4.69089746e-01
-5.82777679e-01 5.04942536e-01 7.22616374e-01 -6.55191720e-01
9.36954379e-01 -1.61707771e+00 -4.81880993e-01 9.37598169e-01
2.18856633e-01 2.03663394e-01 -1.66280806e-01 5.63790977e-01
1.52511939e-01 -1.92812428e-01 1.15734778e-01 -5.32523617e-02
-3.28326315e-01 -1.44271348e-02 -4.38512981e-01 4.81091231e-01
-3.10313612e-01 6.29509687e-01 -6.59886420e-01 -4.80242342e-01
5.24758875e-01 4.50673163e-01 -6.37541890e-01 9.47759375e-02
8.16206932e-02 9.15177822e-01 -9.45638955e-01 2.99087405e-01
8.60767603e-01 -3.40243131e-01 -3.00933383e-02 -2.13755623e-01
4.12691720e-02 6.36924207e-01 -7.58334875e-01 1.33248234e+00
-1.20969009e+00 9.10671771e-01 5.50175384e-02 -1.34111512e+00
8.73702705e-01 -1.21081360e-02 7.77036071e-01 -8.11505497e-01
3.22244287e-01 1.72251344e-01 -1.19464479e-01 -3.54459822e-01
2.01337054e-01 -3.69445264e-01 1.26357704e-01 2.47778982e-01
2.70633399e-01 4.90892865e-02 -2.02394083e-01 2.49408707e-01
1.21119654e+00 -2.52994806e-01 1.13483444e-01 -3.53354096e-01
5.02096117e-01 -1.90164313e-01 4.33954507e-01 1.08413911e+00
-1.34974942e-01 -1.50760800e-01 2.23423943e-01 -3.45653951e-01
-7.56161749e-01 -1.44408464e+00 -5.53829432e-01 6.99563503e-01
6.68832362e-01 2.28421111e-02 -2.67334253e-01 -4.36062664e-01
3.89883630e-02 7.85960436e-01 -4.11078259e-02 -3.92004251e-01
-5.97350597e-01 -8.38293850e-01 5.39885998e-01 4.64869261e-01
5.02246618e-01 -4.13949162e-01 -2.35328287e-01 4.34043199e-01
-2.91204378e-02 -1.51820695e+00 4.51764822e-01 4.50037450e-01
-1.16502655e+00 -7.75571048e-01 -1.29428923e-01 -6.02254450e-01
3.17802489e-01 1.82788178e-01 9.72294211e-01 7.79853240e-02
-3.61378878e-01 3.55856687e-01 -3.02184641e-01 -2.77967334e-01
-3.02081436e-01 2.22704455e-01 4.82559577e-03 -2.80803800e-01
-2.35292360e-01 -1.16947711e+00 -6.35750055e-01 3.27818185e-01
-5.93673550e-02 -1.99180692e-01 7.99855292e-01 5.54597974e-01
4.73325938e-01 6.15167737e-01 6.73263550e-01 -5.72417498e-01
9.97060165e-02 -5.66484034e-01 -1.11487949e+00 1.55813381e-01
-6.41420960e-01 1.02843851e-01 9.67193961e-01 -1.37254283e-01
-1.24627745e+00 -2.05390602e-01 -4.62115496e-01 1.91331029e-01
-1.31173417e-01 2.41573602e-01 -2.06732050e-01 -9.89859402e-01
6.44537032e-01 -1.82566851e-01 -3.94091278e-01 -5.94822943e-01
1.20486334e-01 4.07686591e-01 3.05158168e-01 -5.02515376e-01
9.42221761e-01 4.53342497e-01 9.50793087e-01 -1.40541172e+00
-8.04643035e-01 -4.19592112e-01 -6.92490220e-01 -4.08797592e-01
2.50506699e-01 -6.48426533e-01 -8.05238962e-01 1.16157442e-01
-9.36056316e-01 -7.29904652e-01 -1.21125124e-01 1.02666199e+00
-6.94291234e-01 2.04472736e-01 -5.34700036e-01 -9.47914958e-01
-1.18064031e-01 -6.10987842e-01 6.18846595e-01 2.46302769e-01
1.46002024e-01 -1.62010729e+00 -1.24452062e-01 -5.46454377e-02
4.24730241e-01 1.35896623e-01 1.18623936e+00 -1.08175628e-01
-9.09610152e-01 -4.86314893e-01 -6.88800156e-01 5.10585122e-02
-4.28338021e-01 -4.28054452e-01 -1.07890224e+00 -2.55517781e-01
-1.56420305e-01 -2.26028040e-01 7.52194464e-01 1.05221236e+00
1.43310678e+00 -1.07939720e-01 -1.01170647e+00 9.35317755e-01
1.72212970e+00 1.53637044e-02 4.13777709e-01 -7.61496499e-02
3.87608916e-01 4.11828041e-01 4.70301598e-01 3.55838418e-01
2.57283486e-02 7.98690856e-01 4.92003351e-01 -2.64657617e-01
-6.10073842e-02 -4.02543336e-01 -2.65404940e-01 3.32509965e-01
2.34508470e-01 -7.49405384e-01 -8.30339730e-01 2.01556668e-01
-1.49622846e+00 -8.62308800e-01 -4.21075612e-01 2.16803861e+00
1.03743769e-01 3.93254995e-01 -1.21998645e-01 1.88723385e-01
3.77057821e-01 -7.65853822e-02 -6.26491606e-01 -3.49329680e-01
3.70051235e-01 1.99424699e-01 1.11692226e+00 1.10534430e+00
-1.07956350e+00 6.14506364e-01 7.54860783e+00 1.01003540e+00
-1.36824334e+00 1.02842085e-01 5.87431669e-01 2.92057190e-02
-1.80097502e-02 1.43568292e-01 -9.30837810e-01 2.45186493e-01
1.19852793e+00 2.28215545e-01 1.89537704e-01 5.90820789e-01
6.08033955e-01 -5.37088811e-01 -7.76809216e-01 7.07342207e-01
-5.57621062e-01 -1.34519446e+00 -1.67732179e-01 4.81828041e-02
2.84453899e-01 4.12688792e-01 -1.19063102e-01 2.01332211e-01
9.59160253e-02 -7.78989732e-01 3.03813249e-01 4.88162875e-01
6.69542372e-01 -6.97035432e-01 6.61903381e-01 6.21250987e-01
-9.93036628e-01 -4.84390199e-01 -4.87599015e-01 -1.00773290e-01
7.59618402e-01 1.26774943e+00 -7.91032493e-01 7.35157311e-01
6.76510453e-01 6.86113298e-01 -3.35206479e-01 1.79839849e+00
-1.62542373e-01 1.09987664e+00 -1.00484896e+00 -4.89166118e-02
2.28104562e-01 -1.13459937e-02 5.98754168e-01 1.11962593e+00
4.78331357e-01 -2.78987712e-03 1.25017669e-03 6.45823598e-01
2.02473462e-01 -1.06053829e-01 -3.29101503e-01 4.46990460e-01
4.19022322e-01 1.08633900e+00 -6.57095492e-01 5.23486614e-01
-1.48477018e-01 5.46250880e-01 2.73589462e-01 7.74969995e-01
-8.47980261e-01 -6.58663332e-01 4.54949051e-01 6.14897430e-01
1.27041966e-01 -2.48710737e-01 -4.10221726e-01 -4.16716814e-01
-6.02279343e-02 5.24426043e-01 1.15227625e-01 -4.66666639e-01
-1.00490189e+00 5.21810412e-01 2.51519650e-01 -1.27418613e+00
-4.43862714e-02 -7.98234344e-01 -9.48967099e-01 7.48142838e-01
-2.04015326e+00 -1.04262817e+00 -1.25394687e-01 1.92912191e-01
-2.86731660e-01 -6.94685951e-02 6.69346392e-01 7.21735060e-01
-5.80512762e-01 7.24288583e-01 6.59570694e-01 -2.18185306e-01
3.81610006e-01 -8.66706371e-01 8.33376199e-02 5.56156158e-01
-1.64984331e-01 -8.37302010e-04 1.05143976e+00 -3.69136244e-01
-9.35420871e-01 -1.24731171e+00 9.43051100e-01 -1.10303476e-01
6.31244302e-01 -1.62964866e-01 -5.46138287e-01 4.34514314e-01
-4.98384476e-01 3.19865555e-01 5.06711483e-01 3.70440066e-01
1.31178930e-01 -6.35294259e-01 -1.32703304e+00 1.23620987e-01
8.52726400e-01 -2.02212930e-01 5.39075196e-01 6.11290455e-01
2.97304571e-01 -2.56741256e-01 -8.36739242e-01 3.66984129e-01
5.48231661e-01 -7.86864161e-01 1.30253088e+00 -3.80646735e-01
1.03746159e-02 2.37560451e-01 -4.38370705e-02 -1.06071329e+00
-3.51455390e-01 -6.21964633e-01 -2.47561246e-01 9.35950756e-01
7.75770128e-01 -6.17144287e-01 1.10575581e+00 2.75763631e-01
-3.30995061e-02 -1.09591305e+00 -1.21756673e+00 -1.11263072e+00
2.61955410e-01 -8.89308989e-01 1.90265685e-01 1.12730287e-01
-1.12603493e-01 2.50821143e-01 -4.22588229e-01 8.04408669e-01
9.45476651e-01 2.87285037e-02 3.83455694e-01 -1.51192451e+00
-5.04536271e-01 -2.35448569e-01 -5.93579471e-01 -1.70581210e+00
7.79211000e-02 -8.32087278e-01 8.36800560e-02 -1.37662911e+00
-3.67365956e-01 -1.42848945e+00 -4.23579723e-01 -4.48177829e-02
3.36071223e-01 1.36142552e-01 -3.39744866e-01 -1.45505778e-02
-4.34040934e-01 5.92959344e-01 9.18474555e-01 2.84895953e-02
-2.86232531e-01 8.36705446e-01 -2.99790651e-01 6.44962311e-01
1.08733141e+00 -6.35531783e-01 -7.74693072e-01 -5.03856063e-01
1.41579201e-02 4.84158546e-01 8.59022200e-01 -1.57540905e+00
1.11852042e-01 -1.40050933e-01 4.68532145e-01 -5.33233643e-01
6.97744131e-01 -9.80215013e-01 -3.81053418e-01 4.58626151e-01
-2.83509851e-01 -8.45501125e-01 4.34958249e-01 9.10198927e-01
1.32346272e-01 -2.22553849e-01 1.13231194e+00 5.75475991e-01
-7.13193536e-01 4.27314997e-01 -4.17368442e-01 -2.58263081e-01
8.65786791e-01 -1.56156659e-01 9.44554657e-02 -9.38111126e-01
-8.02472353e-01 2.90717900e-01 -1.30867347e-01 -1.91684559e-01
2.45107502e-01 -1.14407766e+00 -1.67045578e-01 1.72283384e-04
-1.24825090e-01 -4.34408873e-01 4.06529725e-01 1.05342603e+00
-7.25500703e-01 6.72429681e-01 -7.20607862e-02 -6.71000183e-01
-6.96391463e-01 1.71547189e-01 7.79269934e-01 -5.08094728e-01
-7.92765439e-01 1.14839864e+00 -1.23804860e-01 -5.02328098e-01
3.25244367e-01 -1.73044980e-01 2.19664052e-02 -6.83548987e-01
3.44133824e-01 6.65474951e-01 9.67664868e-02 -6.56099558e-01
-2.67486274e-01 6.61476731e-01 2.19198808e-01 6.45527169e-02
1.18969786e+00 -6.00423396e-01 3.30848932e-01 1.05764993e-01
1.42637789e+00 -1.68475926e-01 -1.41536593e+00 -3.84764344e-01
-1.48548871e-01 -2.77881175e-01 8.66652727e-01 -1.00157726e+00
-1.11867774e+00 1.18511295e+00 8.14345300e-01 -6.49834722e-02
1.10137773e+00 -1.12627730e-01 1.09150434e+00 8.81510675e-01
7.28337228e-01 -1.00861895e+00 -3.78649861e-01 2.84104854e-01
3.21013242e-01 -1.18377316e+00 1.91154599e-01 -9.66462374e-01
1.10431023e-01 9.56205487e-01 3.16560060e-01 -6.68720901e-02
1.50599396e+00 2.78731674e-01 -1.02941908e-01 2.24947050e-01
-3.32215458e-01 1.80500701e-01 2.16838628e-01 1.00318933e+00
7.81507939e-02 -1.31834313e-01 1.27937093e-01 4.66270387e-01
-1.00121066e-01 -2.45390832e-01 1.20085135e-01 7.71740556e-01
-8.42658818e-01 -1.08480489e+00 -1.57611277e-02 6.96470261e-01
4.92093191e-02 2.11604938e-01 2.05614269e-01 8.07852566e-01
-3.45919169e-02 1.25141025e+00 -2.00601980e-01 -3.29042405e-01
-5.01994453e-02 -5.52740633e-01 5.78297079e-01 -2.00437531e-01
1.17847109e-02 -3.71263742e-01 1.81250200e-01 -8.36480439e-01
4.79436293e-02 -3.45661223e-01 -1.42022538e+00 -3.53950143e-01
-4.51409787e-01 4.08506036e-01 5.55857003e-01 1.17635024e+00
2.47308284e-01 5.74011147e-01 8.13483596e-01 -8.94072354e-01
-3.58411551e-01 -5.81776023e-01 -1.05613041e+00 -3.78637224e-01
5.75651944e-01 -8.75736773e-01 -2.69836307e-01 -4.64554995e-01]
|
[6.199511528015137, 1.4216413497924805]
|
73c62e71-86d3-4e05-b142-2c92d34654bf
|
t3l-translate-and-test-transfer-learning-for
|
2306.04996
| null |
https://arxiv.org/abs/2306.04996v1
|
https://arxiv.org/pdf/2306.04996v1.pdf
|
T3L: Translate-and-Test Transfer Learning for Cross-Lingual Text Classification
|
Cross-lingual text classification leverages text classifiers trained in a high-resource language to perform text classification in other languages with no or minimal fine-tuning (zero/few-shots cross-lingual transfer). Nowadays, cross-lingual text classifiers are typically built on large-scale, multilingual language models (LMs) pretrained on a variety of languages of interest. However, the performance of these models vary significantly across languages and classification tasks, suggesting that the superposition of the language modelling and classification tasks is not always effective. For this reason, in this paper we propose revisiting the classic "translate-and-test" pipeline to neatly separate the translation and classification stages. The proposed approach couples 1) a neural machine translator translating from the targeted language to a high-resource language, with 2) a text classifier trained in the high-resource language, but the neural machine translator generates "soft" translations to permit end-to-end backpropagation during fine-tuning of the pipeline. Extensive experiments have been carried out over three cross-lingual text classification datasets (XNLI, MLDoc and MultiEURLEX), with the results showing that the proposed approach has significantly improved performance over a competitive baseline.
|
['Massimo Piccardi', 'Gholamreza Haffari', 'Inigo Jauregi Unanue']
|
2023-06-08
| null | null | null | null |
['cross-lingual-transfer']
|
['natural-language-processing']
|
[ 2.10787822e-02 -2.34467447e-01 -3.99275303e-01 -6.05627060e-01
-1.23142123e+00 -6.76135957e-01 9.61907506e-01 1.47089317e-01
-8.40750277e-01 8.32164466e-01 -7.08427578e-02 -6.46784306e-01
5.83713531e-01 -4.86407369e-01 -8.05424750e-01 -4.52997059e-01
5.20934403e-01 9.11050856e-01 -8.88282806e-02 -2.58921921e-01
-1.63030550e-01 -1.21385418e-02 -1.01430464e+00 8.32578003e-01
1.16923296e+00 8.13959837e-01 2.48792604e-01 5.68642378e-01
-5.07363141e-01 5.96158564e-01 -4.05435741e-01 -9.21171129e-01
1.34508863e-01 -4.04638112e-01 -8.70863974e-01 -1.66173220e-01
4.72783089e-01 2.09017128e-01 1.65510252e-01 9.85426009e-01
4.52736914e-01 -1.20043352e-01 6.40239239e-01 -7.28964686e-01
-7.82864928e-01 1.03599322e+00 -4.13175166e-01 -6.40230253e-02
2.15665996e-01 4.44480777e-02 1.03169215e+00 -1.08470047e+00
6.57806516e-01 1.29678953e+00 7.39615619e-01 3.53641093e-01
-1.37785745e+00 -6.93906307e-01 1.40181288e-01 -1.58795670e-01
-1.29735875e+00 -6.53117895e-01 5.05497158e-01 -5.73152959e-01
1.44188452e+00 -1.13486378e-02 9.56669226e-02 1.42665422e+00
7.02357292e-01 8.23304057e-01 1.33266306e+00 -8.36359918e-01
-8.71670842e-02 8.56245697e-01 -5.98711893e-02 6.41553044e-01
-1.25775337e-01 -3.30597237e-02 -6.37202382e-01 1.60305783e-01
-1.96599305e-01 -4.04146731e-01 -1.10461421e-01 -7.22084045e-02
-1.21061456e+00 8.80797267e-01 3.70702505e-01 6.27822638e-01
-1.34314790e-01 -2.27638915e-01 9.85223413e-01 7.01727271e-01
1.10228848e+00 3.09877992e-01 -8.00578356e-01 1.96532230e-03
-1.11737740e+00 -3.05685639e-01 9.19157088e-01 1.04207563e+00
7.66615987e-01 -2.28388086e-02 -1.23807073e-01 1.04253829e+00
1.51045382e-01 4.53511149e-01 1.09947610e+00 3.07417691e-01
1.18148804e+00 6.32534981e-01 -3.09670240e-01 -3.25948775e-01
-2.02684045e-01 -6.19034588e-01 -8.37966323e-01 -5.33269532e-02
1.82566747e-01 -3.52690816e-01 -7.75358856e-01 1.49666452e+00
5.41272908e-02 -4.18016762e-01 5.14985859e-01 5.40229201e-01
5.01547694e-01 7.62027621e-01 3.67042333e-01 5.62829599e-02
1.29558969e+00 -1.38130140e+00 -4.78189349e-01 -5.90072334e-01
1.28933835e+00 -1.15083039e+00 1.54217291e+00 7.65675977e-02
-9.12159204e-01 -6.79951310e-01 -1.07630241e+00 -4.18584764e-01
-1.04829609e+00 7.31167376e-01 7.07427412e-02 6.04943514e-01
-9.62502420e-01 4.07836944e-01 -6.37714326e-01 -6.88674033e-01
8.35681632e-02 1.92961544e-01 -3.91882479e-01 -1.45851389e-01
-1.32124197e+00 1.27496147e+00 7.43934572e-01 1.12521453e-02
-7.37342477e-01 -4.13319796e-01 -6.50275052e-01 -2.46888191e-01
-3.59057486e-02 -5.56239069e-01 1.20674670e+00 -1.48257446e+00
-1.74159718e+00 1.32812500e+00 -3.91930342e-02 -3.87478322e-01
1.01992869e+00 -3.65514398e-01 -3.99594069e-01 -5.27095616e-01
3.92142355e-01 4.52999860e-01 5.74858487e-01 -8.10168386e-01
-8.70987594e-01 -3.33396852e-01 -2.49639183e-01 3.49527568e-01
-5.25786281e-01 4.20921564e-01 -3.72807592e-01 -6.95148408e-01
-4.13328797e-01 -9.62509811e-01 2.18357041e-01 -5.25112092e-01
-4.41834986e-01 -4.40491021e-01 6.76849246e-01 -7.89465904e-01
1.03813410e+00 -1.89962852e+00 2.75552988e-01 -2.38997638e-01
-4.85625833e-01 2.75033414e-01 -1.82485357e-01 6.11019671e-01
-8.69305208e-02 -2.61009987e-02 -3.86709124e-02 -8.76666188e-01
1.35155544e-02 1.37144580e-01 -3.16887230e-01 4.54504728e-01
1.73975542e-01 1.07155240e+00 -6.89386904e-01 -3.88051897e-01
-4.98070661e-03 4.78104860e-01 -2.49741927e-01 2.55746394e-01
-3.03579807e-01 5.20748317e-01 -3.47650766e-01 5.00939548e-01
2.44979307e-01 -1.82532091e-02 3.08238447e-01 -7.31753856e-02
-4.11101699e-01 8.27160358e-01 -5.63027442e-01 2.00086856e+00
-9.90207255e-01 5.65014958e-01 -1.32760644e-01 -9.54051077e-01
8.89697075e-01 5.89757919e-01 2.34137420e-02 -6.91085279e-01
3.81318182e-01 7.01761782e-01 -1.50197759e-01 -4.14122671e-01
3.87289405e-01 -3.65201980e-01 -4.63036329e-01 7.07356811e-01
4.29652125e-01 1.48296893e-01 1.52964681e-01 -2.78466851e-01
5.96928060e-01 4.91870970e-01 2.71208018e-01 -4.41548973e-01
8.67315173e-01 3.42175663e-01 3.90719548e-02 5.28112292e-01
1.11955432e-02 1.11674275e-02 4.55650315e-02 -3.95236254e-01
-9.98329818e-01 -5.49546301e-01 -1.63255736e-01 1.86610258e+00
-4.70805049e-01 -3.30484480e-01 -8.67122829e-01 -9.57912266e-01
-4.65643778e-02 8.27827096e-01 -5.62998056e-01 -9.29016843e-02
-5.81704915e-01 -9.05619562e-01 9.06262159e-01 2.85706699e-01
3.02212179e-01 -1.06184399e+00 -2.56491844e-02 3.54437053e-01
-2.31064841e-01 -1.33974016e+00 -6.60631955e-01 7.39238858e-01
-6.56143963e-01 -5.73629498e-01 -6.04551017e-01 -1.19168782e+00
5.62630951e-01 -1.18731938e-01 1.28823018e+00 -3.77541810e-01
1.83852330e-01 -9.91934538e-02 -4.68683600e-01 -2.01065749e-01
-9.56254363e-01 9.89324391e-01 1.53683603e-01 1.97592825e-01
6.30491674e-01 -1.34554490e-01 2.79717833e-01 1.89527616e-01
-5.37727296e-01 3.34075809e-01 6.69690371e-01 9.76793587e-01
5.00604451e-01 -6.75826222e-02 5.69296539e-01 -1.25448883e+00
8.44241738e-01 -4.65135634e-01 -5.81730485e-01 6.57610416e-01
-6.69191062e-01 2.43333220e-01 1.17863846e+00 -4.95377302e-01
-1.03898156e+00 -1.44038722e-02 -6.18246458e-02 -2.99527377e-01
-1.43340334e-01 8.82013798e-01 -3.45765293e-01 1.53269798e-01
7.53233373e-01 4.13763225e-01 -4.76856500e-01 -6.87862635e-01
5.01854300e-01 1.05424321e+00 3.40167880e-01 -6.61244810e-01
6.54097378e-01 -6.86718151e-02 -5.79785407e-01 -4.10490304e-01
-9.77703154e-01 -1.26640052e-01 -1.29972363e+00 7.09633976e-02
9.75110173e-01 -1.10157371e+00 -1.51864529e-01 4.40022200e-01
-1.18374550e+00 -5.21035016e-01 -2.88440604e-02 6.04373634e-01
-3.27159911e-01 -3.05216819e-01 -7.91693926e-01 -4.38303053e-01
-6.55464053e-01 -1.39651811e+00 1.00368226e+00 -1.03369378e-01
-1.14159286e-01 -1.46459186e+00 1.79562688e-01 4.04523730e-01
5.20968318e-01 -3.03385705e-01 1.05085683e+00 -1.01702464e+00
-1.57931641e-01 -4.53510016e-01 -7.72237331e-02 3.92827421e-01
1.18469104e-01 -1.54702455e-01 -1.25139773e+00 -5.10782003e-01
-1.10369332e-01 -7.43039668e-01 5.15059769e-01 -4.21531349e-02
6.19714916e-01 -2.49682829e-01 -3.06972176e-01 7.09288657e-01
1.38764203e+00 -1.46792963e-01 -9.64552909e-03 4.95616615e-01
7.86820590e-01 8.16991210e-01 3.94331425e-01 -1.65452972e-01
5.57564616e-01 7.86469698e-01 -1.97204903e-01 -1.76173300e-01
-7.47682750e-02 -3.80237848e-01 8.64522696e-01 1.43041801e+00
3.42859060e-01 -2.91724563e-01 -1.12485385e+00 2.55748868e-01
-1.68045783e+00 -4.09999639e-01 3.63100469e-02 2.09258914e+00
1.26786816e+00 3.43927205e-01 -1.13589026e-01 -3.02300096e-01
5.84145606e-01 -1.17939040e-01 -5.10121882e-01 -7.40717828e-01
-2.63566732e-01 1.45874605e-01 4.80079472e-01 6.83453441e-01
-1.25599813e+00 1.58915865e+00 5.37998724e+00 9.78717089e-01
-1.74948823e+00 6.18257463e-01 5.60519636e-01 -7.51682883e-03
1.07203998e-01 -5.28933071e-02 -1.21678817e+00 4.76787955e-01
1.43263030e+00 -2.57950127e-01 3.94641072e-01 9.13808763e-01
4.62619998e-02 3.28248382e-01 -1.48231363e+00 6.23299539e-01
1.86503679e-01 -9.47341979e-01 2.19297886e-01 -2.10851878e-01
6.75769329e-01 6.99722707e-01 -3.80353518e-02 8.00235927e-01
4.85415131e-01 -9.33797061e-01 9.04572129e-01 1.96840927e-01
1.14170051e+00 -6.19134307e-01 7.79757440e-01 7.05332339e-01
-1.19348180e+00 3.24885845e-01 -3.37701142e-01 2.33979002e-01
2.56359037e-02 2.41345555e-01 -1.01587594e+00 7.45344937e-01
5.73730350e-01 8.61455023e-01 -7.17548490e-01 2.28631839e-01
-1.98687091e-01 5.04147708e-01 -5.45166992e-02 -4.99439463e-02
4.95573491e-01 -2.01605231e-01 1.47929460e-01 1.72953141e+00
3.44401389e-01 -7.70140588e-01 5.58266819e-01 6.14198506e-01
-4.41481948e-01 7.92724133e-01 -5.66319466e-01 -1.95875287e-01
-1.64296255e-02 1.29850793e+00 -4.48646396e-01 -6.19569123e-01
-6.89813077e-01 1.21028662e+00 6.95461452e-01 1.20218106e-01
-6.71147346e-01 -3.58181298e-01 2.60394454e-01 -2.83804163e-02
-9.89070758e-02 -2.69932151e-02 -3.44724059e-01 -1.47269130e+00
1.00430772e-01 -1.08011770e+00 4.23690289e-01 -6.29691541e-01
-1.57705665e+00 1.18761969e+00 -3.67740989e-01 -1.20477080e+00
-5.18935978e-01 -8.18880022e-01 -3.46957862e-01 1.44601321e+00
-1.74538195e+00 -1.79824722e+00 1.70309544e-01 6.83025479e-01
9.44993436e-01 -5.96441865e-01 1.08325815e+00 7.20758200e-01
-7.74017870e-01 1.04604626e+00 4.56157804e-01 3.12697262e-01
1.11514652e+00 -1.16228592e+00 4.54752922e-01 7.23062932e-01
5.96394800e-02 3.71478617e-01 3.69795442e-01 -6.34724200e-01
-1.26546574e+00 -1.46150231e+00 1.50974298e+00 -4.93074358e-01
9.95662868e-01 -8.05090547e-01 -1.07790959e+00 9.58846569e-01
5.65854192e-01 -1.62106976e-01 7.53254652e-01 2.12106287e-01
-5.11069596e-01 -4.41969112e-02 -7.79014707e-01 5.31812787e-01
4.44357693e-01 -9.54605818e-01 -5.33740282e-01 6.90630794e-01
5.52350700e-01 -2.91435391e-01 -7.99037337e-01 2.25309525e-02
5.33443511e-01 -4.89884764e-01 4.84398425e-01 -9.10207570e-01
5.15946746e-01 3.40889841e-02 -2.72630990e-01 -1.63419998e+00
1.27498964e-02 -2.52740085e-01 4.86446172e-01 1.38420653e+00
9.04479027e-01 -9.04818356e-01 2.17011690e-01 7.53981471e-02
-3.58096063e-01 -5.89568913e-01 -9.45712924e-01 -6.52726352e-01
8.03951502e-01 -4.16387320e-01 2.23179534e-01 1.36549973e+00
1.65810093e-01 9.46818233e-01 -3.05211037e-01 -8.84740129e-02
3.52138817e-01 6.13020174e-02 7.16317475e-01 -1.15762877e+00
-4.59993333e-01 -6.01023734e-01 1.36813581e-01 -8.96211565e-01
6.85737371e-01 -1.79316747e+00 1.60043210e-01 -1.01713598e+00
1.20419204e-01 -6.02413595e-01 -1.66101247e-01 7.19151735e-01
-1.24161243e-01 1.84881479e-01 -1.74552590e-01 4.42250371e-01
-2.87410110e-01 5.74557841e-01 9.17112291e-01 -2.40090668e-01
-1.47456527e-01 -6.06290512e-02 -4.02157128e-01 6.12809837e-01
7.27141738e-01 -6.79732740e-01 -9.14592221e-02 -9.34632957e-01
2.23277330e-01 -1.18357621e-01 -2.50664264e-01 -7.89916217e-01
3.05319309e-01 1.45837158e-01 3.30951065e-01 -3.64197075e-01
-2.13947613e-02 -6.78511798e-01 -2.58465111e-01 3.04573148e-01
-6.68101788e-01 3.00216019e-01 3.60977530e-01 5.50627671e-02
-2.96776503e-01 -2.67393500e-01 9.36460435e-01 -5.94273098e-02
-1.35082737e-01 2.16107555e-02 -3.91972542e-01 4.44628261e-02
7.74854779e-01 2.10110754e-01 -3.82605255e-01 1.42328173e-01
-6.72063470e-01 2.81607136e-02 4.31096315e-01 8.27344596e-01
-1.96410522e-01 -1.18197620e+00 -9.79740381e-01 4.22579408e-01
4.75061357e-01 -3.99289727e-01 -2.92470574e-01 9.00467455e-01
-3.53332639e-01 8.23444724e-01 -1.11767396e-01 -7.03550935e-01
-9.78537738e-01 4.04961646e-01 7.19287753e-01 -7.06349730e-01
-3.47636074e-01 7.93718457e-01 1.31857276e-01 -1.18645489e+00
1.58517197e-01 -2.68354803e-01 1.10077262e-01 1.53806105e-01
1.86504543e-01 -8.40165168e-02 5.89236498e-01 -9.69982624e-01
-2.02874899e-01 5.60890257e-01 -3.80574316e-01 -1.93927228e-01
1.10903502e+00 -2.83755958e-01 -2.07659364e-01 9.03378785e-01
1.31414783e+00 -7.41366204e-03 -6.56602085e-01 -6.59561813e-01
2.68000901e-01 1.27956763e-01 3.90208364e-02 -9.91636455e-01
-8.08059990e-01 1.24014187e+00 4.68780935e-01 6.69507636e-03
9.26634550e-01 -1.60996646e-01 6.30653918e-01 4.45488721e-01
3.54889512e-01 -1.31590760e+00 -3.35026026e-01 9.71974790e-01
7.46349573e-01 -1.46912348e+00 -3.82903248e-01 -2.49434425e-03
-6.45757914e-01 1.13855827e+00 5.76498091e-01 8.53196681e-02
7.15432703e-01 3.84747535e-01 4.82211828e-01 1.60160288e-01
-1.04756176e+00 6.44542575e-02 4.41506654e-01 6.73823655e-02
1.13868093e+00 2.16147736e-01 -1.60802037e-01 6.12348199e-01
-5.72515845e-01 -3.14779803e-02 -6.32403791e-02 8.17535460e-01
-1.97413012e-01 -1.42893183e+00 -2.51380175e-01 1.84161767e-01
-6.61796272e-01 -4.56075728e-01 -4.15345699e-01 6.78331494e-01
2.17450663e-01 8.14823985e-01 -2.67580189e-02 -3.90686899e-01
3.42879027e-01 7.96321511e-01 1.56872779e-01 -8.73822451e-01
-1.18769264e+00 1.13118999e-01 1.12171873e-01 -1.95855677e-01
-1.08098477e-01 -5.32139361e-01 -8.90356779e-01 -1.80442125e-01
-3.20565641e-01 9.87922326e-02 1.09806263e+00 1.23106766e+00
1.61045089e-01 4.91044849e-01 5.07232189e-01 -7.91035175e-01
-6.20831370e-01 -1.37654126e+00 -2.08903253e-01 2.46624872e-01
7.57544488e-02 -3.19471866e-01 -2.95545369e-01 3.63970876e-01]
|
[11.126908302307129, 9.967049598693848]
|
d5c9a864-e4f3-4d7b-9e72-0113f940fb2f
|
inpaintfusion-incremental-rgb-d-inpainting
| null | null |
https://mugichoko445.github.io/InpaintFusion/
|
https://arbook.icg.tugraz.at/schmalstieg/Schmalstieg_380.pdf
|
InpaintFusion: Incremental RGB-D Inpainting for 3D Scenes
|
State-of-the-art methods for diminished reality propagate pixel information from a keyframe to subsequent frames for real-time inpainting. However, these approaches produce artifacts, if the scene geometry is not sufficiently planar. In this paper, we present InpaintFusion, a new real-time method that extends inpainting to non-planar scenes by considering both color and depth information in the inpainting process. We use an RGB-D sensor for simultaneous localization and mapping, in order to both track the camera and obtain a surfel map in addition to RGB images. We use the RGB-D information in a cost function for both the color and the geometric appearance to derive a global optimization for simultaneous inpainting of color and depth. The inpainted depth is merged in a global map by depth fusion. For the final rendering, we project the map model into image space, where we can use it for effects such as relighting and stereo rendering of otherwise hidden structures. We demonstrate the capabilities of our method by comparing it to inpainting results with methods using planar geometric proxies.
|
['Denis Kalkofen', 'Dieter Schmalstieg', 'Hideo Saito', 'Wolfgang Broll', 'Okan Erat', 'Shohei Mori']
|
2020-10-01
| null | null | null | null |
['video-inpainting']
|
['computer-vision']
|
[ 5.63789427e-01 5.26006892e-02 5.91956794e-01 -3.16886127e-01
-7.93613255e-01 -7.00662017e-01 3.97476614e-01 -1.01445988e-01
-3.42667341e-01 7.88563669e-01 3.14449682e-03 1.82014331e-01
2.50536621e-01 -9.50637221e-01 -8.05146873e-01 -5.61060667e-01
3.91843766e-01 4.86702800e-01 5.37729800e-01 -6.57080859e-02
2.27503404e-01 9.99026299e-01 -1.60874176e+00 4.04123455e-01
7.19246984e-01 8.45957518e-01 2.65183479e-01 8.59913349e-01
-1.84488937e-01 7.88392305e-01 -3.82728785e-01 -2.59502202e-01
5.51795244e-01 -5.40594399e-01 -3.98807883e-01 4.49614078e-01
7.44071126e-01 -7.89302409e-01 -4.44224626e-01 8.59002531e-01
1.39190733e-01 2.90843379e-02 1.98284611e-01 -8.35757852e-01
1.16920844e-01 -2.33484492e-01 -8.46606195e-01 -5.62280297e-01
1.04860985e+00 -8.58197287e-02 3.82881910e-01 -8.56993973e-01
1.05559945e+00 1.24705648e+00 5.66123664e-01 1.24269135e-01
-1.43208277e+00 -3.50603491e-01 -6.12790100e-02 -8.52633491e-02
-1.31582868e+00 -3.83975297e-01 1.33915257e+00 -3.22417378e-01
6.08975947e-01 6.46973133e-01 1.04599679e+00 5.02663672e-01
3.65318716e-01 2.12637171e-01 1.30526829e+00 -6.22410893e-01
2.86182314e-01 -1.64279100e-02 -3.90684754e-01 6.82689369e-01
-1.45366296e-01 4.89554763e-01 -7.25835562e-01 -1.14501834e-01
1.27374375e+00 2.50188380e-01 -5.64509213e-01 -5.16977668e-01
-1.10464859e+00 4.50001299e-01 4.40370142e-01 6.71266839e-02
-4.83676434e-01 4.45333093e-01 -3.85588467e-01 3.22508484e-01
7.65103221e-01 3.96583349e-01 -1.14703275e-01 -3.87336314e-02
-1.18108261e+00 2.45609403e-01 4.96068180e-01 6.80656314e-01
1.30014551e+00 -1.26009807e-01 3.12511265e-01 3.67625266e-01
3.14143836e-01 6.75382912e-01 -8.31214637e-02 -1.84196162e+00
2.66381830e-01 3.72601509e-01 4.63307887e-01 -1.03549063e+00
-1.54749244e-01 5.88763654e-02 -4.53990281e-01 9.30845678e-01
3.58012348e-01 1.51827514e-01 -8.34951699e-01 1.23047495e+00
5.79103649e-01 2.19834775e-01 -1.31826952e-01 9.68563616e-01
4.57713485e-01 7.33106852e-01 -7.17489898e-01 -2.49016181e-01
1.02071857e+00 -6.86174333e-01 -7.17236876e-01 -2.80257255e-01
1.93562418e-01 -1.07109082e+00 7.90943146e-01 7.60391772e-01
-1.53565836e+00 -3.07603568e-01 -9.41426218e-01 -6.57866478e-01
1.81304459e-02 -2.19489440e-01 2.29298100e-01 3.69231582e-01
-1.26921952e+00 7.86193311e-01 -8.90761435e-01 -1.74581841e-01
-2.46013775e-01 2.01614797e-01 -5.62994063e-01 -4.21096325e-01
-7.81988204e-01 8.95596802e-01 -1.26048550e-01 -1.16818309e-01
-6.65572345e-01 -6.66040421e-01 -8.11202586e-01 -1.73998848e-01
2.18816623e-01 -7.35577524e-01 9.42284584e-01 -1.00897527e+00
-1.74354136e+00 9.80192721e-01 -2.79237390e-01 7.06876442e-02
8.65631521e-01 -2.27389440e-01 1.58161819e-01 5.07091582e-01
5.99707551e-02 5.59210002e-01 9.07985389e-01 -1.89099360e+00
-3.78945053e-01 -4.56643671e-01 1.78171858e-01 4.16126609e-01
2.89952844e-01 -3.84410679e-01 -8.05630565e-01 -5.27691066e-01
6.16764724e-01 -4.54742908e-01 -1.66852221e-01 7.34633923e-01
-3.48717302e-01 1.01361072e+00 1.02896845e+00 -1.02752066e+00
6.86546385e-01 -2.21582079e+00 5.19683778e-01 3.18175197e-01
1.69372484e-01 -3.13131064e-01 5.71782105e-02 4.75054860e-01
-3.59873031e-03 -4.40258592e-01 -5.50519228e-01 -1.24341726e+00
-5.84644079e-01 5.44499457e-01 -3.10966790e-01 7.31894791e-01
-1.80974364e-01 4.15637851e-01 -8.96214068e-01 -5.04602194e-01
7.54282713e-01 9.73524809e-01 -4.82854545e-01 4.01933938e-01
-2.87033111e-01 9.29850280e-01 -1.54700249e-01 6.75216854e-01
1.07042265e+00 4.25130695e-01 -1.12240925e-01 -2.47572094e-01
-5.42833447e-01 -2.38402896e-02 -1.20952392e+00 2.34035873e+00
-6.53863549e-01 6.94777846e-01 6.17691398e-01 -1.18096620e-01
7.24069297e-01 1.00397639e-01 6.06979966e-01 -8.13728988e-01
1.41982347e-01 2.34433532e-01 -7.14878201e-01 -1.01867221e-01
7.59932995e-01 -2.26109788e-01 2.76588440e-01 3.96544188e-01
-4.26194936e-01 -9.71984982e-01 -1.77400917e-01 1.32591233e-01
9.31735933e-01 6.33944213e-01 -1.61420867e-01 1.53954908e-01
3.41861039e-01 5.57652526e-02 1.36503682e-01 2.90427119e-01
5.96696377e-01 1.43929827e+00 4.15459424e-01 -3.92445207e-01
-1.28570187e+00 -1.07818437e+00 -1.42264534e-02 1.40051737e-01
3.46640110e-01 -2.51946241e-01 -8.97079647e-01 -6.73393384e-02
-1.51388738e-02 9.38815236e-01 -7.50895739e-01 1.34542033e-01
-6.04932249e-01 -1.54317960e-01 -3.73832881e-02 -3.80112194e-02
4.67176110e-01 -5.40472150e-01 -1.19548047e+00 2.80978113e-01
-2.62644231e-01 -1.09696710e+00 -3.14423770e-01 8.53599161e-02
-1.20987356e+00 -1.02939904e+00 -8.13940823e-01 -2.93773234e-01
7.30120122e-01 3.55677754e-01 1.06583345e+00 6.43591732e-02
-2.69689083e-01 4.78282481e-01 -3.20012689e-01 1.82071596e-01
-5.48216283e-01 -6.27331793e-01 -5.05569756e-01 2.67608315e-01
-6.20357394e-01 -9.70139623e-01 -7.25255132e-01 2.93672383e-01
-1.18147302e+00 6.84077144e-01 1.76400602e-01 4.00948226e-01
7.76867747e-01 -2.53082514e-01 -6.94621027e-01 -6.74876630e-01
-3.44327302e-04 1.17790289e-01 -1.02356136e+00 2.46617831e-02
-2.12380096e-01 -4.87771630e-02 2.25126967e-01 -1.90537110e-01
-1.29471087e+00 4.45440799e-01 -2.00439870e-01 -8.89841676e-01
1.68386236e-01 6.46239966e-02 -1.68672219e-01 -3.69756073e-01
4.55117255e-01 2.06877798e-01 7.29853958e-02 -6.37338758e-01
5.54650486e-01 2.14544654e-01 6.33685052e-01 -4.23173130e-01
8.07510316e-01 1.11552942e+00 3.42291206e-01 -8.60319257e-01
-4.40501899e-01 -2.64882326e-01 -7.80127764e-01 -4.16369796e-01
9.53430593e-01 -8.24619055e-01 -4.20974344e-01 4.91173834e-01
-1.70754135e+00 -5.57765186e-01 -6.14117324e-01 3.28188956e-01
-7.89634347e-01 5.59835136e-01 -6.50922060e-01 -8.08764219e-01
2.27849171e-01 -1.13202763e+00 1.54124320e+00 -7.33664557e-02
1.91928729e-01 -9.20887470e-01 3.46942335e-01 9.08505768e-02
1.07126802e-01 8.00266802e-01 4.77740079e-01 9.04142618e-01
-1.10425806e+00 -1.17167152e-01 -1.53384089e-01 1.81965262e-01
1.05914563e-01 4.82814573e-02 -1.31236815e+00 3.76179069e-02
3.36340606e-01 1.56520054e-01 6.36498332e-01 3.05779725e-01
7.35639215e-01 5.03492765e-02 -2.41563469e-01 1.19581819e+00
1.91407275e+00 3.10720444e-01 1.16859019e+00 3.77455711e-01
9.56804752e-01 7.25802660e-01 7.00676024e-01 4.69727933e-01
2.28283927e-01 1.00832021e+00 7.64817238e-01 -5.37939310e-01
-4.35185552e-01 -2.16981471e-01 1.72335088e-01 3.77668470e-01
-3.24099422e-01 -1.73578992e-01 -7.05818355e-01 2.50325918e-01
-1.70803165e+00 -6.84179127e-01 -5.27069032e-01 2.58753252e+00
7.29091525e-01 -3.15115571e-01 -4.92242843e-01 2.06410885e-01
2.79523104e-01 -1.25076532e-01 -2.35829160e-01 -2.77129650e-01
-2.81939864e-01 2.44815379e-01 8.41172218e-01 1.39777040e+00
-4.29976642e-01 6.73644245e-01 5.61954641e+00 4.00318414e-01
-1.20013034e+00 3.89675766e-01 4.77344692e-01 -1.09804317e-01
-7.34189332e-01 3.90156865e-01 -4.17336375e-02 1.53540507e-01
4.51351076e-01 4.02124912e-01 6.84807062e-01 3.59391302e-01
3.40239078e-01 -8.11442733e-01 -9.91674662e-01 1.29570329e+00
3.13782275e-01 -1.06726170e+00 -1.73596501e-01 1.31930798e-01
7.25062430e-01 -1.46227047e-01 -1.70693830e-01 -6.04686022e-01
4.33241986e-02 -5.90940833e-01 1.08759832e+00 8.28067780e-01
9.55227137e-01 -7.02325046e-01 3.51843864e-01 1.62018150e-01
-8.84815812e-01 3.88721615e-01 -1.43680051e-01 -9.16133448e-02
7.92413592e-01 8.79019022e-01 -3.55430305e-01 8.24251354e-01
4.41253424e-01 6.12883925e-01 -4.87337261e-01 9.53638911e-01
-3.25061589e-01 -2.76256502e-01 -6.16814494e-01 8.17593157e-01
9.51814186e-03 -6.88171983e-01 4.50564861e-01 6.02479935e-01
6.39097095e-01 3.62299502e-01 -1.06944166e-01 1.08167052e+00
3.05305243e-01 -2.22846746e-01 -5.79891860e-01 5.02549529e-01
-7.56529421e-02 1.03781188e+00 -8.47289622e-01 -2.36980438e-01
-2.41063029e-01 1.79567969e+00 -2.65240297e-02 5.50896585e-01
-7.67657936e-01 -3.35531265e-01 4.07435268e-01 6.47726893e-01
-7.19831362e-02 -4.40346181e-01 -4.03130561e-01 -1.29845214e+00
5.94742298e-02 -2.85483539e-01 -2.74234265e-01 -1.48670781e+00
-4.98777151e-01 6.45152926e-01 1.10165946e-01 -1.31379426e+00
-2.13506460e-01 -3.45572710e-01 -2.91356593e-01 1.05809081e+00
-1.77122772e+00 -1.20793581e+00 -6.20455265e-01 8.27904344e-01
2.27552071e-01 6.22223020e-01 7.58982897e-01 3.62660795e-01
1.29504070e-01 -2.44110033e-01 3.20889413e-01 -3.17446709e-01
7.39952981e-01 -1.11317253e+00 2.40371808e-01 8.93542051e-01
-1.08499110e-01 9.88341793e-02 9.22343552e-01 -6.26323760e-01
-1.61032474e+00 -6.72624171e-01 5.92031360e-01 -3.66680682e-01
1.38161704e-02 -3.99141937e-01 -7.01576591e-01 7.50933349e-01
7.09710717e-02 4.92754951e-02 -6.60795644e-02 -6.85172796e-01
-1.10286921e-01 -1.99950323e-01 -1.59357393e+00 3.69394332e-01
7.44355619e-01 -6.03504717e-01 -1.73562929e-01 1.69626743e-01
7.02083647e-01 -9.28628087e-01 -5.36186993e-01 -1.07665837e-01
7.13968098e-01 -1.69303191e+00 1.12467706e+00 5.00245392e-01
4.45002645e-01 -6.93708420e-01 -3.62171292e-01 -1.15861058e+00
2.03247815e-01 -7.99989700e-01 1.42235056e-01 9.11604404e-01
-4.40429486e-02 -5.98715663e-01 8.50419581e-01 5.88055730e-01
-2.60152936e-01 -2.08881795e-01 -1.16593611e+00 -3.40427995e-01
-3.12833518e-01 -5.92948556e-01 3.80342245e-01 7.68166423e-01
-3.79104286e-01 -1.69850975e-01 -4.44753796e-01 3.61982554e-01
9.98641491e-01 2.33053774e-01 8.41382086e-01 -1.07618141e+00
-5.24482846e-01 2.23863851e-02 -3.79360139e-01 -1.10717726e+00
-3.53810281e-01 -1.77194044e-01 1.41963765e-01 -1.64065742e+00
-2.24165589e-01 -4.41138715e-01 5.15200973e-01 1.53876290e-01
3.26569676e-01 5.17101526e-01 3.19817811e-01 2.66338557e-01
9.73773524e-02 4.53554541e-01 1.41034234e+00 2.00131893e-01
-3.47197533e-01 -5.01303971e-01 7.41992891e-02 7.55427003e-01
2.88392097e-01 -5.51455021e-01 -3.72571498e-01 -8.13179910e-01
3.68442148e-01 6.86070025e-01 4.94802803e-01 -1.11699951e+00
3.19434740e-02 -6.65275753e-02 5.99084675e-01 -7.01053560e-01
1.29448521e+00 -1.35915446e+00 8.21894825e-01 3.94036174e-01
1.46630660e-01 1.51807874e-01 3.65072712e-02 3.65253091e-01
-2.59901851e-01 -1.16001554e-01 8.42503786e-01 -3.37268889e-01
-3.13606173e-01 1.88559368e-01 2.11779792e-02 -4.21867698e-01
1.01521015e+00 -5.22853792e-01 1.31687358e-01 -6.91989362e-01
-8.44593763e-01 -2.84225285e-01 1.43024683e+00 -1.88827246e-01
1.10788429e+00 -1.16127145e+00 -4.85984147e-01 5.79739690e-01
-2.03785136e-01 3.92419606e-01 3.47714335e-01 7.55569875e-01
-1.57885480e+00 -1.45824790e-01 -9.97456014e-02 -8.29434931e-01
-1.28886759e+00 5.38732171e-01 5.46712935e-01 -6.00094646e-02
-7.69563794e-01 5.84519744e-01 3.47084820e-01 -1.80590793e-01
1.19716763e-01 -5.00903726e-01 4.41912353e-01 -2.12889001e-01
5.41876197e-01 3.37223709e-01 1.28643006e-01 -6.64495587e-01
-1.98126838e-01 1.26532114e+00 4.89115119e-01 -8.79035115e-01
1.27918434e+00 -4.90184814e-01 -4.49370325e-01 4.63762790e-01
1.27342045e+00 4.85258609e-01 -1.60957253e+00 1.07036635e-01
-6.53745949e-01 -1.11299002e+00 3.86229515e-01 -6.97156012e-01
-1.24161828e+00 9.74480033e-01 5.81892252e-01 -1.09766781e-01
1.34539926e+00 -1.86959222e-01 6.50817573e-01 -2.11513028e-01
7.57002115e-01 -7.25275457e-01 -2.09245622e-01 2.18807995e-01
9.99342561e-01 -7.43957579e-01 4.57572907e-01 -8.56619477e-01
-1.23198546e-01 1.47086132e+00 5.19053824e-02 -1.95058987e-01
5.65831244e-01 4.37824309e-01 2.04665765e-01 -1.57499030e-01
-1.61422893e-01 5.74172810e-02 -9.31783989e-02 5.98469913e-01
2.23172873e-01 -2.21690565e-01 -3.44620794e-02 -4.66137290e-01
-6.71795681e-02 -4.24514338e-02 8.83143723e-01 1.09218645e+00
-2.37584263e-01 -1.39994061e+00 -9.61947024e-01 -3.31178933e-01
-2.31887456e-02 4.20655161e-02 -4.29374427e-01 6.85034156e-01
3.88341635e-01 7.58703530e-01 2.39972100e-01 -1.40491456e-01
4.48036700e-01 -3.35810274e-01 9.41524029e-01 -5.30449450e-01
-3.55533183e-01 3.89984787e-01 8.43613520e-02 -1.13764715e+00
-4.85182971e-01 -5.87365389e-01 -1.14878035e+00 -2.14546129e-01
-9.34563726e-02 -2.67769784e-01 1.00153244e+00 4.12840724e-01
1.34127393e-01 3.68573546e-01 5.82643509e-01 -1.58714938e+00
3.87209237e-01 -4.71326768e-01 -8.14641833e-01 2.70710468e-01
7.75561273e-01 -4.94402587e-01 -5.26554286e-01 1.17654435e-01]
|
[9.40311336517334, -2.99514102935791]
|
ddacbbae-4e1c-4c7f-a71a-e05e8105bf3a
|
thermal-object-detection-using-domain
|
2006.00821
| null |
https://arxiv.org/abs/2006.00821v2
|
https://arxiv.org/pdf/2006.00821v2.pdf
|
Exploring Thermal Images for Object Detection in Underexposure Regions for Autonomous Driving
|
Underexposure regions are vital to construct a complete perception of the surroundings for safe autonomous driving. The availability of thermal cameras has provided an essential alternate to explore regions where other optical sensors lack in capturing interpretable signals. A thermal camera captures an image using the heat difference emitted by objects in the infrared spectrum, and object detection in thermal images becomes effective for autonomous driving in challenging conditions. Although object detection in the visible spectrum domain imaging has matured, thermal object detection lacks effectiveness. A significant challenge is scarcity of labeled data for the thermal domain which is desiderata for SOTA artificial intelligence techniques. This work proposes a domain adaptation framework which employs a style transfer technique for transfer learning from visible spectrum images to thermal images. The framework uses a generative adversarial network (GAN) to transfer the low-level features from the visible spectrum domain to the thermal domain through style consistency. The efficacy of the proposed method of object detection in thermal images is evident from the improved results when used styled images from publicly available thermal image datasets (FLIR ADAS and KAIST Multi-Spectral).
|
['Witold Pedrycz', 'Muhammd Aasim Rafique', 'Ahmad Muqeem Sheri', 'Shoaib Azam', 'Moongu Jeon', 'Farzeen Munir']
|
2020-06-01
| null | null | null | null |
['robust-object-detection']
|
['computer-vision']
|
[ 8.63799036e-01 -3.37801784e-01 1.64589629e-01 -4.97099519e-01
-4.73163724e-01 -8.52666199e-01 6.12334073e-01 -8.55162740e-01
-5.03444135e-01 6.78322911e-01 -4.12126392e-01 -1.26088083e-01
2.18358368e-01 -6.69155538e-01 -6.25750303e-01 -1.16445553e+00
5.28500736e-01 -9.36189666e-02 1.10304527e-01 -4.57001716e-01
1.72909960e-01 3.90844852e-01 -1.88522291e+00 3.39149415e-01
8.15601528e-01 1.21429837e+00 4.91627425e-01 7.53343940e-01
-3.50048803e-02 6.31857157e-01 -4.70258325e-01 2.57131737e-02
8.06640089e-01 -3.10781002e-01 -3.81629288e-01 2.02718988e-01
6.62539005e-01 -5.43316185e-01 -2.03189597e-01 1.15190184e+00
1.55994385e-01 2.85236984e-01 9.13244903e-01 -1.50042462e+00
-8.82672131e-01 -2.51726657e-01 -7.58890152e-01 3.31835151e-01
6.67090192e-02 4.44755226e-01 3.82354945e-01 -5.43409228e-01
2.55387902e-01 1.07708371e+00 4.60789651e-01 5.87464571e-01
-9.35452580e-01 -8.14260781e-01 -3.93097967e-01 3.67056847e-01
-9.17838752e-01 -2.33923733e-01 1.11675858e+00 -4.34433222e-01
7.22763360e-01 1.20681569e-01 4.89048958e-01 1.41239405e+00
6.84895337e-01 3.97441119e-01 1.96105421e+00 -4.24382448e-01
2.00193241e-01 6.29005730e-01 -1.12995960e-01 4.17381465e-01
5.46536967e-02 8.24513495e-01 -6.16195917e-01 2.17027202e-01
3.56551468e-01 1.13939278e-01 -2.87278816e-02 8.83536562e-02
-6.65450275e-01 6.37313366e-01 5.46687067e-01 -1.49529716e-02
-2.08684221e-01 8.02251548e-02 2.42524385e-01 5.25628746e-01
5.60265899e-01 1.81348592e-01 -5.78311197e-02 1.30162716e-01
-6.20941222e-01 -2.19960764e-01 3.03636324e-02 8.62024426e-01
1.05348980e+00 4.41091567e-01 3.31239760e-01 7.67975628e-01
3.50527853e-01 1.28536975e+00 3.97455454e-01 -9.15279388e-01
1.51881024e-01 4.74051982e-01 1.15606494e-01 -5.07226110e-01
1.50687657e-02 -1.30227283e-01 -5.65107465e-01 1.08429825e+00
1.46664530e-01 -2.73829490e-01 -1.20771468e+00 1.31849635e+00
3.48695278e-01 -9.91119593e-02 3.65243107e-01 1.12645876e+00
5.97662807e-01 7.68138707e-01 -1.57172725e-01 8.51762071e-02
1.37520361e+00 -7.74808824e-01 -6.71425164e-01 -8.56606960e-01
-5.31796739e-03 -8.75993848e-01 9.27863121e-01 4.77477610e-01
-3.27164888e-01 -1.03785563e+00 -1.40922260e+00 -8.37287009e-02
-6.87392950e-01 3.62581871e-02 5.81721783e-01 1.11379445e+00
-1.05782533e+00 -1.23949997e-01 -4.21309978e-01 -6.73944592e-01
4.30959344e-01 4.91545424e-02 -3.72054428e-01 -1.64834723e-01
-1.21524310e+00 1.06647086e+00 6.51027143e-01 1.65341526e-01
-1.02576327e+00 -4.71377969e-01 -6.70467257e-01 -6.38877332e-01
3.06701243e-01 -4.69562560e-01 8.91269803e-01 -1.56776166e+00
-1.57933664e+00 9.99684155e-01 5.68128414e-02 -5.67610860e-01
5.16942739e-01 -2.49591127e-01 -6.35257065e-01 4.73878175e-01
-1.23266459e-01 5.81309736e-01 1.65885854e+00 -1.42248166e+00
-7.41272330e-01 -5.12527525e-01 -1.53631661e-02 4.11180913e-01
-3.10773820e-01 -3.55177857e-02 2.51446843e-01 -1.59662619e-01
-2.37758055e-01 -1.13831830e+00 2.02983290e-01 1.23214439e-01
-1.16270900e-01 2.34603390e-01 1.75705409e+00 -6.21290922e-01
2.58698076e-01 -2.12594390e+00 -4.00370687e-01 -9.01682749e-02
-1.75819099e-01 2.86324859e-01 1.75460160e-01 4.14594799e-01
-1.47474825e-01 -3.36534858e-01 -3.90868723e-01 -8.89254287e-02
-4.00166422e-01 1.57121480e-01 -4.32538599e-01 5.89091480e-01
2.52556592e-01 7.50328124e-01 -5.80638826e-01 -2.94959277e-01
7.44016886e-01 4.61673707e-01 1.62771627e-01 3.55151325e-01
-1.84243396e-01 7.96762943e-01 -6.11014962e-01 5.10472357e-01
9.94958043e-01 4.87576962e-01 -4.98181731e-01 -3.80626142e-01
-3.13035071e-01 -4.97213006e-01 -4.98757541e-01 1.55192912e+00
-5.14723659e-01 8.82609844e-01 2.01567441e-01 -8.63191724e-01
1.23439240e+00 8.29269290e-02 4.03570294e-01 -1.05839741e+00
2.16964439e-01 5.04334867e-02 -3.38061631e-01 -7.93132544e-01
6.66197598e-01 -5.73127151e-01 -5.16914986e-02 5.42307615e-01
-2.96619087e-01 -7.49473929e-01 -4.96261865e-01 -1.31455824e-01
6.37065411e-01 5.32282114e-01 -1.20432332e-01 -8.22916180e-02
4.96916354e-01 3.24388623e-01 1.95212170e-01 4.31392044e-01
-2.84476429e-01 5.57191491e-01 -3.62001330e-01 -5.18219233e-01
-1.34742081e+00 -1.10508657e+00 -1.23774491e-01 1.00342119e+00
2.96778828e-01 7.40608335e-01 -6.57962978e-01 -3.02936882e-01
5.44085540e-02 7.17738271e-01 -7.56521642e-01 -5.21151602e-01
-2.35990092e-01 -5.56324720e-01 4.81324852e-01 5.62185705e-01
1.23902297e+00 -1.13492358e+00 -1.22346056e+00 -1.90758422e-01
-9.07381102e-02 -1.38625753e+00 -6.41622841e-02 4.15061772e-01
-6.30271852e-01 -7.94353485e-01 -4.05242115e-01 -4.93106395e-01
4.11813468e-01 6.86813354e-01 5.76135218e-01 -5.08268833e-01
-8.15609813e-01 5.08670807e-01 -3.82556349e-01 -9.20049310e-01
-6.77031636e-01 -4.51295853e-01 -8.16794336e-02 3.71708155e-01
4.83531654e-01 -1.58814639e-01 -7.06347227e-01 1.88873783e-01
-1.11297202e+00 8.12516809e-02 7.14323401e-01 6.73657417e-01
4.80051711e-02 3.28574598e-01 5.10119855e-01 -7.39608824e-01
1.64042741e-01 -2.24125847e-01 -6.49852753e-01 -1.61082968e-02
-5.02039075e-01 5.57404496e-02 7.13131011e-01 -1.12394705e-01
-1.93580258e+00 3.04588348e-01 3.80576640e-01 -4.15404648e-01
-5.32456875e-01 -7.25620892e-03 1.05262056e-01 -3.17711800e-01
9.46545303e-01 4.24524188e-01 2.60763645e-01 -8.85457620e-02
3.26324195e-01 1.05145538e+00 7.70422995e-01 -4.29836214e-01
1.26817644e+00 1.01066184e+00 8.87587592e-02 -1.27807367e+00
-6.26870275e-01 -7.43368804e-01 -6.12341881e-01 -7.38114715e-01
1.29209781e+00 -9.16507065e-01 -4.22321409e-01 7.31829941e-01
-7.31803954e-01 -1.99720159e-01 8.40698332e-02 2.62943715e-01
-5.56093991e-01 3.89617205e-01 -2.04902142e-01 -1.34804130e+00
-2.90653527e-01 -8.92985940e-01 1.30484819e+00 3.70308757e-01
3.61656845e-01 -9.41957116e-01 -1.38426349e-01 9.69164431e-01
4.56387579e-01 5.27936757e-01 6.56804681e-01 2.48647124e-01
-5.41116178e-01 -1.71159878e-01 -3.23986948e-01 8.65101337e-01
3.44880760e-01 -7.81819373e-02 -1.78250539e+00 -2.75202781e-01
4.61667120e-01 -5.68298042e-01 1.06892419e+00 3.42344999e-01
9.16086674e-01 3.10514599e-01 -1.93157464e-01 6.25075161e-01
1.77081895e+00 5.92724085e-01 6.67896330e-01 4.44523543e-01
7.10956216e-01 7.50670254e-01 8.83100688e-01 -2.97672600e-02
-2.01766282e-01 5.13705671e-01 6.53685629e-01 -5.67480147e-01
-8.74901190e-02 1.47355109e-01 6.72146559e-01 -1.94692776e-01
-1.64251551e-01 -1.56469405e-01 -7.47631311e-01 3.78666818e-01
-1.37903392e+00 -1.04404724e+00 -1.57050878e-01 2.19850397e+00
4.85455990e-01 3.54362354e-02 -3.40242200e-02 1.68751910e-01
6.58800304e-01 7.86830336e-02 -8.64208221e-01 -5.69749594e-01
-1.83998466e-01 2.18809530e-01 1.01018834e+00 2.43476957e-01
-1.04357481e+00 7.93835104e-01 5.99031496e+00 5.08742154e-01
-1.45893431e+00 8.78151059e-02 6.14463091e-01 2.52471954e-01
4.93184663e-02 -1.90659836e-01 -3.45872343e-01 4.02374774e-01
1.01970005e+00 -2.83671133e-02 4.29211617e-01 7.43352115e-01
3.82818013e-01 -5.34543157e-01 -8.96676064e-01 1.02104640e+00
2.67482817e-01 -4.86064285e-01 -2.95307428e-01 6.81702197e-02
8.06171238e-01 4.43967469e-02 8.52612197e-01 -1.53367147e-01
1.01351373e-01 -1.09382570e+00 7.04963803e-01 5.55748880e-01
1.06492627e+00 -8.40643048e-01 5.32728016e-01 3.48172098e-01
-1.19825065e+00 -3.01945955e-01 -5.97724259e-01 1.84226141e-03
-1.86807215e-01 2.31808677e-01 -1.22256255e+00 7.42749333e-01
8.95279706e-01 6.83449447e-01 -6.01006687e-01 4.22198623e-01
4.11222177e-03 6.00583553e-01 -2.65443712e-01 2.19698697e-01
4.22550708e-01 -6.44620717e-01 3.78903657e-01 9.48496103e-01
2.86328495e-01 -9.91324112e-02 6.63984865e-02 1.08933151e+00
3.74611259e-01 -4.43524897e-01 -1.27753401e+00 -2.87346337e-02
1.50658861e-01 1.51288295e+00 -5.50561011e-01 -1.98370546e-01
-3.85637969e-01 1.04958510e+00 -3.70710731e-01 7.29063392e-01
-8.68341684e-01 -1.24473363e-01 5.83658695e-01 6.01910008e-03
2.01123450e-02 -2.55106330e-01 -4.01530325e-01 -6.28143191e-01
-4.61259671e-02 -5.43197870e-01 1.84509158e-01 -1.33961344e+00
-1.23975301e+00 4.84191746e-01 1.68196872e-01 -1.56029594e+00
-2.10256781e-03 -9.22374666e-01 -4.83236820e-01 1.17780566e+00
-1.60607743e+00 -1.57747245e+00 -8.54971588e-01 6.28125310e-01
7.04222322e-01 -1.31922275e-01 5.03474951e-01 -1.06799752e-01
-1.30171731e-01 2.50937968e-01 2.34112576e-01 -2.34178811e-01
1.03208220e+00 -1.16739511e+00 1.13634296e-01 9.52757597e-01
-4.14754599e-01 2.78433323e-01 9.53596056e-01 -4.26123619e-01
-1.63914943e+00 -1.33296406e+00 -4.14323032e-01 -6.24450564e-01
2.11238474e-01 -3.32422704e-01 -6.72911108e-01 5.26341081e-01
7.10868716e-01 -2.41159707e-01 4.09169972e-01 -7.26453364e-01
-3.02787185e-01 -5.20779967e-01 -1.37269926e+00 1.47737458e-01
5.04274786e-01 -7.93675303e-01 -6.50488913e-01 1.55681923e-01
3.55284393e-01 -1.22810788e-01 -4.94982570e-01 1.22870140e-01
6.54937506e-01 -1.17435229e+00 9.18003500e-01 -3.43133770e-02
2.94611484e-01 -4.94295627e-01 -5.69835417e-02 -1.31091928e+00
9.29527432e-02 -3.68736506e-01 6.07260823e-01 9.86535549e-01
1.91945016e-01 -9.52185690e-01 8.36212814e-01 4.77049083e-01
-3.00618440e-01 1.82321310e-01 -7.32088387e-01 -7.15081930e-01
1.38158454e-02 -4.54422057e-01 1.23269551e-01 7.84826756e-01
-5.54588497e-01 2.85093844e-01 -5.82257688e-01 3.40601146e-01
1.08724606e+00 1.37219980e-01 7.14885950e-01 -1.08993089e+00
-2.55513549e-01 1.75871715e-01 -5.53803563e-01 -4.61363256e-01
1.02203064e-01 -6.30682409e-01 2.60388881e-01 -1.22791231e+00
2.45932922e-01 -7.15389624e-02 -1.02081388e-01 1.90338939e-01
-1.52901500e-01 5.62778413e-01 -2.79095080e-02 8.12576041e-02
1.53871244e-02 5.02361357e-01 1.32388628e+00 -3.80607933e-01
2.39803746e-01 -2.09488213e-01 -3.09268057e-01 4.48936641e-01
9.59137261e-01 -2.51683146e-01 -8.60849380e-01 -2.74110079e-01
5.19730896e-02 -1.33999437e-01 5.21734536e-01 -1.36124516e+00
1.47381470e-01 -3.43546808e-01 5.98173916e-01 -6.11486733e-01
7.66198099e-01 -1.29092109e+00 3.61616313e-01 4.18115586e-01
-1.22816183e-01 -2.62015462e-01 4.00044143e-01 8.25091660e-01
-7.24302828e-02 7.24061653e-02 1.14950895e+00 -2.32419029e-01
-1.36871338e+00 -2.09280904e-02 -5.12952566e-01 -3.39945525e-01
1.44234157e+00 -8.58811498e-01 -5.50027549e-01 -3.45194936e-01
-9.01401564e-02 -1.10323913e-01 7.36992180e-01 6.39123380e-01
7.16293275e-01 -1.07225871e+00 -5.25380194e-01 4.88632143e-01
5.53317249e-01 -8.83201584e-02 3.96838427e-01 4.58546340e-01
-6.34278536e-01 2.38023818e-01 -8.02221179e-01 -8.52503121e-01
-1.47334588e+00 4.78910714e-01 5.64190865e-01 6.67231917e-01
-6.03701890e-01 6.71964705e-01 5.79485536e-01 -1.60095915e-01
-3.56539398e-01 -3.58965635e-01 -7.56469145e-02 -4.61311698e-01
4.90153164e-01 2.64918923e-01 1.40396133e-01 -7.70757139e-01
-1.58147007e-01 7.03997076e-01 1.34471223e-01 -2.50857800e-01
1.11516392e+00 -4.10231829e-01 -2.18251385e-02 6.35883272e-01
1.26284027e+00 -5.35394788e-01 -1.54955399e+00 -1.20709307e-01
-5.57365477e-01 -5.76921046e-01 4.69588637e-01 -9.51565921e-01
-8.81828487e-01 9.99751210e-01 1.24879169e+00 1.39609590e-01
1.46414161e+00 -3.70636046e-01 7.06372917e-01 3.23712409e-01
5.32964945e-01 -1.42105782e+00 3.61826241e-01 2.33233348e-01
6.39110744e-01 -1.56590104e+00 -1.97800562e-01 -3.30783129e-01
-9.70777988e-01 1.20692563e+00 6.61467791e-01 4.62677469e-03
1.33002624e-01 3.35676551e-01 5.93887866e-01 -1.96132436e-01
-4.93953943e-01 -1.57734632e-01 2.20955387e-01 1.09079397e+00
1.21052772e-01 -1.48116169e-03 4.48220462e-01 -4.03576195e-01
-6.31343722e-02 -2.13424161e-01 6.85331702e-01 1.00885689e+00
-7.07334101e-01 -7.43742764e-01 -9.94176090e-01 3.33648950e-01
-2.80098200e-01 1.86637491e-01 -6.10716283e-01 5.50887644e-01
2.36101210e-01 1.39209211e+00 -1.67622074e-01 -5.11014581e-01
-1.08890198e-01 3.31977785e-01 5.04967272e-01 -4.53709573e-01
-1.56015426e-01 -1.51743412e-01 -1.10160545e-01 -3.79890233e-01
-6.89771116e-01 -4.40339386e-01 -8.82877767e-01 -6.96492493e-02
-2.21954599e-01 -1.80020675e-01 1.17047870e+00 8.39338541e-01
-1.39396891e-01 5.01828611e-01 8.59492660e-01 -8.20765913e-01
-4.21513110e-01 -9.90740895e-01 -8.01967144e-01 4.88063544e-01
5.62486470e-01 -8.23516965e-01 -5.11930346e-01 5.33826530e-01]
|
[9.215988159179688, -1.8126875162124634]
|
053ba867-10db-4bf3-88dd-a0d5120b0591
|
fast-effective-and-self-supervised
|
2104.08027
| null |
https://arxiv.org/abs/2104.08027v2
|
https://arxiv.org/pdf/2104.08027v2.pdf
|
Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders
|
Pretrained Masked Language Models (MLMs) have revolutionised NLP in recent years. However, previous work has indicated that off-the-shelf MLMs are not effective as universal lexical or sentence encoders without further task-specific fine-tuning on NLI, sentence similarity, or paraphrasing tasks using annotated task data. In this work, we demonstrate that it is possible to turn MLMs into effective universal lexical and sentence encoders even without any additional data and without any supervision. We propose an extremely simple, fast and effective contrastive learning technique, termed Mirror-BERT, which converts MLMs (e.g., BERT and RoBERTa) into such encoders in 20-30 seconds without any additional external knowledge. Mirror-BERT relies on fully identical or slightly modified string pairs as positive (i.e., synonymous) fine-tuning examples, and aims to maximise their similarity during identity fine-tuning. We report huge gains over off-the-shelf MLMs with Mirror-BERT in both lexical-level and sentence-level tasks, across different domains and different languages. Notably, in the standard sentence semantic similarity (STS) tasks, our self-supervised Mirror-BERT model even matches the performance of the task-tuned Sentence-BERT models from prior work. Finally, we delve deeper into the inner workings of MLMs, and suggest some evidence on why this simple approach can yield effective universal lexical and sentence encoders.
|
['Nigel Collier', 'Anna Korhonen', 'Ivan Vulić', 'Fangyu Liu']
|
2021-04-16
| null |
https://aclanthology.org/2021.emnlp-main.109
|
https://aclanthology.org/2021.emnlp-main.109.pdf
|
emnlp-2021-11
|
['cross-lingual-semantic-textual-similarity']
|
['natural-language-processing']
|
[ 6.28793836e-01 2.15842545e-01 -1.57123998e-01 -6.80140138e-01
-1.02626920e+00 -7.02947795e-01 7.18595028e-01 4.53857213e-01
-7.10927606e-01 8.21283460e-01 4.14630532e-01 -2.97799140e-01
3.38676780e-01 -6.45791650e-01 -8.48696411e-01 -2.03199938e-01
2.43863776e-01 5.00119686e-01 2.83929348e-01 -8.14660668e-01
3.17732722e-01 2.16023698e-02 -1.43838704e+00 7.82633543e-01
9.20523882e-01 6.07731998e-01 5.88256538e-01 6.51817024e-01
-4.15433586e-01 6.08905196e-01 -7.39789784e-01 -8.77965569e-01
1.29894987e-01 -6.30020320e-01 -1.02226436e+00 -4.22812074e-01
6.83347046e-01 1.26247913e-01 -8.76710340e-02 9.20006335e-01
8.19740355e-01 5.59544191e-02 3.85673404e-01 -6.75223947e-01
-1.08134890e+00 1.08243203e+00 -1.53162435e-01 5.43697238e-01
4.22606111e-01 3.07073295e-01 1.29919815e+00 -1.07559597e+00
6.16869926e-01 1.36795115e+00 9.68504131e-01 6.24572933e-01
-1.64331627e+00 -6.31794393e-01 -3.66619825e-02 2.30582908e-01
-1.37039161e+00 -8.67132664e-01 2.70921171e-01 -2.06634849e-01
1.63573611e+00 3.41970652e-01 9.77695957e-02 1.19785869e+00
1.13260977e-01 6.02859437e-01 1.06294310e+00 -7.16959655e-01
5.55592626e-02 4.63029236e-01 -1.11028351e-01 4.85621154e-01
2.33344976e-02 6.32699877e-02 -8.57791901e-01 -6.17442131e-02
5.24255514e-01 -3.20731550e-01 -1.69671372e-01 -1.96742564e-02
-1.34097099e+00 9.73710120e-01 4.29768533e-01 5.53087711e-01
-1.07361868e-01 -1.61459282e-01 7.55393028e-01 7.73632765e-01
5.17392814e-01 1.07571888e+00 -6.72109246e-01 -6.80664778e-02
-1.13072681e+00 -4.26988602e-02 6.78578138e-01 9.55915093e-01
9.32989776e-01 -7.71273952e-03 -4.38257933e-01 1.02788210e+00
-3.28086585e-01 3.06378841e-01 9.93313551e-01 -7.32674062e-01
6.15808249e-01 3.93691093e-01 -1.93416670e-01 -6.50281131e-01
-2.82204002e-01 -5.60399830e-01 -6.78780317e-01 -4.34121221e-01
3.77546041e-03 6.07539900e-02 -6.24296427e-01 2.25111723e+00
-2.11803585e-01 4.27120999e-02 2.17947885e-01 6.39625728e-01
9.19311345e-01 4.31861579e-01 1.57587677e-01 -2.15241179e-01
1.35838962e+00 -8.53813767e-01 -4.80083257e-01 -6.75841272e-01
1.00599062e+00 -8.80541444e-01 1.97797203e+00 -4.85805832e-02
-1.29831290e+00 -8.33690345e-01 -1.08887184e+00 -2.78177023e-01
-3.84243280e-01 -4.80640903e-02 6.06648803e-01 6.22093022e-01
-1.12524617e+00 8.38074863e-01 -3.69992793e-01 -7.14246213e-01
2.17091590e-01 2.35243946e-01 -3.55428457e-01 6.68384060e-02
-1.73457456e+00 1.33484459e+00 5.58866322e-01 -3.65147144e-01
-5.60170949e-01 -9.41127956e-01 -9.32351172e-01 -4.99968324e-03
2.55058438e-01 -9.53703403e-01 1.45320916e+00 -1.08813190e+00
-1.42489934e+00 1.31821299e+00 -4.66028064e-01 -7.73293734e-01
6.77831396e-02 -2.03154758e-02 -4.27219868e-01 1.56088201e-02
4.03951764e-01 1.00900602e+00 7.15781033e-01 -8.23177159e-01
-2.02346697e-01 -1.72878150e-02 1.12612672e-01 3.97665203e-01
-5.56843698e-01 3.72397006e-01 1.71946391e-01 -8.35047424e-01
-1.98641896e-01 -5.96729696e-01 -1.88943557e-02 -3.03029358e-01
-2.90765285e-01 -2.53879458e-01 7.48404190e-02 -6.18111253e-01
1.34579086e+00 -2.07277870e+00 1.05928913e-01 -3.32896829e-01
-1.69707283e-01 4.71140891e-01 -4.42535907e-01 7.48578250e-01
-1.38381183e-01 2.16038570e-01 -6.29838169e-01 -5.38424015e-01
2.91825496e-02 3.33675534e-01 -5.10473073e-01 -2.11519916e-02
4.34958935e-01 1.28486300e+00 -1.01289570e+00 -4.58359480e-01
-2.99558509e-02 1.17513552e-01 -5.01669884e-01 1.61887661e-01
-3.43805760e-01 3.07937831e-01 4.22002226e-01 2.69343674e-01
2.16098353e-01 -3.36121351e-01 1.29033327e-01 -2.16580465e-01
9.33264419e-02 1.10583770e+00 -6.91258788e-01 2.11282539e+00
-9.29768145e-01 6.73739672e-01 -2.76578605e-01 -1.00840247e+00
9.21747983e-01 2.72390366e-01 -1.15960017e-01 -1.01514924e+00
-2.38906860e-01 5.43574035e-01 1.08415283e-01 -3.00900161e-01
5.56138813e-01 -8.06613445e-01 -2.11222783e-01 6.15485549e-01
2.72859573e-01 -3.53929460e-01 2.60322928e-01 3.33748430e-01
1.14233208e+00 -2.05555335e-01 5.63957512e-01 -3.46950650e-01
6.70624673e-01 5.62991984e-02 2.35480592e-01 9.13543105e-01
9.21006873e-02 6.15422130e-01 1.20434016e-02 -8.66354704e-02
-1.18641472e+00 -1.19353127e+00 -2.01689973e-01 1.51390100e+00
-1.60845235e-01 -6.29324853e-01 -7.48960853e-01 -4.19320375e-01
5.54616638e-02 1.07870114e+00 -3.76515001e-01 -5.99373996e-01
-6.38745785e-01 -6.29228354e-01 1.04304504e+00 5.31274974e-01
4.12728965e-01 -1.28035593e+00 -2.21352085e-01 3.35774332e-01
-4.19244856e-01 -1.19777310e+00 -7.67748892e-01 3.41312259e-01
-7.54683256e-01 -4.55585897e-01 -4.00856346e-01 -9.54931915e-01
3.75814140e-01 2.97393858e-01 1.55125487e+00 -1.74749106e-01
-2.20086902e-01 -7.47698843e-02 -2.59999275e-01 -1.61769435e-01
-8.08860064e-01 3.70978117e-01 2.47067735e-01 -2.43408620e-01
5.80681145e-01 -7.51855493e-01 -3.36132258e-01 1.12109393e-01
-8.17128718e-01 3.62083279e-02 8.95410299e-01 9.63049710e-01
4.13505107e-01 -3.78659695e-01 1.11346400e+00 -1.06456816e+00
9.63149309e-01 -3.37342739e-01 2.59683076e-02 3.91518980e-01
-6.32061005e-01 3.72890532e-01 8.35358739e-01 -4.77053970e-01
-8.53001714e-01 -2.50671774e-01 -3.79711896e-01 -1.35893794e-02
-4.35044169e-02 6.19151473e-01 4.31300513e-02 -9.46327019e-03
1.24799132e+00 5.18225133e-01 -2.72568054e-02 -6.11128509e-01
6.92380726e-01 9.10034001e-01 6.79535270e-01 -5.29315710e-01
5.92996359e-01 2.53649592e-01 -5.05376577e-01 -7.11914718e-01
-1.26600170e+00 -3.79496306e-01 -7.14025199e-01 5.07029355e-01
6.02044284e-01 -1.06023824e+00 -4.62614894e-01 3.26273814e-02
-1.31718755e+00 -4.31425661e-01 -6.12909138e-01 9.27440524e-02
-6.28547966e-01 3.96471918e-01 -7.56272316e-01 -3.79275411e-01
-6.20795608e-01 -7.57934928e-01 1.06911016e+00 -1.45221680e-01
-8.57204020e-01 -1.15705097e+00 -7.86499977e-02 3.43483955e-01
7.29383290e-01 -4.09256905e-01 1.07900894e+00 -8.87867570e-01
-1.10733949e-01 8.70434269e-02 -1.94640994e-01 5.95178962e-01
3.17078680e-01 -6.52172565e-01 -9.51797485e-01 -2.87080109e-01
1.66151047e-01 -5.35334527e-01 9.76626039e-01 -5.70673309e-02
8.43990088e-01 -5.78814447e-01 -1.61495782e-03 6.59696519e-01
1.26654196e+00 -3.14090699e-01 3.20241898e-01 3.53281140e-01
4.61310834e-01 5.69132924e-01 3.14079374e-01 1.40586779e-01
4.35988307e-01 7.75583267e-01 -1.57073513e-01 2.90928427e-02
-4.28043813e-01 -5.61981201e-01 6.83791101e-01 1.08867013e+00
5.81557095e-01 2.89710592e-02 -7.17745662e-01 6.17146671e-01
-1.60798526e+00 -9.99851763e-01 1.30973533e-01 2.31479383e+00
1.77811778e+00 3.65079224e-01 2.65364554e-02 -2.41471436e-02
6.59928679e-01 1.93306223e-01 -5.24702847e-01 -8.52211595e-01
-6.99163079e-01 6.59125507e-01 3.14423651e-01 6.69172943e-01
-8.99885833e-01 1.34069896e+00 6.44684076e+00 1.05412376e+00
-1.18954778e+00 5.17981350e-01 2.71730393e-01 -3.20021302e-01
-6.71234727e-01 1.40277147e-01 -8.79195213e-01 5.23672104e-01
1.24536479e+00 -4.66353744e-01 6.94265187e-01 4.88266408e-01
-4.63309102e-02 1.86600700e-01 -1.42120600e+00 9.14036334e-01
2.60231435e-01 -1.62983847e+00 2.28468329e-01 -5.53633273e-01
6.55566037e-01 2.97814339e-01 -4.98115905e-02 5.07242382e-01
3.01088959e-01 -1.11276972e+00 6.32934690e-01 6.93730265e-02
1.18529761e+00 -3.98268163e-01 6.91369057e-01 4.93609130e-01
-8.57902408e-01 2.92010363e-02 -5.57219684e-01 -3.20477039e-01
2.40912110e-01 5.36330044e-01 -1.04609227e+00 3.10732245e-01
3.85968298e-01 6.45502865e-01 -8.56446207e-01 5.07289588e-01
-3.69607419e-01 6.64288878e-01 -1.25758290e-01 -8.34788904e-02
2.39371687e-01 2.75326520e-01 4.25964266e-01 1.72181094e+00
7.00046942e-02 -2.44566992e-01 -5.82990050e-02 9.85749960e-01
-1.30147874e-01 2.84756422e-01 -5.02705812e-01 -1.84441209e-01
6.52909935e-01 9.57533956e-01 -2.51464993e-01 -5.77049375e-01
-3.57127368e-01 1.36601675e+00 5.83972931e-01 1.61910594e-01
-5.35668373e-01 -5.06099820e-01 7.44048178e-01 1.58793300e-01
2.05027014e-01 -9.32617784e-02 -6.91570342e-01 -1.26788938e+00
-1.89461708e-02 -1.05262589e+00 4.04399544e-01 -6.74331009e-01
-1.80005538e+00 6.83369815e-01 -2.70068020e-01 -7.94557095e-01
-3.74495685e-01 -3.81868899e-01 -5.56642294e-01 1.16614950e+00
-1.49789798e+00 -9.36954677e-01 1.42825931e-01 5.12995005e-01
7.21427023e-01 -2.77181923e-01 1.03263938e+00 2.12072611e-01
-3.12557369e-01 9.96283889e-01 -3.87324244e-02 -2.91623585e-02
1.13084674e+00 -1.27340996e+00 1.01198792e+00 7.40223110e-01
5.08754194e-01 1.00550807e+00 7.21665561e-01 -5.64886808e-01
-9.21484649e-01 -9.54337239e-01 1.67535317e+00 -6.17389083e-01
9.01682794e-01 -7.65151858e-01 -1.12285984e+00 5.56842208e-01
4.06716824e-01 -2.79536247e-01 7.93201327e-01 3.26812983e-01
-6.19754612e-01 -3.92853618e-02 -9.88364875e-01 7.55455852e-01
1.47428727e+00 -1.20559728e+00 -1.04542124e+00 7.36323476e-01
1.32786322e+00 -1.91479668e-01 -6.30863845e-01 4.05829817e-01
1.37704313e-01 -9.36950684e-01 1.12601733e+00 -8.87539864e-01
4.52414632e-01 4.69977856e-02 -3.04085553e-01 -1.52103484e+00
-4.20211703e-01 -7.34466255e-01 1.82883307e-01 1.20567632e+00
6.10426188e-01 -8.04730296e-01 4.66491193e-01 2.58637398e-01
-4.26821798e-01 -6.76092923e-01 -9.96825874e-01 -1.19914925e+00
3.56700748e-01 -3.84502202e-01 6.05878055e-01 1.11450136e+00
2.71504939e-01 9.57705021e-01 -2.65114792e-02 -1.61412761e-01
2.29171440e-01 1.08447127e-01 3.95915776e-01 -9.18264627e-01
-6.50748491e-01 -6.60797238e-01 -2.38604859e-01 -1.08972514e+00
4.70837057e-01 -1.49652779e+00 2.20908485e-02 -1.31427193e+00
2.55098641e-01 -3.60359043e-01 -2.84198135e-01 4.98132080e-01
-5.88300049e-01 3.33036721e-01 1.32084042e-01 2.30619386e-01
-4.62809384e-01 5.15266836e-01 9.10426557e-01 -2.88407672e-02
-1.30356416e-01 -2.27895066e-01 -1.13162363e+00 3.56529772e-01
9.13344085e-01 -4.77719903e-01 -4.69092697e-01 -7.18794823e-01
4.89402413e-01 -3.64916235e-01 3.26650888e-01 -6.91645026e-01
2.58990407e-01 1.08486768e-02 -1.15762554e-01 -9.55091044e-02
3.03109139e-01 -1.78343043e-01 -1.69062719e-01 4.10282850e-01
-8.60434949e-01 1.77948952e-01 3.46751809e-01 1.45160779e-01
-2.83098578e-01 -4.93185252e-01 6.90602779e-01 -5.24514496e-01
-7.22751319e-01 -1.41462713e-01 -1.28634460e-02 6.41135275e-01
5.31994820e-01 -1.79117501e-01 -4.88632292e-01 -3.37415755e-01
-5.62881291e-01 -9.81981978e-02 4.97961193e-01 5.56610286e-01
4.29839045e-01 -1.27989745e+00 -9.75488007e-01 3.31186891e-01
3.95960957e-01 -2.93240160e-01 4.82294485e-02 7.05510616e-01
-5.19980565e-02 7.11715817e-01 -2.20006257e-02 -4.37956154e-01
-1.08429933e+00 6.71000600e-01 1.91761717e-01 -6.84505329e-02
-3.56884211e-01 1.31985450e+00 1.59272686e-01 -8.63913238e-01
-2.82115452e-02 -2.52783328e-01 2.45065302e-01 6.46232292e-02
4.29780126e-01 -7.07491562e-02 3.58631223e-01 -4.55765277e-01
-5.16588807e-01 3.27985585e-01 -1.72993302e-01 -3.02075803e-01
1.14612377e+00 -2.89506823e-01 -3.03906024e-01 6.98050141e-01
1.30791306e+00 1.47168683e-02 -7.02308953e-01 -8.07791173e-01
3.03058058e-01 -2.92128325e-01 -2.99698114e-01 -9.28748548e-01
-3.37698132e-01 1.15755713e+00 -3.69103067e-02 6.71419352e-02
9.94185865e-01 1.46096691e-01 1.17212081e+00 5.67309320e-01
3.38380605e-01 -9.99294281e-01 2.79827833e-01 6.58131778e-01
9.56423700e-01 -1.32502246e+00 -3.42481703e-01 -3.28687102e-01
-9.07675564e-01 7.20767558e-01 5.30203938e-01 1.02345563e-01
9.23918188e-02 1.80263638e-01 -1.46614667e-02 1.24491267e-01
-1.08259165e+00 -2.89631993e-01 1.92484528e-01 4.97878522e-01
7.03581393e-01 2.43268926e-02 -3.35604727e-01 6.95544362e-01
-7.63275743e-01 -2.20441893e-01 2.22946495e-01 6.07296884e-01
-6.62782490e-01 -1.20823550e+00 -2.13866886e-02 4.65293735e-01
-2.20899045e-01 -9.66702819e-01 -4.92659807e-01 3.08463693e-01
1.11262009e-01 9.20283556e-01 6.73428550e-02 -4.20352966e-01
3.44990194e-01 3.99000674e-01 5.59257984e-01 -1.24062908e+00
-1.01189530e+00 -6.18015885e-01 3.31395000e-01 -3.96870643e-01
-4.75765875e-04 -5.60518444e-01 -1.24188733e+00 -3.13625962e-01
-8.76048356e-02 5.89912534e-02 4.72950071e-01 9.97887194e-01
6.76002085e-01 2.95483917e-01 4.79269028e-01 -5.06859779e-01
-1.06290889e+00 -1.14196515e+00 -2.62764730e-02 5.86577237e-01
2.79428959e-01 -2.73544699e-01 -3.41611207e-01 -2.30944250e-02]
|
[10.90769100189209, 8.744810104370117]
|
fadeb983-d1ad-4e93-b479-ba79944e72f0
|
end-to-end-dense-video-captioning-as-sequence-1
|
2204.08121
| null |
https://arxiv.org/abs/2204.08121v2
|
https://arxiv.org/pdf/2204.08121v2.pdf
|
End-to-end Dense Video Captioning as Sequence Generation
|
Dense video captioning aims to identify the events of interest in an input video, and generate descriptive captions for each event. Previous approaches usually follow a two-stage generative process, which first proposes a segment for each event, then renders a caption for each identified segment. Recent advances in large-scale sequence generation pretraining have seen great success in unifying task formulation for a great variety of tasks, but so far, more complex tasks such as dense video captioning are not able to fully utilize this powerful paradigm. In this work, we show how to model the two subtasks of dense video captioning jointly as one sequence generation task, and simultaneously predict the events and the corresponding descriptions. Experiments on YouCook2 and ViTT show encouraging results and indicate the feasibility of training complex tasks such as end-to-end dense video captioning integrated into large-scale pretrained models.
|
['Ashish V. Thapliyal', 'Radu Soricut', 'William Yang Wang', 'Bo Pang', 'Wanrong Zhu']
|
2022-04-18
| null |
https://aclanthology.org/2022.coling-1.498
|
https://aclanthology.org/2022.coling-1.498.pdf
|
coling-2022-10
|
['dense-video-captioning']
|
['computer-vision']
|
[ 5.32797217e-01 2.45082274e-01 -1.49616420e-01 -3.80518049e-01
-1.11576939e+00 -4.38973784e-01 8.98427010e-01 -2.58730769e-01
-1.00091904e-01 1.00218987e+00 7.43432820e-01 3.97122698e-03
7.63702869e-01 -3.22875232e-01 -1.02111292e+00 -4.79151785e-01
1.08206244e-02 9.12554502e-01 6.81997091e-02 -1.12045528e-02
-8.12462494e-02 -4.56283428e-02 -1.40905190e+00 6.91463947e-01
3.64123732e-01 8.05852652e-01 5.32157362e-01 1.04931319e+00
-1.01077735e-01 1.09620333e+00 -6.48815870e-01 -4.65956360e-01
-8.84992033e-02 -1.03598642e+00 -8.11619997e-01 4.31155205e-01
4.38493639e-01 -6.23846233e-01 -7.61688590e-01 5.76364636e-01
4.49937791e-01 3.31876844e-01 5.67450762e-01 -1.40144563e+00
-8.74389172e-01 7.86575735e-01 -3.94505739e-01 3.14637125e-01
6.34088635e-01 3.48316431e-01 1.00065362e+00 -7.25963414e-01
8.31908345e-01 1.07730353e+00 4.37032551e-01 1.01346254e+00
-1.10981691e+00 -6.05816364e-01 2.70308942e-01 3.14480454e-01
-1.08798885e+00 -5.19380450e-01 5.97638369e-01 -4.47911501e-01
1.01839113e+00 1.43451318e-02 6.19264603e-01 1.63227236e+00
-1.62758261e-01 1.16607237e+00 6.29816711e-01 6.12317352e-03
3.26121948e-03 -2.57495373e-01 -2.41349578e-01 4.92068738e-01
-5.76169305e-02 -7.73461387e-02 -4.69601035e-01 -8.22052360e-02
8.86319041e-01 -4.51754965e-02 -3.88388842e-01 -1.33959696e-01
-1.68242311e+00 9.43119824e-01 2.51902372e-01 9.39200521e-02
-7.73898482e-01 6.51239634e-01 5.90456724e-01 -1.15767837e-01
4.08433884e-01 4.36652899e-01 -9.67678055e-02 -4.19654727e-01
-1.25193691e+00 5.06611168e-01 6.41272902e-01 1.35513353e+00
6.86243832e-01 9.35511217e-02 -6.19186223e-01 5.28085887e-01
1.48371190e-01 4.45171922e-01 6.23538017e-01 -8.51350546e-01
7.06351399e-01 -8.18775371e-02 2.55983412e-01 -6.05092108e-01
8.36004391e-02 5.35546988e-02 -7.59432733e-01 -3.88504118e-01
9.95468199e-02 -3.85181427e-01 -1.52476847e+00 1.89337695e+00
-9.59994197e-02 8.26374531e-01 1.78493947e-01 1.03706956e+00
1.04394102e+00 1.38805985e+00 6.26711130e-01 -2.24757046e-01
1.28992581e+00 -1.47426939e+00 -6.98554993e-01 -5.56596875e-01
2.57188290e-01 -7.03592539e-01 6.75898433e-01 -1.15907565e-01
-1.45763898e+00 -7.55514562e-01 -5.77391982e-01 -1.41084418e-01
9.33531821e-02 9.23948549e-03 6.96481109e-01 9.66101605e-03
-1.22056675e+00 1.27468362e-01 -9.08520103e-01 -4.16436970e-01
4.79360670e-01 9.17692780e-02 -3.91965806e-01 -1.97412774e-01
-1.18354464e+00 6.29369378e-01 7.30333805e-01 1.14283979e-01
-1.62079132e+00 -6.39559686e-01 -1.23378718e+00 2.86371976e-01
2.19520777e-01 -1.09138846e+00 1.67855918e+00 -1.19666731e+00
-1.17552900e+00 8.15675318e-01 -4.91907418e-01 -7.24638879e-01
3.07865262e-01 -2.78249830e-01 -2.40586326e-01 4.20860648e-01
2.76406497e-01 1.39789546e+00 9.71256435e-01 -1.29928052e+00
-6.91667438e-01 2.35662475e-01 2.73458799e-03 3.49677175e-01
1.06753916e-01 3.22479278e-01 -7.79226661e-01 -7.35209703e-01
-6.05136156e-01 -9.98679757e-01 -5.10089219e-01 -4.67368782e-01
-4.42190021e-01 -1.34012610e-01 7.75658548e-01 -8.76389861e-01
9.35855806e-01 -1.93085992e+00 3.81920755e-01 -3.71191263e-01
1.96390212e-01 3.48406672e-01 -4.02194619e-01 5.65424800e-01
-2.29619473e-01 1.40046000e-01 -2.41488010e-01 -8.72944176e-01
-7.06297308e-02 1.71811774e-01 -6.40482068e-01 1.30928690e-02
5.15527964e-01 1.35036564e+00 -1.21935391e+00 -4.91670787e-01
1.26317769e-01 6.22274637e-01 -4.64380503e-01 8.66220593e-01
-7.72714198e-01 4.00043249e-01 -4.60627645e-01 2.25456789e-01
7.46060684e-02 -7.81322181e-01 8.61814320e-02 -2.73956661e-03
2.87873447e-01 9.25722122e-02 -5.81110954e-01 1.80434656e+00
-3.79138410e-01 9.92445886e-01 -1.41394526e-01 -9.17259216e-01
5.80757558e-01 8.50467563e-01 4.54329342e-01 -2.59758174e-01
-7.24662170e-02 -1.09617211e-01 -4.86949623e-01 -6.87255144e-01
7.65146911e-01 -3.83545160e-01 -3.08829963e-01 5.28637290e-01
3.75237286e-01 8.63350462e-03 3.79038453e-01 5.29580176e-01
1.00746429e+00 5.56049287e-01 2.55489707e-01 4.88844812e-01
1.03071205e-01 2.98489422e-01 2.47220337e-01 6.36599898e-01
-2.69014146e-02 1.37982535e+00 4.28221047e-01 -4.02172118e-01
-1.47982526e+00 -9.02721286e-01 5.57827055e-01 1.02190828e+00
6.22635186e-02 -3.87496114e-01 -7.41512120e-01 -6.78072333e-01
-3.59055758e-01 6.79943264e-01 -7.35744357e-01 9.41370148e-03
-8.11756492e-01 -5.70703387e-01 5.29693365e-01 7.87858188e-01
3.44983160e-01 -1.64015579e+00 -3.81272763e-01 4.38958764e-01
-7.96528816e-01 -1.48975801e+00 -7.87932515e-01 -1.91999391e-01
-5.87198019e-01 -7.37188101e-01 -1.32655263e+00 -1.27914059e+00
6.73388958e-01 2.22781822e-01 1.61510110e+00 4.89203036e-02
1.14579201e-02 3.76960963e-01 -4.68880355e-01 -5.98367155e-02
-7.67579675e-01 1.95857257e-01 -4.09822285e-01 6.50299266e-02
1.98384374e-01 -3.58643413e-01 -3.86621296e-01 7.82313719e-02
-9.62791443e-01 6.09678507e-01 6.59226120e-01 7.80258179e-01
7.25033402e-01 -6.18998230e-01 8.18179071e-01 -7.73651600e-01
4.71459717e-01 -9.73676801e-01 -1.23989969e-01 1.98767841e-01
-5.98523431e-02 -7.84696043e-02 6.14196062e-01 -4.25600678e-01
-1.15272880e+00 2.45182917e-01 -3.04790586e-01 -9.25702989e-01
-6.00202262e-01 5.28456211e-01 -4.41822298e-02 5.90665102e-01
3.05158317e-01 6.39705598e-01 -1.31500632e-01 -2.57430285e-01
5.70848167e-01 3.18964750e-01 9.85984445e-01 -3.41606557e-01
8.07279289e-01 1.81113288e-01 -4.28598672e-01 -6.17185593e-01
-8.84287179e-01 -5.04342496e-01 -2.08719641e-01 -1.83020130e-01
1.45084631e+00 -1.41433680e+00 -1.99536890e-01 3.34350497e-01
-1.48737693e+00 -4.64058697e-01 -3.67211789e-01 3.83642793e-01
-1.03354692e+00 3.15143943e-01 -6.67221069e-01 -3.89727682e-01
-3.25150430e-01 -1.12039828e+00 1.52839041e+00 4.02939677e-01
-3.30207735e-01 -1.00072289e+00 3.14838946e-01 4.85458404e-01
2.29417294e-01 3.84292066e-01 4.52707112e-01 -6.66596949e-01
-8.49957228e-01 -2.56464332e-01 -2.24226177e-01 1.58289775e-01
-1.31523296e-01 -7.70170614e-02 -7.28221297e-01 -1.90068722e-01
-4.05668855e-01 -6.74077749e-01 9.62593317e-01 4.21617240e-01
9.68875408e-01 -2.67741740e-01 -4.48202312e-01 6.89821780e-01
1.29319119e+00 2.34428316e-01 9.66036975e-01 5.72358258e-02
1.01605475e+00 2.06237376e-01 4.31681484e-01 2.43965030e-01
4.09425229e-01 6.80603623e-01 3.28474224e-01 -2.98748583e-01
-4.88161922e-01 -7.40706027e-01 4.50338513e-01 4.57308829e-01
-1.22095622e-01 -8.51393759e-01 -6.16521001e-01 8.52074146e-01
-2.01053739e+00 -1.50826669e+00 1.10007830e-01 1.66880536e+00
6.93418741e-01 -7.22935125e-02 1.61986798e-01 -5.84010899e-01
1.11139774e+00 4.86977905e-01 -3.91568005e-01 -2.38032356e-01
-6.57873377e-02 -2.38967258e-02 6.27438426e-02 2.58432746e-01
-1.05869567e+00 1.15310562e+00 6.82289219e+00 4.58471000e-01
-1.01381445e+00 5.30534722e-02 9.26729262e-01 -1.04117632e-01
-3.67806375e-01 -8.05123672e-02 -7.57261217e-01 7.17808366e-01
1.24946225e+00 -3.07525575e-01 2.04946041e-01 8.62727165e-01
4.17862356e-01 1.85852468e-01 -1.27196205e+00 1.11278629e+00
4.69777346e-01 -1.64288890e+00 5.19417405e-01 -1.64767548e-01
1.08412433e+00 1.36086926e-01 -2.50268698e-01 5.78726053e-01
2.13966027e-01 -1.15563238e+00 7.85319328e-01 3.41353148e-01
8.56962800e-01 -3.48160863e-01 6.45809710e-01 -2.05963701e-02
-1.10114539e+00 2.58891046e-01 -1.66205555e-01 -8.64523500e-02
1.10231483e+00 1.14131659e-01 -1.06190932e+00 2.46970057e-01
2.68288732e-01 8.90294731e-01 -2.70359933e-01 1.31117237e+00
-3.94491881e-01 8.49104047e-01 3.68559472e-02 1.81598157e-01
6.16528392e-01 1.58391058e-01 4.90624309e-01 1.49653375e+00
4.22080368e-01 1.68323502e-01 2.98589587e-01 7.25145996e-01
-4.96035576e-01 -2.12390080e-01 -4.97047752e-01 -5.40110350e-01
1.30755007e-01 1.17480934e+00 -7.45316565e-01 -8.39312971e-01
-5.17500818e-01 1.29104078e+00 1.00855574e-01 5.33343673e-01
-1.28953719e+00 3.30049880e-02 7.44047582e-01 6.12924546e-02
7.31706679e-01 -1.94822922e-01 1.61527619e-01 -1.46350741e+00
-1.41516536e-01 -9.16267812e-01 3.85521442e-01 -1.45859981e+00
-9.21380818e-01 9.16456163e-01 3.47229652e-02 -1.14177549e+00
-9.63757157e-01 -3.99930961e-02 -8.25043619e-01 8.45957458e-01
-1.33182323e+00 -1.38019586e+00 -5.58244050e-01 5.62711656e-01
1.25233889e+00 4.58456054e-02 8.18609893e-01 2.35030353e-01
-5.41524351e-01 2.12262765e-01 -7.96824619e-02 2.38834098e-01
6.10482991e-01 -1.14835119e+00 9.62457955e-01 1.08778167e+00
3.32423955e-01 8.84632394e-02 9.96820986e-01 -7.88669944e-01
-1.05392361e+00 -1.32226598e+00 1.05688572e+00 -5.08064628e-01
4.65316564e-01 -4.87173766e-01 -8.27664614e-01 1.10210609e+00
5.75135410e-01 -9.48489383e-02 4.57443684e-01 -1.93372712e-01
-2.22100988e-01 4.74615186e-01 -6.37212992e-01 5.05301416e-01
1.06970990e+00 -4.84992623e-01 -7.26329505e-01 5.91118991e-01
1.02151084e+00 -5.06756365e-01 -5.90242207e-01 -4.61631268e-02
2.03762218e-01 -5.98232329e-01 1.00577819e+00 -9.77051795e-01
1.14441061e+00 -1.37586042e-01 2.11194187e-01 -1.29402196e+00
-4.73029017e-01 -1.06693554e+00 -3.67470950e-01 1.30439031e+00
5.68239510e-01 1.91802174e-01 1.02844107e+00 4.54817206e-01
-6.28562689e-01 -4.65222508e-01 -3.82336736e-01 -4.05147761e-01
-3.37309659e-01 -1.48719236e-01 4.79366332e-01 6.67454362e-01
-1.67336553e-01 7.77350545e-01 -9.74038422e-01 1.16033871e-02
4.53235716e-01 3.19580168e-01 7.61455595e-01 -7.09652126e-01
-4.56476957e-01 -1.60474613e-01 -4.47749138e-01 -1.34061706e+00
3.43896925e-01 -6.32751942e-01 5.82891226e-01 -2.04672098e+00
6.05210900e-01 5.07000554e-03 1.14479974e-01 4.87174481e-01
-6.75751030e-01 6.92163348e-01 2.86301494e-01 4.05374914e-01
-1.06214559e+00 6.11201704e-01 1.25648904e+00 -9.80200395e-02
8.29582103e-03 -2.62526125e-01 -7.23629713e-01 2.72066683e-01
4.42544907e-01 -2.41099715e-01 -6.78944170e-01 -7.33999014e-01
-1.22008003e-01 5.37033916e-01 4.48710442e-01 -1.06622314e+00
-1.13006122e-01 -2.83361912e-01 4.90093261e-01 -4.60320026e-01
4.92313087e-01 -4.15647894e-01 4.54295754e-01 -1.85025638e-04
-5.89675307e-01 2.07222521e-01 2.34637141e-01 7.10704684e-01
-5.45024753e-01 -2.58442410e-03 5.84333956e-01 -3.72870058e-01
-1.26551020e+00 6.65545940e-01 -4.67986166e-01 2.81001329e-01
1.36032176e+00 -1.65212482e-01 -2.03600913e-01 -9.40816224e-01
-1.02902400e+00 3.29937786e-01 5.07044733e-01 7.52793849e-01
7.00370073e-01 -1.33728230e+00 -1.17026746e+00 -1.34805560e-01
4.38885428e-02 8.07825550e-02 4.50898558e-01 3.35052550e-01
-3.97284597e-01 5.12009323e-01 -3.12243462e-01 -5.34365892e-01
-1.14476585e+00 8.90873432e-01 1.88082643e-02 -4.08884346e-01
-8.01932037e-01 1.01073647e+00 6.79252625e-01 4.43962753e-01
1.05053335e-01 -1.32977990e-02 -1.58765525e-01 1.07841194e-02
8.80441546e-01 -1.96091443e-01 -3.49580199e-01 -9.52256382e-01
-2.52654422e-02 1.69086933e-01 -2.27313027e-01 -1.19033262e-01
1.40444088e+00 -2.96639234e-01 3.69078189e-01 -1.39797097e-02
1.26996148e+00 -6.02378786e-01 -1.67289746e+00 1.19644158e-01
-3.86849016e-01 -2.53120661e-01 -3.30109537e-01 -6.07253194e-01
-9.98290539e-01 7.32104957e-01 -1.08131710e-02 1.42777979e-01
1.11795211e+00 4.04874623e-01 1.33448088e+00 1.00564003e-01
1.67393431e-01 -5.13622582e-01 1.81399688e-01 4.23815131e-01
8.52264166e-01 -1.14822245e+00 -4.12200361e-01 -2.68916249e-01
-1.07511067e+00 8.38561416e-01 7.17727721e-01 -1.41391173e-01
-5.53062409e-02 6.84441105e-02 -6.85611740e-02 -1.09319843e-01
-9.85525548e-01 -2.40425333e-01 1.01255208e-01 7.42558479e-01
4.78994995e-01 -1.36452973e-01 1.18500389e-01 3.78192604e-01
-2.21559568e-03 3.14597070e-01 7.61900544e-01 7.45022953e-01
-3.43463391e-01 -8.70291173e-01 -2.41898075e-01 3.58985156e-01
-4.62184250e-01 -3.59784007e-01 -2.07621723e-01 5.32051027e-01
-1.41413346e-01 6.52986109e-01 4.14734870e-01 -2.64620125e-01
7.26256222e-02 1.68270528e-01 2.96815485e-01 -9.19354200e-01
-4.44281131e-01 3.14395949e-02 3.68425399e-01 -5.39613247e-01
-4.88119304e-01 -7.99448311e-01 -1.16442537e+00 -1.38157725e-01
1.86697155e-01 5.62596142e-01 4.33386445e-01 8.67271483e-01
3.89277637e-01 5.17158031e-01 3.33653927e-01 -1.33284521e+00
-1.76006585e-01 -8.87130022e-01 -1.00032806e-01 8.23178351e-01
3.98498714e-01 -2.86681831e-01 -1.82482272e-01 9.38895762e-01]
|
[10.477128028869629, 0.7270507216453552]
|
0ac58112-988f-4063-ae97-3afc4bd2a68a
|
optimizing-neural-network-hyperparameters
|
1609.08703
| null |
http://arxiv.org/abs/1609.08703v1
|
http://arxiv.org/pdf/1609.08703v1.pdf
|
Optimizing Neural Network Hyperparameters with Gaussian Processes for Dialog Act Classification
|
Systems based on artificial neural networks (ANNs) have achieved
state-of-the-art results in many natural language processing tasks. Although
ANNs do not require manually engineered features, ANNs have many
hyperparameters to be optimized. The choice of hyperparameters significantly
impacts models' performances. However, the ANN hyperparameters are typically
chosen by manual, grid, or random search, which either requires expert
experiences or is computationally expensive. Recent approaches based on
Bayesian optimization using Gaussian processes (GPs) is a more systematic way
to automatically pinpoint optimal or near-optimal machine learning
hyperparameters. Using a previously published ANN model yielding
state-of-the-art results for dialog act classification, we demonstrate that
optimizing hyperparameters using GP further improves the results, and reduces
the computational time by a factor of 4 compared to a random search. Therefore
it is a useful technique for tuning ANN models to yield the best performances
for natural language processing tasks.
|
['Franck Dernoncourt', 'Ji Young Lee']
|
2016-09-27
| null | null | null | null |
['dialog-act-classification']
|
['natural-language-processing']
|
[-4.77033854e-02 1.17078505e-01 3.55744734e-03 -6.68128908e-01
-6.58707619e-01 -4.32829022e-01 8.00940752e-01 4.36421782e-01
-8.25913131e-01 6.93623126e-01 -1.02795556e-01 -3.29002798e-01
-1.95531994e-01 -7.58664072e-01 -1.61986351e-01 -7.54123449e-01
2.84657598e-01 1.12223327e+00 3.10635805e-01 -1.41326517e-01
6.29974186e-01 6.01077020e-01 -1.48143399e+00 -1.03289755e-02
9.28136289e-01 8.69912148e-01 1.37745127e-01 7.91252315e-01
-5.56316376e-01 2.45614424e-01 -7.42341042e-01 -3.23948056e-01
1.20884046e-01 -8.21923539e-02 -6.20970786e-01 -2.63584733e-01
-1.33065626e-01 1.98201373e-01 2.64216542e-01 9.32642043e-01
3.99031609e-01 3.83556306e-01 9.94927168e-01 -9.32419002e-01
-3.59611630e-01 6.34673536e-01 -7.73708075e-02 -1.54774278e-01
2.11370274e-01 3.72169912e-01 9.19147074e-01 -6.04084611e-01
-1.51929902e-02 1.50614762e+00 7.08914459e-01 4.70967799e-01
-1.59309638e+00 -3.73145700e-01 -1.22504681e-01 -1.26357749e-01
-1.38885808e+00 -4.05727983e-01 4.48918998e-01 -5.10129094e-01
1.21979201e+00 -4.45615463e-02 4.89675939e-01 9.33307052e-01
2.80390620e-01 4.99229223e-01 1.03060305e+00 -8.70705247e-01
6.80944264e-01 4.37192231e-01 4.37948048e-01 6.70985401e-01
1.58423036e-01 -7.09082261e-02 -4.29871589e-01 -4.46820199e-01
5.46108723e-01 -3.64524990e-01 1.12028439e-02 -1.67776421e-01
-9.99155104e-01 1.26932025e+00 -3.03346552e-02 3.70076448e-01
-6.64988220e-01 9.35500190e-02 1.53394237e-01 -3.36580351e-02
2.61452317e-01 1.01162505e+00 -8.10035646e-01 -5.70839226e-01
-7.07886279e-01 3.34202856e-01 1.26584816e+00 6.10249162e-01
7.24727571e-01 -7.91202411e-02 -2.68354923e-01 1.26279378e+00
6.37556612e-01 4.05043781e-01 5.97832680e-01 -9.41416442e-01
3.21285605e-01 8.01319897e-01 3.57313246e-01 -7.94757903e-01
-5.90375006e-01 -7.07340166e-02 -7.15890050e-01 3.37355733e-01
8.21457446e-01 -4.33880568e-01 -9.19285178e-01 1.47875333e+00
1.95434064e-01 -5.77140808e-01 2.29105517e-01 4.75461930e-01
3.57275754e-01 8.03205848e-01 4.15172368e-01 -1.00835994e-01
1.49776590e+00 -7.65115738e-01 -7.15532005e-01 -4.76027727e-01
3.92750502e-01 -6.38253808e-01 1.38293099e+00 6.18747771e-01
-9.26251233e-01 -4.25806671e-01 -9.32964802e-01 3.61837894e-01
-5.36665499e-01 2.40206316e-01 6.39841557e-01 1.21854472e+00
-9.21091378e-01 7.25460470e-01 -1.02671230e+00 -2.47198090e-01
1.06173761e-01 7.09714293e-01 -3.71299125e-02 5.42765796e-01
-1.13852966e+00 1.15149784e+00 9.38278675e-01 8.98919851e-02
-2.83942401e-01 -5.13728499e-01 -7.75317311e-01 4.05233711e-01
4.57341105e-01 -5.30035257e-01 1.55951536e+00 -5.08995235e-01
-2.34249830e+00 3.52136195e-01 -2.15012804e-01 -7.10041821e-01
3.28954995e-01 -3.22552264e-01 5.66893630e-02 -5.23824133e-02
-5.80264807e-01 7.88490117e-01 8.47001970e-01 -9.36695695e-01
-3.83724064e-01 -6.25739917e-02 -2.11454302e-01 5.38197607e-02
-4.91994143e-01 1.74592838e-01 -4.39058602e-01 -3.64353240e-01
2.02404279e-02 -1.17307651e+00 -4.98103797e-01 -3.97969127e-01
-3.34025800e-01 -5.11983335e-01 3.32513392e-01 -4.15495604e-01
1.21453416e+00 -1.86847377e+00 -6.19326085e-02 3.53666395e-01
-1.87334299e-01 4.64736253e-01 2.33484909e-01 4.15466785e-01
3.17865878e-01 2.23275602e-01 -2.78428525e-01 -3.94033283e-01
2.96376318e-01 3.43374938e-01 1.20214028e-02 2.82204319e-02
2.77264178e-01 6.31144345e-01 -5.11216581e-01 -4.75251049e-01
4.75372761e-01 4.79440749e-01 -4.20926481e-01 3.38192642e-01
-6.11898243e-01 1.98828191e-01 -3.59138161e-01 2.72169024e-01
2.17057824e-01 -3.15411150e-01 9.32747349e-02 3.61227654e-02
8.02347064e-02 2.72587448e-01 -1.37084794e+00 1.20924842e+00
-5.42439699e-01 5.00247955e-01 -2.40325823e-01 -6.91479087e-01
1.23098087e+00 3.77797365e-01 1.55877545e-01 1.29007408e-03
3.53377730e-01 -5.07892072e-02 1.26040965e-01 -3.94005805e-01
5.34318864e-01 -1.00443944e-01 -4.66113426e-02 4.61598724e-01
1.35669768e-01 -3.88822436e-01 3.36456984e-01 -4.12047356e-01
8.63159359e-01 -5.16953915e-02 6.95621789e-01 -3.84098828e-01
7.02533543e-01 1.07839100e-01 4.47425038e-01 1.08105981e+00
-1.48825860e-02 4.19164538e-01 5.91176510e-01 -4.77887392e-01
-9.42398250e-01 -7.06263959e-01 -2.47909203e-02 1.27908814e+00
-4.45641667e-01 -2.93290317e-01 -1.06723988e+00 -4.03972715e-01
-2.14955479e-01 1.29839540e+00 -3.31214309e-01 4.45708521e-02
-4.05505598e-01 -1.18425298e+00 5.26967764e-01 4.74523395e-01
5.20066857e-01 -1.41106498e+00 -7.22713709e-01 4.87203032e-01
9.72245112e-02 -1.22847950e+00 3.11179794e-02 3.59991074e-01
-9.70620751e-01 -7.57340014e-01 -6.15487278e-01 -1.82422861e-01
6.38937712e-01 -4.92211968e-01 1.01177227e+00 -2.25022033e-01
9.01717618e-02 1.63597241e-01 -3.14481109e-01 -7.89807498e-01
-9.25804019e-01 4.32218194e-01 1.31576195e-01 -1.73218533e-01
6.63319886e-01 -2.21587628e-01 -1.99421972e-01 2.61910141e-01
-7.11033404e-01 -2.03511834e-01 7.60733008e-01 8.79770100e-01
2.98968673e-01 3.95541400e-01 4.26063657e-01 -9.41791177e-01
1.22192693e+00 2.37898845e-02 -1.08191228e+00 3.05347592e-01
-1.06073463e+00 7.78281868e-01 6.21793091e-01 -4.68773425e-01
-1.26814198e+00 3.75467539e-01 -1.19658902e-01 -4.39942069e-02
-7.25858569e-01 5.14542043e-01 5.65083660e-02 4.90439832e-02
9.21376705e-01 -1.67122558e-01 4.10855189e-02 -4.14983630e-01
1.89466611e-01 7.26952851e-01 9.12595466e-02 -5.87306440e-01
3.50266486e-01 -5.50203361e-02 5.40131666e-02 -1.00799608e+00
-6.18702829e-01 -3.83088142e-01 -8.84104013e-01 -3.12335659e-02
1.08324909e+00 -3.40493500e-01 -1.06653345e+00 3.59294415e-01
-1.14561498e+00 -3.88404429e-01 5.48215695e-02 5.41806281e-01
-4.34376180e-01 2.51991868e-01 -4.54464048e-01 -1.25860679e+00
-5.88922560e-01 -1.36062038e+00 8.98635447e-01 5.35951793e-01
-7.83912957e-01 -1.05214608e+00 -5.28252907e-02 3.60217750e-01
5.01525164e-01 -2.55542338e-01 1.15082848e+00 -1.23545134e+00
-3.78945053e-01 -3.59227866e-01 1.26018614e-01 3.25331807e-01
-7.25244656e-02 1.73037097e-01 -1.00390518e+00 1.17970593e-01
-3.07020359e-02 5.13974056e-02 4.37130719e-01 6.39204323e-01
1.00021732e+00 -3.17354202e-01 -2.81426311e-01 5.94290867e-02
1.16395807e+00 5.83005250e-01 4.92113292e-01 4.96976167e-01
1.68225259e-01 6.84547067e-01 5.49390674e-01 5.00673950e-01
1.91750042e-02 5.69053710e-01 9.89851207e-02 3.54472816e-01
6.24286413e-01 1.45589143e-01 2.41590947e-01 3.89263928e-01
-1.52767198e-02 -3.38820130e-01 -1.41709816e+00 2.89440989e-01
-1.96375871e+00 -6.25915289e-01 1.07446238e-01 2.33108234e+00
9.39387679e-01 5.26154518e-01 2.17050910e-01 2.24692971e-01
6.09926045e-01 -1.80539221e-01 -3.26233923e-01 -6.19865000e-01
2.89685726e-01 2.38226548e-01 4.70381469e-01 7.35519469e-01
-1.18643415e+00 1.10003793e+00 6.75038815e+00 8.31962645e-01
-8.39137971e-01 -2.57669210e-01 5.75239956e-01 8.22077468e-02
3.14812839e-01 -8.40005800e-02 -1.24766672e+00 3.59530658e-01
1.27583838e+00 1.14566684e-01 4.96255994e-01 1.04502296e+00
5.28624058e-01 -5.15504301e-01 -9.09642696e-01 8.92485857e-01
-2.14743763e-01 -1.18237495e+00 -9.15586501e-02 1.07719630e-01
4.42375511e-01 -1.59586191e-01 -2.41081357e-01 3.36529523e-01
7.42506742e-01 -1.19581676e+00 2.76126027e-01 5.76313615e-01
1.72955915e-02 -9.00276542e-01 1.04779184e+00 5.66856146e-01
-6.08128369e-01 -1.61223516e-01 -4.78974104e-01 1.62199348e-01
2.19366044e-01 6.59454584e-01 -1.37913167e+00 -4.35070321e-02
6.38669670e-01 -1.94042146e-01 -5.79771936e-01 9.79939044e-01
-3.51327717e-01 8.62974048e-01 -6.45352542e-01 -8.35486114e-01
3.11690897e-01 -3.51340979e-01 5.36904573e-01 1.36888325e+00
1.21436737e-01 2.77961344e-02 4.03674580e-02 9.29928064e-01
4.93743986e-01 1.93849146e-01 -2.19889760e-01 -2.78715104e-01
5.16497314e-01 1.15084982e+00 -9.84773517e-01 -3.00377309e-01
5.00325970e-02 4.50354129e-01 2.06216604e-01 3.20822090e-01
-5.36805034e-01 -6.23392880e-01 4.79492456e-01 -1.46173492e-01
3.37981403e-01 -5.13502717e-01 -5.61184406e-01 -6.22723877e-01
-3.23697180e-01 -9.13407087e-01 2.76372969e-01 -5.49976885e-01
-1.26978815e+00 7.76248157e-01 4.12998259e-01 -7.73923814e-01
-8.84455740e-01 -9.64106023e-01 -4.41238493e-01 8.77060473e-01
-8.06784749e-01 -5.98662615e-01 -1.70772329e-01 2.06880793e-02
5.90967357e-01 -5.55155516e-01 1.18052614e+00 -4.00590837e-01
-5.07271409e-01 2.20071211e-01 6.81435019e-02 -4.05617282e-02
5.35125315e-01 -1.46762371e+00 2.47940421e-01 4.74384487e-01
1.23669490e-01 6.33566082e-01 1.04260361e+00 -3.48803639e-01
-1.03137052e+00 -5.27514935e-01 9.15399790e-01 -3.37312132e-01
5.97315967e-01 -2.45532781e-01 -1.00848174e+00 3.21572572e-01
2.76689649e-01 -7.39398479e-01 7.23324239e-01 4.67709869e-01
2.03444641e-02 1.06160298e-01 -1.14973390e+00 9.09462094e-01
2.45159090e-01 -1.86445475e-01 -8.09033334e-01 2.89101481e-01
4.65013683e-01 -1.35766372e-01 -9.86547709e-01 3.05879384e-01
4.85266149e-01 -8.72094870e-01 8.17182839e-01 -4.25249100e-01
-2.77471989e-02 -2.19188467e-01 7.97580257e-02 -1.50830078e+00
-1.71611279e-01 -8.72582018e-01 2.14173235e-02 1.25722122e+00
9.05446351e-01 -8.71203065e-01 9.45313931e-01 1.42627859e+00
2.54641652e-01 -7.40034938e-01 -5.53506851e-01 -6.32303655e-01
-1.05480358e-01 -5.72149634e-01 5.92488289e-01 5.12568057e-01
-5.63926399e-02 6.68133795e-01 8.76823440e-02 1.17093183e-01
3.84454191e-01 -2.35718504e-01 8.24610174e-01 -1.61065900e+00
-4.35129911e-01 -8.39106321e-01 -1.89456895e-01 -8.84296298e-01
3.40277135e-01 -2.32279897e-01 4.60979611e-01 -1.24982595e+00
-2.77601540e-01 -4.99211222e-01 1.52821839e-01 6.54313624e-01
-2.74148703e-01 -3.23353767e-01 5.60828485e-02 -2.58812448e-03
-1.98077813e-01 4.46365505e-01 5.53875923e-01 1.10184945e-01
-7.53031731e-01 3.80358368e-01 -4.64872569e-01 9.97111738e-01
1.18045735e+00 -6.14977658e-01 -3.16203088e-01 -6.22995198e-02
1.18112050e-01 -5.54472208e-02 -5.51913343e-02 -9.35214400e-01
5.23689747e-01 -2.47307450e-01 3.97835582e-01 -5.39029300e-01
7.35077441e-01 -7.71407366e-01 1.31220715e-02 3.68571162e-01
-5.29307723e-01 1.63598776e-01 3.61531913e-01 5.84385335e-01
-2.01759607e-01 -1.02224648e+00 8.83617878e-01 -2.27833137e-01
-2.80214339e-01 -1.69568941e-01 -9.44522917e-01 -2.67941952e-01
7.74754107e-01 -9.23505500e-02 6.08116947e-02 -5.62101841e-01
-7.71094441e-01 -1.62066482e-02 5.71285114e-02 1.30244359e-01
1.96734875e-01 -5.83739698e-01 -3.54137599e-01 2.32595623e-01
-1.76255018e-01 4.44520712e-02 -3.05444777e-01 4.69808221e-01
-7.81873345e-01 8.07768345e-01 1.02522388e-01 -7.76815236e-01
-1.37655675e+00 2.27958888e-01 2.85365403e-01 -6.99149251e-01
-2.82380670e-01 6.94318175e-01 -4.49237972e-01 -6.60689890e-01
2.85074174e-01 -3.98790509e-01 -5.04481196e-01 -2.57421583e-02
2.86909401e-01 3.13164413e-01 6.53952658e-02 -2.14077234e-01
-5.32067195e-02 3.89739573e-01 -5.57920374e-02 -6.24431610e-01
1.20441663e+00 2.24360794e-01 5.93613014e-02 5.23003340e-01
5.47594845e-01 -2.46951878e-01 -1.07765007e+00 -1.95550665e-01
5.02964616e-01 -2.25933418e-01 2.16341257e-01 -7.51539946e-01
-4.85060811e-01 8.11366498e-01 3.88095260e-01 5.38647175e-01
8.15920889e-01 -3.14456761e-01 3.18894595e-01 1.18769968e+00
3.22752774e-01 -1.24405634e+00 -6.37975112e-02 9.20594394e-01
7.87902355e-01 -1.31214094e+00 2.17725411e-02 -2.37125739e-01
-9.59486067e-01 1.36849833e+00 4.97435719e-01 3.24588567e-02
9.02092159e-01 3.56094390e-01 1.59571290e-01 -9.48744714e-02
-7.31768072e-01 -4.12707403e-02 3.19703966e-01 4.89605218e-01
4.12958592e-01 3.80749926e-02 -1.56806335e-01 4.27006245e-01
-5.03200173e-01 -2.44438276e-01 1.52394518e-01 7.27346778e-01
-5.72294056e-01 -1.37019658e+00 -6.91868424e-01 4.56339359e-01
-4.74952221e-01 -1.65650591e-01 -3.29731971e-01 8.41635466e-01
-5.26371360e-01 1.19896555e+00 -6.20254017e-02 -8.11147466e-02
2.61658847e-01 6.67350709e-01 2.11147517e-01 -5.90968490e-01
-9.19996858e-01 -7.65784457e-02 2.58977979e-01 -6.74737468e-02
-2.18657359e-01 -6.33482575e-01 -1.26272774e+00 -1.00568481e-01
-6.10740602e-01 2.69360185e-01 1.04888523e+00 1.16125619e+00
3.79601568e-01 2.28180289e-01 -3.36300628e-03 -9.01846409e-01
-8.87427449e-01 -1.37309980e+00 -1.56930029e-01 -5.12421206e-02
-3.69038105e-01 -6.83254719e-01 -3.45400304e-01 1.08077228e-01]
|
[6.882308006286621, 3.9900219440460205]
|
666435cc-06ea-4231-9992-2dac6be7686e
|
minimally-supervised-structure-rich-text
|
2102.11479
| null |
https://arxiv.org/abs/2102.11479v1
|
https://arxiv.org/pdf/2102.11479v1.pdf
|
Minimally-Supervised Structure-Rich Text Categorization via Learning on Text-Rich Networks
|
Text categorization is an essential task in Web content analysis. Considering the ever-evolving Web data and new emerging categories, instead of the laborious supervised setting, in this paper, we focus on the minimally-supervised setting that aims to categorize documents effectively, with a couple of seed documents annotated per category. We recognize that texts collected from the Web are often structure-rich, i.e., accompanied by various metadata. One can easily organize the corpus into a text-rich network, joining raw text documents with document attributes, high-quality phrases, label surface names as nodes, and their associations as edges. Such a network provides a holistic view of the corpus' heterogeneous data sources and enables a joint optimization for network-based analysis and deep textual model training. We therefore propose a novel framework for minimally supervised categorization by learning from the text-rich network. Specifically, we jointly train two modules with different inductive biases -- a text analysis module for text understanding and a network learning module for class-discriminative, scalable network learning. Each module generates pseudo training labels from the unlabeled document set, and both modules mutually enhance each other by co-training using pooled pseudo labels. We test our model on two real-world datasets. On the challenging e-commerce product categorization dataset with 683 categories, our experiments show that given only three seed documents per category, our framework can achieve an accuracy of about 92%, significantly outperforming all compared methods; our accuracy is only less than 2% away from the supervised BERT model trained on about 50K labeled documents.
|
['Jiawei Han', 'Jingbo Shang', 'Luna Xin Dong', 'Chenwei Zhang', 'Xinyang Zhang']
|
2021-02-23
| null | null | null | null |
['product-categorization', 'text-categorization']
|
['miscellaneous', 'natural-language-processing']
|
[ 4.06647205e-01 2.74462581e-01 -6.29690528e-01 -6.51075244e-01
-7.58308470e-01 -8.83211076e-01 7.21839666e-01 5.13925433e-01
-3.42328399e-01 1.34414360e-01 2.83035189e-01 -4.04047430e-01
-1.08122632e-01 -8.30449402e-01 -5.36709309e-01 -4.85348046e-01
1.31437242e-01 8.30877602e-01 -1.50580853e-01 3.39410864e-02
9.65264812e-02 -2.59585947e-01 -1.43357348e+00 5.14385581e-01
1.03818536e+00 1.44076979e+00 -5.94684258e-02 2.92742848e-01
-7.06414819e-01 7.90318012e-01 -2.03656331e-01 -7.33755767e-01
1.61972553e-01 -1.18052281e-01 -1.02537525e+00 4.86543298e-01
3.47627103e-01 -1.62225869e-02 -1.86983943e-01 1.25275540e+00
1.19013682e-01 -6.78440034e-02 9.86169219e-01 -1.12292683e+00
-6.34249747e-01 1.33295202e+00 -6.90155685e-01 -1.42586678e-01
-9.86902788e-02 -2.57276744e-01 1.57098997e+00 -9.14645970e-01
5.84531903e-01 1.17005372e+00 6.29583836e-01 3.61855567e-01
-1.39976919e+00 -5.46095669e-01 4.46419030e-01 -1.39407679e-01
-1.19525671e+00 -2.15901196e-01 9.47756350e-01 -5.70424736e-01
4.02923137e-01 -3.95577922e-02 1.09068334e-01 1.30833328e+00
-1.55910626e-01 9.38579917e-01 7.66690075e-01 -4.46757555e-01
2.95431167e-01 5.56129217e-01 9.28836703e-01 5.42483509e-01
3.16447735e-01 -4.79232013e-01 -2.20691964e-01 -1.15584791e-01
4.03461307e-02 3.56395006e-01 1.16301805e-01 -6.13682151e-01
-1.17871499e+00 1.07916903e+00 4.77722079e-01 3.26669067e-01
-1.91069245e-01 -1.09686658e-01 7.79567122e-01 3.12545925e-01
8.83981049e-01 2.27053389e-01 -7.21424699e-01 2.81678647e-01
-6.61140680e-01 -8.60756114e-02 1.01362658e+00 1.24520826e+00
7.85004497e-01 -2.90595800e-01 -3.58712561e-02 1.28892303e+00
4.12160277e-01 4.56700027e-01 7.45920062e-01 -4.07889903e-01
8.54805350e-01 1.01521194e+00 -2.83833802e-01 -1.04558539e+00
-5.23504317e-01 -7.56624877e-01 -1.19479978e+00 -4.68861818e-01
4.08814222e-01 -9.18327644e-02 -9.52263117e-01 1.62889934e+00
2.81525671e-01 -5.90992510e-01 -2.74861790e-02 6.47245467e-01
7.91696429e-01 6.80270731e-01 2.91480005e-01 4.60936688e-02
1.56115627e+00 -1.27587116e+00 -5.58430433e-01 -4.48514014e-01
8.61392856e-01 -5.24673879e-01 1.28968704e+00 3.02965730e-01
-5.99516928e-01 -4.68917519e-01 -9.20026779e-01 -6.59196079e-02
-8.16392481e-01 2.65572876e-01 5.84807158e-01 4.19566691e-01
-5.81315875e-01 3.35600257e-01 -3.69247615e-01 -4.39235985e-01
5.94335973e-01 1.42149076e-01 -2.03318045e-01 -1.08410262e-01
-1.08425915e+00 3.25859100e-01 7.36605346e-01 -1.22438349e-01
-6.76517367e-01 -5.38361192e-01 -8.88692558e-01 3.35305691e-01
6.64689004e-01 -4.75640833e-01 1.21074700e+00 -1.28560460e+00
-1.15504301e+00 1.07897627e+00 -5.37231117e-02 -3.42454642e-01
3.42055678e-01 2.80553196e-02 -3.97913218e-01 9.90860313e-02
3.52836102e-01 5.09994388e-01 8.08347940e-01 -1.52417505e+00
-8.61845195e-01 -5.76446593e-01 -1.17334381e-01 7.70250987e-03
-1.02443528e+00 -1.71732262e-01 -5.78411877e-01 -8.44364047e-01
4.13895577e-01 -9.68310654e-01 -1.44299075e-01 -1.37207806e-01
-9.06551480e-01 -5.63639760e-01 6.89498007e-01 -4.03737217e-01
1.17152047e+00 -2.08178067e+00 -3.72152142e-02 4.48154449e-01
5.56202948e-01 -1.88488886e-01 -2.50245094e-01 1.62064373e-01
-1.31406873e-01 3.63768518e-01 -2.37255499e-01 -5.85564733e-01
3.96714866e-01 1.21713998e-02 -5.44033825e-01 7.79552311e-02
4.51670438e-02 1.05896616e+00 -9.43453550e-01 -5.67775130e-01
-2.02451020e-01 1.63951740e-01 -3.95870656e-01 -2.32530106e-02
-5.56511223e-01 6.89688139e-03 -6.36929393e-01 8.16564858e-01
5.98503709e-01 -7.07877278e-01 5.53231359e-01 -2.37708613e-01
3.34744781e-01 2.97845006e-01 -1.03385639e+00 1.51683664e+00
-5.97676754e-01 4.97301579e-01 1.31998867e-01 -1.36368573e+00
9.04374182e-01 1.56666394e-02 4.37514335e-01 -5.99640191e-01
4.22068298e-01 3.01312149e-01 -1.87001213e-01 -3.31839502e-01
3.89006019e-01 4.55457792e-02 -4.00759220e-01 9.44847405e-01
3.58276039e-01 3.33760619e-01 3.70546043e-01 5.31376064e-01
8.26246381e-01 -2.83804566e-01 -1.14446476e-01 -5.23535907e-01
4.43162173e-01 9.37983692e-02 2.68554926e-01 7.60703206e-01
5.03126532e-02 3.91340584e-01 7.64332891e-01 -5.31319141e-01
-1.15287519e+00 -7.94899642e-01 -3.07985067e-01 1.73493266e+00
1.65942311e-01 -6.57952130e-01 -6.88015461e-01 -1.32094729e+00
1.93678439e-01 5.21298885e-01 -7.85479188e-01 -2.02802822e-01
-2.99314350e-01 -8.81860793e-01 2.63448119e-01 5.37001252e-01
4.26275969e-01 -7.87038326e-01 5.00574350e-01 1.43404275e-01
-4.13840652e-01 -1.09611249e+00 -5.80596745e-01 6.81828022e-01
-7.69914627e-01 -9.14724648e-01 -4.45285589e-01 -1.11619270e+00
8.44824851e-01 4.72001821e-01 1.42915738e+00 -4.21221852e-02
6.22865707e-02 1.58916652e-01 -4.72025573e-01 -3.08318764e-01
-6.78693891e-01 7.27035999e-01 1.73099134e-02 3.01999778e-01
7.25099146e-01 -4.10371035e-01 -5.30142426e-01 3.42261940e-01
-1.08972633e+00 4.78481166e-02 5.49545586e-01 9.85528111e-01
4.20495570e-01 4.31789845e-01 7.37858951e-01 -1.36289573e+00
4.15291309e-01 -8.82398784e-01 -2.86263913e-01 1.52125672e-01
-9.22622621e-01 1.09803163e-01 9.78470385e-01 -6.21805131e-01
-9.84525919e-01 3.21930908e-02 1.25504449e-01 6.32054731e-02
-2.25305483e-01 8.19706559e-01 -3.43286693e-01 4.24801469e-01
7.06433833e-01 2.41099343e-01 -1.21381313e-01 -6.80923343e-01
5.42545795e-01 1.16283190e+00 3.16751391e-01 -7.82058597e-01
9.77330208e-01 4.64866847e-01 -4.32008296e-01 -3.96180958e-01
-1.52841008e+00 -7.39062250e-01 -1.03351498e+00 -1.01365233e-02
5.72678745e-01 -9.18793619e-01 -4.86488700e-01 3.62196803e-01
-8.07678699e-01 -3.04837316e-01 -1.25494257e-01 1.73336118e-01
-1.21010654e-01 2.75129557e-01 -6.91990495e-01 -3.92252624e-01
-4.48059857e-01 -7.95727909e-01 9.87538815e-01 -1.39332995e-01
-8.79847482e-02 -1.32621014e+00 -1.77899987e-01 5.90170681e-01
2.84150004e-01 -1.51253209e-01 1.32342625e+00 -1.40508950e+00
-2.45158821e-01 -4.43866581e-01 -5.74495852e-01 5.23320019e-01
2.40810364e-01 -2.32279524e-01 -1.10582793e+00 -2.83752054e-01
-2.18635485e-01 -6.23478532e-01 1.21237004e+00 3.41345370e-02
1.52134311e+00 -5.11844635e-01 -5.59903681e-01 4.61488038e-01
1.23091364e+00 -3.47445719e-02 -4.02745493e-02 4.75058496e-01
1.06876600e+00 9.76860464e-01 3.93289775e-01 2.71403790e-01
5.22308409e-01 3.19155842e-01 3.52996916e-01 -1.56522796e-01
1.20868891e-01 -4.24363881e-01 1.29244983e-01 1.03857768e+00
5.70060313e-01 -4.82452095e-01 -1.01632035e+00 4.59731132e-01
-1.87759840e+00 -6.61606789e-01 -2.26026662e-02 1.88891411e+00
1.17618227e+00 4.54169571e-01 1.52423963e-01 8.81183147e-02
9.58835483e-01 2.08949223e-01 -7.17954993e-01 1.35361895e-01
-3.04980967e-02 -2.14020997e-01 3.63045245e-01 1.87541947e-01
-1.58888352e+00 7.99303412e-01 5.50974941e+00 8.56105804e-01
-8.80098581e-01 5.11328802e-02 9.94210422e-01 1.30980596e-01
-5.14630735e-01 -1.63120091e-01 -8.91246438e-01 7.74602354e-01
9.51963961e-01 -2.28465706e-01 1.08685993e-01 1.28011584e+00
-2.86990792e-01 4.03230578e-01 -1.25811696e+00 7.65784502e-01
2.21603021e-01 -1.30110455e+00 4.28955764e-01 2.08423361e-01
8.37671041e-01 5.82183748e-02 1.69496000e-01 5.94318032e-01
7.16811836e-01 -7.10097849e-01 7.54613638e-01 -1.35589033e-01
7.10221052e-01 -6.46513581e-01 7.63319969e-01 4.67251807e-01
-1.18345094e+00 -3.40669721e-01 -3.65896642e-01 3.75831634e-01
-2.16369480e-01 9.35136199e-01 -4.84457791e-01 3.90538037e-01
7.55868852e-01 1.05463195e+00 -6.32516026e-01 4.72281694e-01
1.08280070e-01 8.10439467e-01 -2.21847728e-01 -1.91455647e-01
4.61188227e-01 -2.51222879e-01 1.00838691e-01 1.40464783e+00
-8.49661306e-02 -1.98530391e-01 4.37904716e-01 7.93490052e-01
-6.80623412e-01 3.59966397e-01 -4.83058989e-01 -2.16773227e-01
4.36305970e-01 1.74187338e+00 -1.14135885e+00 -6.14662170e-01
-5.22456050e-01 7.28868961e-01 4.72576976e-01 2.92185932e-01
-5.24289072e-01 -6.16667092e-01 3.15163136e-01 -9.76557732e-02
9.82966349e-02 7.50831142e-02 -4.52459842e-01 -1.50448155e+00
1.82364509e-01 -9.09641325e-01 6.80236995e-01 -4.36063975e-01
-1.83605182e+00 7.28556156e-01 -3.12902421e-01 -1.28823447e+00
-1.02630883e-01 -9.30570781e-01 -2.91015685e-01 4.64381218e-01
-1.51596212e+00 -1.30121493e+00 -3.74859989e-01 3.06450158e-01
8.06250751e-01 -3.47101569e-01 5.67346334e-01 4.02743489e-01
-7.23939478e-01 6.87391460e-01 5.86027682e-01 6.92753196e-01
8.31930220e-01 -1.57531404e+00 4.18049246e-01 4.67611969e-01
2.22987264e-01 7.48207808e-01 1.98354945e-01 -5.02683699e-01
-1.19953156e+00 -1.50549543e+00 9.39379096e-01 -6.80151701e-01
1.09181798e+00 -8.60176086e-01 -1.04854620e+00 8.19351852e-01
8.25044960e-02 -1.40450284e-01 8.97186756e-01 4.95002240e-01
-8.73996615e-01 -2.89179116e-01 -8.42422843e-01 5.25636852e-01
1.08741224e+00 -5.93235195e-01 -5.77357471e-01 7.84863770e-01
9.40261424e-01 2.14861222e-02 -7.63987899e-01 8.82231593e-02
3.24874222e-01 -4.28236544e-01 8.37641001e-01 -8.45439076e-01
7.15931058e-01 6.05504960e-02 -1.26610622e-01 -1.30738378e+00
-3.63242477e-01 -2.47793153e-01 2.34647729e-02 1.77267408e+00
6.55011714e-01 -6.08628154e-01 8.18444908e-01 4.66273487e-01
-1.28081530e-01 -4.59542185e-01 -3.16060573e-01 -6.57871306e-01
2.67641068e-01 -4.29186583e-01 4.29336220e-01 1.29652250e+00
2.95027882e-01 9.14899468e-01 3.41197550e-02 -1.81992427e-01
9.15045500e-01 4.19102788e-01 6.16930485e-01 -1.68838179e+00
1.24472019e-04 -6.44964278e-01 9.99443829e-02 -1.27598822e+00
6.37360394e-01 -1.47086465e+00 1.63135231e-01 -1.38213325e+00
8.27214658e-01 -7.32899606e-01 -2.20004410e-01 6.17227256e-01
-1.53750867e-01 1.44625217e-01 -1.62632391e-01 4.43931609e-01
-8.60719323e-01 4.90477383e-01 9.07207966e-01 -7.33550906e-01
2.25281511e-02 3.13162990e-02 -1.15678632e+00 8.68891418e-01
6.98444784e-01 -5.82648456e-01 -4.09919411e-01 -3.83343250e-01
4.10435379e-01 -5.64984202e-01 1.45497382e-01 -4.48380917e-01
2.45386168e-01 -3.56410095e-03 3.34128946e-01 -6.71670437e-01
-2.44082361e-01 -1.00790322e+00 -5.01497030e-01 1.57711744e-01
-1.03109550e+00 -3.59798729e-01 -2.11099207e-01 7.30898440e-01
-2.17754617e-01 -4.16350543e-01 5.96714854e-01 -5.69501929e-02
-5.97737014e-01 4.68577147e-01 -2.27888748e-01 1.79076418e-01
5.78833759e-01 1.74419224e-01 -6.39185965e-01 -1.84225887e-01
-7.54472911e-01 4.16575849e-01 2.04803705e-01 6.80944204e-01
7.02050403e-02 -1.49590206e+00 -5.83487630e-01 2.30791137e-01
5.47956884e-01 1.17589273e-01 6.49460778e-02 3.51325750e-01
-5.87729132e-03 5.35594404e-01 3.66288543e-01 -7.57620215e-01
-8.99046659e-01 9.13133323e-01 1.13002494e-01 -3.86958748e-01
-5.29782176e-01 4.70123768e-01 4.36522961e-01 -9.05113935e-01
4.56596553e-01 -2.87334204e-01 -3.82328570e-01 3.95072579e-01
4.14189845e-01 -7.56271789e-03 1.29129261e-01 -4.94512290e-01
-5.62751591e-02 5.37250042e-01 -4.68430012e-01 1.08700298e-01
1.30543923e+00 -3.92345309e-01 -3.17408919e-01 5.72001517e-01
1.56118512e+00 -1.79805204e-01 -9.07992721e-01 -9.10385072e-01
4.09665227e-01 -6.40910193e-02 1.45745546e-01 -8.25567305e-01
-1.19195986e+00 8.52411330e-01 2.28380054e-01 7.71033049e-01
8.15410376e-01 3.97699654e-01 7.08144486e-01 7.39714205e-01
9.47768316e-02 -1.29520619e+00 5.41342378e-01 6.54860556e-01
5.37619233e-01 -1.58292353e+00 -2.53523469e-01 -6.24586284e-01
-5.84448099e-01 1.08910668e+00 6.57741368e-01 2.06908941e-01
9.09990132e-01 7.97014534e-02 2.01353580e-01 -2.77088046e-01
-8.61697078e-01 -2.19908357e-01 4.16927457e-01 4.25951868e-01
4.50568646e-01 -1.78650394e-01 7.67007545e-02 1.07840359e+00
-1.52709976e-01 -4.34021235e-01 1.70646474e-01 6.86214924e-01
-4.83736187e-01 -9.00752962e-01 -2.57138666e-02 8.55237603e-01
-4.63474989e-01 -1.46908402e-01 -4.11996663e-01 5.89261353e-01
-1.89022854e-01 9.03818369e-01 3.83432984e-01 -4.71988708e-01
1.76616639e-01 3.15146118e-01 -2.94072509e-01 -7.79663682e-01
-5.17766416e-01 1.26792029e-01 8.58030990e-02 -1.69119075e-01
-2.71193117e-01 -5.70953310e-01 -1.05997145e+00 -2.97658533e-01
-4.00092453e-01 3.40436578e-01 6.78923607e-01 9.05027926e-01
3.79709750e-01 4.45854157e-01 1.18050420e+00 -6.05356932e-01
-8.10591877e-01 -1.14626420e+00 -7.03483403e-01 7.43092895e-01
1.79921404e-01 -4.62117940e-01 -7.73829818e-01 5.07691145e-01]
|
[10.360474586486816, 6.674609184265137]
|
31e02a7b-05c3-45bc-b66b-7ea80d3d1b19
|
learning-illumination-from-diverse-portraits
|
2008.02396
| null |
https://arxiv.org/abs/2008.02396v1
|
https://arxiv.org/pdf/2008.02396v1.pdf
|
Learning Illumination from Diverse Portraits
|
We present a learning-based technique for estimating high dynamic range (HDR), omnidirectional illumination from a single low dynamic range (LDR) portrait image captured under arbitrary indoor or outdoor lighting conditions. We train our model using portrait photos paired with their ground truth environmental illumination. We generate a rich set of such photos by using a light stage to record the reflectance field and alpha matte of 70 diverse subjects in various expressions. We then relight the subjects using image-based relighting with a database of one million HDR lighting environments, compositing the relit subjects onto paired high-resolution background imagery recorded during the lighting acquisition. We train the lighting estimation model using rendering-based loss functions and add a multi-scale adversarial loss to estimate plausible high frequency lighting detail. We show that our technique outperforms the state-of-the-art technique for portrait-based lighting estimation, and we also show that our method reliably handles the inherent ambiguity between overall lighting strength and surface albedo, recovering a similar scale of illumination for subjects with diverse skin tones. We demonstrate that our method allows virtual objects and digital characters to be added to a portrait photograph with consistent illumination. Our lighting inference runs in real-time on a smartphone, enabling realistic rendering and compositing of virtual objects into live video for augmented reality applications.
|
['Christoph Rhemann', 'Rohit Pandey', 'Wan-Chun Ma', 'Sean Fanello', 'Paul Debevec', 'Jay Busch', 'Jason Dourgarian', 'Chloe LeGendre']
|
2020-08-05
| null | null | null | null |
['lighting-estimation']
|
['computer-vision']
|
[ 7.25343645e-01 -3.14988554e-01 3.75635058e-01 -4.30379629e-01
-1.03049052e+00 -7.58083940e-01 3.92449975e-01 -9.87088621e-01
2.69908849e-02 7.97961414e-01 1.51996225e-01 5.37766367e-02
4.94298756e-01 -6.31501675e-01 -1.05795979e+00 -7.37526417e-01
2.16719642e-01 -4.41893972e-02 -5.52290559e-01 -1.06499039e-01
-1.60735562e-01 6.69544339e-01 -1.87063551e+00 1.67778075e-01
6.21899009e-01 9.64652896e-01 -1.44858629e-01 1.31689847e+00
4.85341996e-01 7.58021057e-01 -7.66260207e-01 -7.40579128e-01
8.45147610e-01 -3.86312664e-01 -6.79200590e-02 5.53307712e-01
1.47592854e+00 -8.44001293e-01 -4.82653975e-01 7.38277376e-01
7.82983184e-01 7.42088482e-02 3.73768687e-01 -1.11905766e+00
-7.47197211e-01 -3.44628334e-01 -9.52777803e-01 -3.63114774e-01
1.09178984e+00 4.26028430e-01 6.04628384e-01 -8.37317228e-01
6.99471354e-01 1.14284277e+00 9.97707963e-01 5.17798543e-01
-1.59521818e+00 -9.01374042e-01 6.80161733e-03 -1.85559541e-01
-1.66010559e+00 -7.82098174e-01 9.75840449e-01 -4.97306734e-02
5.16740322e-01 6.70304775e-01 1.05769181e+00 1.48311782e+00
1.25455827e-01 2.79733360e-01 1.91452825e+00 -2.70585477e-01
1.41069945e-02 2.10624918e-01 -5.76482773e-01 7.24203825e-01
-1.91590548e-01 1.47409156e-01 -6.33970380e-01 -2.11782083e-01
1.12770891e+00 -1.89954370e-01 -6.33693278e-01 -3.00522950e-02
-1.11135936e+00 2.27285936e-01 4.89544123e-01 -3.41375679e-01
-1.50851160e-01 4.01070058e-01 -2.07543969e-01 2.69800782e-01
5.20250499e-01 1.42054647e-01 -1.76775843e-01 1.90031812e-01
-6.58322275e-01 3.78448367e-02 6.92741275e-01 9.54567850e-01
6.90095007e-01 2.80927360e-01 1.61463112e-01 8.15160990e-01
1.80561006e-01 1.16350341e+00 -1.31719068e-01 -1.33683288e+00
1.10190824e-01 -1.86702907e-01 3.03083092e-01 -8.65623236e-01
1.32687204e-02 -1.63299859e-01 -8.48770380e-01 6.02382541e-01
2.09454358e-01 -1.82974339e-01 -9.48060989e-01 1.64850307e+00
5.73460698e-01 5.42052329e-01 -4.22565565e-02 1.12541306e+00
6.54094636e-01 6.53193712e-01 -3.68601322e-01 -3.58532518e-01
1.15451992e+00 -3.65990520e-01 -7.97010124e-01 -8.57083872e-03
-2.96390921e-01 -9.52189028e-01 1.60589468e+00 8.69432747e-01
-1.17967927e+00 -5.64931631e-01 -1.01856983e+00 -4.62962657e-01
3.86976004e-02 -4.23416533e-02 6.94517851e-01 1.12891710e+00
-1.20176530e+00 1.88923493e-01 -3.31753135e-01 -1.83459491e-01
2.90103793e-01 1.39322296e-01 -2.71601975e-01 -3.83785963e-01
-8.70284677e-01 6.26495600e-01 -6.63243473e-01 2.10999563e-01
-1.12903678e+00 -9.83397305e-01 -9.03779209e-01 -5.54385424e-01
-3.53344902e-02 -8.15099716e-01 7.37637579e-01 -1.36551523e+00
-1.93503916e+00 1.13907278e+00 -1.41310498e-01 4.45802510e-02
6.59479201e-01 -1.49328172e-01 -6.07231498e-01 7.28501678e-02
-2.50902981e-01 4.50136513e-01 1.18831599e+00 -2.02273726e+00
-3.99265252e-02 -3.77389342e-01 4.04960327e-02 6.92417204e-01
-8.28728601e-02 -3.07474304e-02 -4.38343525e-01 -4.63707954e-01
-1.52300522e-01 -9.46033776e-01 1.77715167e-01 5.86970925e-01
-6.01089358e-01 8.97133231e-01 7.09522963e-01 -8.14708769e-01
5.30609131e-01 -2.00953889e+00 -1.64473698e-01 1.89819887e-01
4.41109911e-02 -3.37019414e-01 -1.94168210e-01 -1.31516561e-01
-5.16518876e-02 -1.80887699e-01 -1.31583750e-01 -7.26513565e-01
-9.90481526e-02 2.50164747e-01 -6.65226102e-01 9.14876521e-01
-3.33687365e-01 6.63448989e-01 -7.78146803e-01 -3.65471154e-01
5.98858833e-01 1.44615650e+00 -4.18114215e-01 4.15635198e-01
3.64409527e-03 7.90482461e-01 2.00637683e-01 9.76340711e-01
8.91476393e-01 2.79439241e-01 1.33358866e-01 -5.55390954e-01
1.35021240e-01 -2.01641291e-01 -1.13636541e+00 1.69615245e+00
-1.05878270e+00 9.20662284e-01 2.12995201e-01 1.78659528e-01
8.37776184e-01 -2.82354951e-02 3.32788795e-01 -8.19088459e-01
-1.78616326e-02 -3.59366648e-02 -7.91486740e-01 -2.92949617e-01
5.10656714e-01 -3.56340438e-01 -5.33056930e-02 2.28156105e-01
-4.22233105e-01 -8.73303592e-01 -7.01400518e-01 -1.91645354e-01
8.57522368e-01 5.03421426e-01 2.93574445e-02 2.83082306e-01
1.91533506e-01 -6.68151855e-01 2.06847921e-01 3.22716236e-01
1.63094074e-01 1.08612525e+00 -1.85208857e-01 -5.57302177e-01
-1.35158467e+00 -1.66433442e+00 -4.31411177e-01 7.73847520e-01
1.30086794e-01 7.70842060e-02 -7.32838631e-01 -8.24197233e-02
-1.47042304e-01 8.49266171e-01 -7.51148820e-01 3.36698830e-01
-4.58287358e-01 -7.83825397e-01 4.25768197e-01 -2.38364772e-03
8.20664942e-01 -4.19037580e-01 -3.43085825e-01 -5.05465448e-01
-4.35305268e-01 -1.21628225e+00 -7.14486361e-01 -3.22645247e-01
-2.24863216e-01 -7.23640800e-01 -7.69374132e-01 -4.45769370e-01
8.16774666e-01 4.66560632e-01 1.51618159e+00 -2.59862274e-01
-7.66393661e-01 1.00730562e+00 1.15915015e-01 -1.88728586e-01
-3.51656109e-01 -7.41361320e-01 2.33580321e-01 4.58189160e-01
-3.15954059e-01 -8.19076002e-01 -9.77262199e-01 6.15711391e-01
-7.74987519e-01 3.41404349e-01 -1.59859806e-01 5.07016063e-01
1.08842778e+00 8.61537084e-02 -2.45030165e-01 -5.62392533e-01
1.47441506e-01 -1.50755262e-02 -7.92621434e-01 1.10076241e-01
-7.66910538e-02 -7.52676368e-01 7.22324967e-01 -6.91134036e-01
-1.28895843e+00 4.87463884e-02 -8.96397382e-02 -7.22344220e-01
-8.66072625e-02 -7.55609214e-01 -5.31038046e-01 -6.59873366e-01
8.51190567e-01 2.22236082e-01 -5.10185122e-01 4.52302949e-04
6.28518939e-01 5.97021103e-01 8.93674135e-01 -5.28085589e-01
1.20866132e+00 9.13991094e-01 1.88072696e-01 -1.19035959e+00
-8.87176752e-01 1.36304826e-01 -2.38871694e-01 -6.15591109e-01
8.41436088e-01 -1.29914904e+00 -1.12152278e+00 6.16295218e-01
-7.66387522e-01 -9.32597280e-01 -6.01988256e-01 1.18237674e-01
-8.04798007e-01 1.31072074e-01 -3.86356473e-01 -1.02721739e+00
-1.05592512e-01 -9.18231666e-01 1.65665996e+00 2.15108410e-01
1.07343420e-01 -8.12863886e-01 2.14835107e-01 7.65684307e-01
2.51261860e-01 7.83420742e-01 2.52413154e-01 7.98652828e-01
-8.55008662e-01 2.99324151e-02 -1.33462578e-01 4.43184704e-01
2.54878789e-01 1.96162648e-02 -1.65056801e+00 -3.56371075e-01
1.13654882e-01 -5.53815842e-01 2.14318827e-01 4.32638735e-01
1.46905911e+00 -4.42911267e-01 2.89260745e-01 1.48669207e+00
1.85161269e+00 -2.52152413e-01 1.11730134e+00 3.45601700e-02
1.20620763e+00 3.78427237e-01 3.81600529e-01 4.66908455e-01
4.79395986e-01 8.05152237e-01 4.86994982e-01 -6.27511203e-01
-5.21863461e-01 -3.84303123e-01 5.57060957e-01 2.68552929e-01
-2.49506474e-01 -4.20876145e-01 -3.35604221e-01 3.96661982e-02
-8.67181063e-01 -9.82811391e-01 6.82456195e-02 2.55621076e+00
1.09527910e+00 -3.97776812e-01 -9.05013978e-02 -6.29928112e-02
5.03224790e-01 2.46625811e-01 -8.45483661e-01 -4.30284023e-01
-6.43950045e-01 1.63648278e-01 7.31746078e-01 8.07124496e-01
-6.56332493e-01 7.06390023e-01 7.06371021e+00 3.49320173e-01
-1.10444105e+00 -4.25067134e-02 1.13410807e+00 -4.91016448e-01
-8.92461360e-01 -3.90553296e-01 -4.99160498e-01 2.68982410e-01
7.60261893e-01 3.08505803e-01 1.29042375e+00 6.40398324e-01
2.38809615e-01 -1.18435554e-01 -8.34270418e-01 1.49411118e+00
7.47727752e-01 -7.73046970e-01 -1.64333045e-01 3.29500228e-01
1.14786541e+00 -2.02451766e-01 7.09044278e-01 -2.72863388e-01
4.29658622e-01 -1.13129807e+00 5.68891644e-01 8.11977923e-01
1.77674246e+00 -6.42179608e-01 -7.50450417e-02 -3.11057866e-01
-9.40139651e-01 1.52554765e-01 -2.54932821e-01 2.42405489e-01
2.02099949e-01 3.74299556e-01 -6.21487141e-01 1.68285400e-01
6.95974231e-01 4.41440374e-01 -4.05263424e-01 3.78257960e-01
-2.70179957e-01 2.23456249e-01 -4.65993583e-01 5.07332087e-01
-5.66629052e-01 -4.83164370e-01 5.85112154e-01 9.02219057e-01
2.65859425e-01 3.18212956e-01 -5.55512309e-02 7.40573525e-01
-3.99335474e-01 -1.70194268e-01 -7.66856253e-01 6.66853428e-01
1.79093838e-01 1.40211105e+00 -2.29436979e-01 -1.36903048e-01
-2.05209538e-01 1.65179336e+00 -1.93379521e-01 7.28658378e-01
-1.24074590e+00 -1.00096136e-01 1.04512131e+00 4.16656107e-01
-1.83241084e-01 -1.57085508e-01 -3.27357143e-01 -1.42058241e+00
1.64135545e-02 -1.06645226e+00 -3.13312739e-01 -1.80791938e+00
-1.32786107e+00 6.62270904e-01 -1.82694972e-01 -1.23090017e+00
1.62678584e-01 -4.46227640e-01 -3.82363766e-01 9.36892986e-01
-1.60745883e+00 -1.52645230e+00 -9.49097574e-01 1.03379154e+00
4.63016689e-01 2.36324802e-01 9.47138906e-01 2.48918280e-01
-3.21483523e-01 7.38463938e-01 1.85598489e-02 -2.07305640e-01
9.91982818e-01 -1.35901177e+00 4.77566719e-01 8.02730441e-01
1.31990924e-01 4.33278918e-01 9.26409423e-01 -4.68143404e-01
-2.04313827e+00 -1.12049472e+00 8.16359073e-02 -7.51629353e-01
3.13524008e-01 -8.04764807e-01 -4.11943734e-01 8.80640984e-01
2.20475748e-01 3.39523613e-01 8.59939814e-01 -2.31717229e-01
-6.71233237e-01 -5.69752455e-01 -1.61826956e+00 1.09603393e+00
1.27832139e+00 -7.75207758e-01 3.85976061e-02 5.82733572e-01
7.88334250e-01 -8.16304982e-01 -1.11219084e+00 1.89672448e-02
1.11294818e+00 -1.11034667e+00 1.38314748e+00 2.70989507e-01
2.31657118e-01 -5.33743501e-01 -6.57611907e-01 -1.17868590e+00
1.33336559e-01 -1.27332866e+00 9.73622650e-02 1.15690804e+00
-6.56398907e-02 -7.18915164e-01 5.25366902e-01 9.25588727e-01
2.00892493e-01 -2.89073825e-01 -6.35611653e-01 -3.55852038e-01
-3.89524996e-01 -5.14177382e-01 6.22069955e-01 8.71075749e-01
-9.27103400e-01 -9.12332255e-03 -9.26273108e-01 3.69825602e-01
1.30557775e+00 4.78606783e-02 1.38105047e+00 -7.26570666e-01
-4.11470383e-01 3.80800664e-01 -2.40708396e-01 -8.32032144e-01
1.85149342e-01 -2.65710920e-01 1.62425525e-02 -1.14317393e+00
3.64010110e-02 -4.79840666e-01 2.96916634e-01 1.43955857e-01
-2.58003119e-02 1.27605474e+00 6.97018579e-02 7.97981471e-02
-2.68300354e-01 3.25959146e-01 1.42074180e+00 3.75164226e-02
-2.10838675e-01 -2.55501002e-01 -6.54063880e-01 7.13627696e-01
3.86112005e-01 7.52567351e-02 -5.71522117e-01 -5.59810460e-01
5.26004851e-01 9.27860942e-03 8.57339859e-01 -1.11255109e+00
-3.35603476e-01 -1.54396757e-01 1.22177899e+00 -1.37208730e-01
9.34239507e-01 -1.01714265e+00 8.83232415e-01 -1.53253719e-01
-3.26177776e-01 -2.05706373e-01 1.72404706e-01 4.94699448e-01
6.10494435e-01 7.81637728e-01 1.03192604e+00 -7.51789361e-02
-1.07685685e-01 4.50388104e-01 1.65867046e-01 -7.79737085e-02
8.00615489e-01 -5.44677734e-01 -2.49575153e-01 -7.29851663e-01
-3.28045398e-01 -4.04903650e-01 1.39327121e+00 3.90850939e-02
7.50001967e-01 -1.36609507e+00 -6.40699685e-01 5.08074343e-01
-4.43528369e-02 -2.05548003e-01 5.76043189e-01 3.04662734e-01
-7.64151871e-01 -5.80614328e-01 -2.11490408e-01 -6.07283235e-01
-1.42038536e+00 4.32654619e-01 6.21206403e-01 2.36763284e-01
-8.73537302e-01 7.46885896e-01 4.83194202e-01 -2.17201680e-01
1.69591516e-01 -6.85425475e-02 4.28217709e-01 -4.53542143e-01
7.62957811e-01 1.96976751e-01 -1.32759660e-01 -8.50947201e-01
-1.03352010e-01 1.26186335e+00 5.11647224e-01 -4.11845714e-01
1.03776240e+00 -6.80552304e-01 2.41996735e-01 5.61273336e-01
1.38018453e+00 6.23264730e-01 -1.71135414e+00 8.51252154e-02
-1.51259744e+00 -1.12415612e+00 1.97695658e-01 -1.06532812e+00
-1.00651634e+00 4.68229979e-01 9.26455319e-01 -2.57131130e-01
1.81653428e+00 -4.59135056e-01 8.39885414e-01 1.29534379e-01
6.62881076e-01 -7.44863212e-01 1.58715025e-02 -2.56259590e-01
9.59078610e-01 -1.21231878e+00 2.99512178e-01 -4.44416225e-01
-7.48944044e-01 7.68880904e-01 2.72718608e-01 -2.55437881e-01
2.58278072e-01 6.29618168e-01 3.89765024e-01 1.63939133e-01
-3.85080695e-01 5.06217927e-02 2.70306855e-01 9.96257186e-01
1.02621451e-01 2.84166951e-02 7.23508000e-01 -2.98750937e-01
-5.91881931e-01 -1.62234023e-01 6.25909090e-01 3.15286487e-01
1.14523932e-01 -4.70757097e-01 -8.71468604e-01 2.53516566e-02
-3.81400645e-01 -1.47108495e-01 -2.56126791e-01 5.90949357e-01
2.07801402e-01 8.48456204e-01 1.43100932e-01 -3.62208426e-01
3.17443192e-01 -4.49319303e-01 1.03579378e+00 -1.38719589e-01
-4.73327309e-01 2.01268554e-01 9.46996957e-02 -8.68268549e-01
-3.93594772e-01 -4.90815639e-01 -4.77526575e-01 -5.75546265e-01
-1.64130807e-03 -6.78977430e-01 1.03715503e+00 1.61997363e-01
5.19978106e-02 3.87173802e-01 1.25992620e+00 -1.37224281e+00
2.15377361e-01 -3.46561134e-01 -7.60198653e-01 4.99648482e-01
7.27419317e-01 -2.95347810e-01 -7.52277434e-01 4.33065981e-01]
|
[9.896072387695312, -2.914862871170044]
|
fb60acec-9fa5-45fe-8697-de98fc73b657
|
novel-classification-of-ischemic-heart
|
2011.09801
| null |
https://arxiv.org/abs/2011.09801v1
|
https://arxiv.org/pdf/2011.09801v1.pdf
|
Novel Classification of Ischemic Heart Disease Using Artificial Neural Network
|
Ischemic heart disease (IHD), particularly in its chronic stable form, is a subtle pathology due to its silent behavior before developing in unstable angina, myocardial infarction or sudden cardiac death. Machine learning techniques applied to parameters extracted form heart rate variability (HRV) signal seem to be a valuable support in the early diagnosis of some cardiac diseases. However, so far, IHD patients were identified using Artificial Neural Networks (ANNs) applied to a limited number of HRV parameters and only to very few subjects. In this study, we used several linear and non-linear HRV parameters applied to ANNs, in order to confirm these results on a large cohort of 965 sample of subjects and to identify which features could discriminate IHD patients with high accuracy. By using principal component analysis and stepwise regression, we reduced the original 17 parameters to five, used as inputs, for a series of ANNs. The highest accuracy of 82% was achieved using meanRR, LFn, SD1, gender and age parameters and two hidden neurons.
|
['Agostino Accardo', 'Gianfranco Sinagra', 'Luca Restivo', 'Marco Merlo', 'Giulia Silveri']
|
2020-11-19
| null | null | null | null |
['heart-rate-variability']
|
['medical']
|
[ 8.29122365e-02 -1.05138846e-01 -1.22047924e-01 -2.36344814e-01
1.68222919e-01 -1.93188876e-01 1.03548467e-01 3.72429013e-01
-4.08022374e-01 1.08294415e+00 -4.02909145e-02 -4.03558046e-01
-3.69927734e-01 -7.34312415e-01 3.49584609e-01 -6.66802764e-01
-4.70404804e-01 6.51576400e-01 -3.52761209e-01 -2.75578424e-02
4.75703515e-02 7.93503046e-01 -1.39135683e+00 -2.12681815e-01
7.27225065e-01 6.67484403e-01 -2.88563758e-01 7.76518941e-01
3.48819852e-01 3.34108144e-01 -6.57717228e-01 2.39422575e-01
2.76733260e-03 -8.72744143e-01 -4.80421692e-01 -4.14159596e-01
-2.82755524e-01 -2.71508098e-01 2.03867525e-01 2.65629470e-01
1.04241359e+00 1.98237970e-02 7.86101103e-01 -7.94543624e-01
-1.90983132e-01 4.45586205e-01 2.95264949e-03 5.07137954e-01
1.63712382e-01 1.68883875e-01 4.35862511e-01 -6.46634519e-01
4.38256562e-01 8.71600330e-01 1.00411129e+00 4.50337917e-01
-1.59594047e+00 -5.64304829e-01 -7.04391479e-01 1.57562897e-01
-1.29377544e+00 -4.01029259e-01 8.52337182e-01 -7.15399086e-01
8.87580454e-01 3.78945947e-01 1.15675581e+00 8.32338929e-01
2.49092281e-01 -3.92794132e-01 1.43417311e+00 -5.32092333e-01
2.24161819e-01 3.15951645e-01 7.78812230e-01 2.83999950e-01
7.42494464e-01 3.79993856e-01 1.16771117e-01 -4.50324386e-01
6.85157657e-01 6.16160873e-03 -2.06311986e-01 1.55049548e-01
-1.26524627e+00 9.01924551e-01 -7.53302574e-02 8.08259308e-01
-7.17727005e-01 -3.73961896e-01 5.86982965e-01 5.22825181e-01
1.92460477e-01 3.72298092e-01 -7.64819026e-01 -1.88060924e-01
-6.99390352e-01 -3.23347487e-02 8.78741264e-01 -2.28764698e-01
3.24265301e-01 2.86779583e-01 -1.02827743e-01 7.00185418e-01
9.05920416e-02 7.03320026e-01 7.30068326e-01 -8.83602262e-01
-1.75817966e-01 7.29334354e-01 -2.71073341e-01 -1.29245782e+00
-1.15921140e+00 -7.91806340e-01 -1.56980503e+00 2.66319036e-01
3.98092479e-01 -2.84845412e-01 -4.78604645e-01 1.46913004e+00
1.66797861e-01 -2.37183869e-01 1.90830156e-01 8.05183887e-01
6.31177664e-01 2.38754183e-01 3.38431090e-01 -9.15343702e-01
1.48947906e+00 7.21495524e-02 -5.67491591e-01 2.27990493e-01
5.02463222e-01 -4.46075588e-01 6.97653115e-01 4.00792211e-01
-7.17670977e-01 -7.40888357e-01 -8.52696359e-01 3.88732851e-01
-2.60344654e-01 4.02727991e-01 2.42620170e-01 1.12189603e+00
-7.77374268e-01 1.11443114e+00 -6.86040759e-01 -7.04962552e-01
2.09813207e-01 3.33829433e-01 -2.72836924e-01 5.31423986e-01
-1.35205460e+00 1.05706584e+00 5.86484611e-01 1.64593607e-01
7.82880858e-02 -6.15918994e-01 -4.57829237e-01 -4.63550910e-02
-2.63453841e-01 -1.32915723e+00 1.57199547e-01 -4.96469617e-01
-1.33667934e+00 1.03457737e+00 -1.08220577e-01 -4.96317387e-01
4.12223220e-01 5.23564145e-02 -5.15249431e-01 3.58140707e-01
-2.55412340e-01 6.60315976e-02 6.89976513e-01 -7.54640460e-01
1.00549296e-01 -7.73187339e-01 -6.96376264e-01 -3.26053262e-01
-1.44474566e-01 -7.05210418e-02 6.98014677e-01 -4.56447005e-01
4.25620556e-01 -9.38569307e-01 -1.67267844e-01 -4.24753368e-01
-2.48923153e-01 9.64949206e-02 6.21142685e-01 -1.10695779e+00
1.33188856e+00 -2.04204297e+00 1.37979880e-01 6.32056415e-01
5.02413034e-01 4.45580959e-01 4.66252863e-01 1.67905971e-01
-2.78368413e-01 5.19156933e-01 -2.55227387e-01 1.72714382e-01
-2.56309330e-01 3.65177900e-01 1.72391698e-01 5.01862407e-01
1.63854823e-01 6.62584245e-01 -3.47580552e-01 -6.66910708e-01
4.79249805e-01 8.35933805e-01 -1.56044379e-01 3.16133127e-02
5.88992774e-01 7.86534846e-01 -3.68466407e-01 4.20185000e-01
5.24324119e-01 8.75888467e-02 3.06530476e-01 -1.58234358e-01
-1.01906024e-01 -3.71755987e-01 -8.67415249e-01 8.00867796e-01
-1.48285583e-01 5.02725482e-01 -4.72784996e-01 -1.23585379e+00
1.51416874e+00 7.08676398e-01 7.55302370e-01 -7.24705994e-01
4.15661573e-01 3.58323038e-01 4.61256027e-01 -7.04270542e-01
-4.72341120e-01 -4.20015931e-01 2.73025811e-01 3.43373835e-01
-2.63768077e-01 4.72721905e-01 3.12536418e-01 -5.53727090e-01
6.41349435e-01 -3.48392040e-01 9.86423373e-01 -2.95979679e-01
8.88050318e-01 -1.46242052e-01 6.12064600e-01 6.57802463e-01
-3.85231048e-01 6.09391391e-01 8.26130688e-01 -8.86647105e-01
-1.02289057e+00 -7.85791159e-01 -6.24750257e-01 1.82111248e-01
-7.23018527e-01 1.44081026e-01 -2.48138025e-01 6.62352890e-03
-5.83867310e-03 4.71583873e-01 -4.43430215e-01 -3.14327836e-01
-6.90345824e-01 -1.30967140e+00 6.80321574e-01 1.81912348e-01
4.06625062e-01 -1.14938819e+00 -1.00541151e+00 5.13251841e-01
4.71873544e-02 -6.41696334e-01 4.70690966e-01 2.18262240e-01
-1.61964428e+00 -1.30484831e+00 -9.28131700e-01 -5.19103944e-01
2.69147754e-01 -4.59653884e-01 1.15429890e+00 3.26786995e-01
-7.50040710e-01 -1.18378624e-01 -1.46628663e-01 -4.96209472e-01
-7.48578012e-01 3.20841998e-01 2.11844727e-01 -2.29564816e-01
2.25311816e-01 -1.22584701e+00 -7.56463110e-01 1.42624781e-01
-1.96161315e-01 -4.19404417e-01 8.36007357e-01 6.17766440e-01
3.14840525e-01 -2.03406513e-01 9.58379686e-01 -7.68309951e-01
5.19192338e-01 -3.33899468e-01 -1.69630423e-01 -2.93735892e-01
-1.14512455e+00 -3.11019540e-01 5.20583212e-01 -3.17081213e-01
-3.95291060e-01 -1.48909926e-01 -8.32383707e-02 -1.86577603e-01
-6.25465274e-01 4.47231501e-01 1.71336815e-01 2.98192739e-01
9.72011387e-01 2.41423965e-01 7.69532263e-01 -5.02190948e-01
-3.37924153e-01 4.07106817e-01 2.89574534e-01 -2.16124147e-01
6.85635507e-01 2.44476441e-02 6.96173787e-01 -1.30453050e+00
-2.23424047e-01 -4.23818260e-01 -9.84124839e-01 -1.39368147e-01
9.43963885e-01 -6.44739270e-01 -9.96311367e-01 3.98424923e-01
-7.12436855e-01 -4.49887896e-03 -2.87795216e-01 9.12304699e-01
-2.69734472e-01 2.35678673e-01 -3.81138980e-01 -1.03264153e+00
-1.01832497e+00 -3.60688776e-01 2.01476455e-01 1.54914409e-01
-6.58995092e-01 -1.04614425e+00 5.84672570e-01 4.47182953e-02
7.11901307e-01 9.66422915e-01 1.52000582e+00 -8.54819715e-01
3.80263537e-01 -3.72113466e-01 3.76328118e-02 4.94457126e-01
1.96535155e-01 1.56618342e-01 -7.17884958e-01 -1.50165275e-01
1.98406711e-01 2.00034395e-01 5.57955146e-01 6.94419503e-01
5.94128191e-01 1.96351614e-02 -2.01311305e-01 4.01318163e-01
1.35272264e+00 6.30887389e-01 6.75594449e-01 3.09737653e-01
3.79232228e-01 4.25117493e-01 -6.03720993e-02 4.88923788e-01
-1.10501386e-01 3.62860411e-01 -9.49392989e-02 -3.10719579e-01
6.15284666e-02 6.42453969e-01 1.50457108e-02 8.14477026e-01
-1.04598367e+00 3.87004465e-01 -8.93772662e-01 4.93044443e-02
-1.29433095e+00 -1.28856862e+00 -6.86616600e-01 2.42576718e+00
5.54252803e-01 2.15636581e-01 6.51759624e-01 7.36437976e-01
6.84537411e-01 -5.63169271e-02 -3.31321388e-01 -5.73320270e-01
-3.70974332e-01 3.54855955e-01 1.92296859e-02 -8.63101613e-03
-9.59912717e-01 -2.05929711e-01 6.75516367e+00 -3.85506421e-01
-1.32812560e+00 -2.09279522e-01 7.28531480e-01 2.60713965e-01
3.80873889e-01 -8.65005180e-02 -3.68312180e-01 2.94072866e-01
1.36872816e+00 -1.50176525e-01 2.52006739e-01 6.31787717e-01
7.36884952e-01 1.68392416e-02 -5.05467057e-01 1.02058792e+00
-8.31998736e-02 -7.83877194e-01 -3.27350944e-01 7.40344524e-02
1.55382603e-01 -4.48096752e-01 -1.88389942e-01 3.42134416e-01
-8.90481710e-01 -9.83049452e-01 -1.83782905e-01 9.23498869e-01
7.63125956e-01 -6.03816807e-01 1.25468481e+00 1.44768238e-01
-7.79103756e-01 -1.88603312e-01 -9.73456353e-02 -3.55846226e-01
1.94617603e-02 9.54776704e-01 -9.43907678e-01 2.76957750e-01
4.86424476e-01 5.25805414e-01 -6.11578465e-01 8.91463399e-01
3.38370442e-01 8.16367447e-01 -2.69092917e-01 -4.37437333e-02
-5.48099101e-01 -3.22740465e-01 7.58360803e-01 8.75208735e-01
3.95483464e-01 2.81879961e-01 -2.29042783e-01 8.61981452e-01
6.97893798e-01 4.55393136e-01 -4.74311501e-01 1.41950762e-02
1.97817057e-01 1.16515875e+00 -8.68553579e-01 -3.38144422e-01
-4.97469008e-02 4.43048865e-01 -4.48084712e-01 2.65907645e-01
-5.46108246e-01 -6.35968566e-01 4.41309571e-01 4.63690370e-01
-2.86987364e-01 -7.71039501e-02 -6.34212255e-01 -8.59081447e-01
-2.10235745e-01 -7.67806053e-01 5.42222321e-01 -2.62798667e-01
-9.03259635e-01 6.62561476e-01 9.49271321e-02 -1.17692423e+00
-4.95018572e-01 -4.64746952e-01 -7.56616354e-01 1.25008929e+00
-1.13231885e+00 -6.67544007e-01 -5.05752265e-01 3.90383512e-01
-1.14113398e-01 -3.09637606e-01 1.43636024e+00 3.97704720e-01
-8.11138570e-01 1.01820849e-01 -3.45759131e-02 1.07920147e-01
6.01834118e-01 -1.24654758e+00 -4.08542067e-01 3.57786030e-01
-6.49237692e-01 6.36061966e-01 7.83712924e-01 -5.18500805e-01
-7.06236780e-01 -6.73950911e-01 1.35775506e+00 -2.45270114e-02
1.24693504e-02 2.61801630e-01 -8.42627466e-01 2.08138660e-01
-2.08030149e-01 -1.31015226e-01 1.01251328e+00 1.25920594e-01
3.07738811e-01 -1.93053350e-01 -1.21864700e+00 1.71789259e-01
4.03329521e-01 -2.06681639e-01 -7.36250877e-01 -7.06003979e-02
1.14142813e-01 1.16145149e-01 -1.55458200e+00 5.26209950e-01
1.01677883e+00 -1.27923226e+00 1.29363906e+00 -5.40977359e-01
9.93230864e-02 -5.67553751e-02 2.99348652e-01 -9.72304344e-01
-4.58955616e-01 -4.98864651e-01 -7.20453740e-04 1.12782800e+00
2.91690409e-01 -1.12209761e+00 4.90094244e-01 5.13224483e-01
9.63885561e-02 -5.32713413e-01 -8.56306732e-01 -5.61178207e-01
-3.26712020e-02 1.84994191e-01 2.12412596e-01 9.23732102e-01
-2.47170642e-01 4.17609513e-01 -2.67720789e-01 -2.51837105e-01
7.89465189e-01 8.57023150e-02 4.90034431e-01 -2.14608216e+00
-4.01967159e-03 -4.24743712e-01 -7.05558598e-01 6.50376856e-01
-1.52811944e-01 -7.20521033e-01 -7.08727717e-01 -1.29775310e+00
3.47548798e-02 -4.03015584e-01 -5.68566680e-01 3.42449307e-01
-1.19951591e-01 4.67742622e-01 -1.08502917e-02 3.25960994e-01
5.79785526e-01 -6.00920469e-02 8.25179815e-01 2.50872076e-01
-1.01846647e+00 5.85025966e-01 -4.60392386e-01 6.79738879e-01
1.21410489e+00 -6.65916383e-01 -3.75700027e-01 6.21939957e-01
-5.09374365e-02 4.00511652e-01 5.45357049e-01 -1.25510573e+00
-5.16830683e-01 3.67840640e-02 1.04668915e+00 -4.10712928e-01
2.03907087e-01 -7.93361664e-01 7.74929523e-01 1.22737837e+00
7.17381528e-03 -6.04751334e-02 -2.72117630e-02 -5.93431108e-03
-2.84456350e-02 -6.04873896e-02 7.30735898e-01 -7.19914660e-02
-2.24625155e-01 -1.49745792e-01 -6.86496913e-01 1.81151647e-03
1.03298831e+00 -4.72416103e-01 -1.12515286e-01 -2.88736057e-02
-1.49665618e+00 -3.75031114e-01 4.01363932e-02 -1.11263387e-01
4.67326790e-01 -1.04037499e+00 -9.17202175e-01 3.10697854e-01
-8.86831731e-02 -4.07667249e-01 5.22625983e-01 1.53436267e+00
-9.54759240e-01 6.73813701e-01 -8.18058014e-01 -6.10892117e-01
-1.61123955e+00 4.92926180e-01 4.20135319e-01 -1.55823782e-01
-1.00775778e+00 -1.55983835e-01 -8.95478785e-01 -2.73492932e-02
-9.02012512e-02 -2.81164259e-01 -1.01219285e+00 6.91014826e-01
3.58513504e-01 9.40463006e-01 -7.37057328e-02 -6.29238546e-01
-4.67837572e-01 1.01341236e+00 5.53261161e-01 2.26374865e-01
1.20499146e+00 -1.84108123e-01 -3.17144126e-01 7.41872728e-01
1.23188233e+00 -2.51045078e-01 -1.59978837e-01 1.90841585e-01
-1.62158728e-01 3.67529169e-02 -4.95445505e-02 -6.66864634e-01
-9.35028195e-01 7.81553090e-01 1.21212471e+00 5.10502100e-01
1.24510765e+00 -3.37948114e-01 4.50773448e-01 5.79025209e-01
-6.48821220e-02 -6.79451883e-01 -5.46682358e-01 -4.17726710e-02
6.82137728e-01 -1.00440657e+00 2.15577886e-01 -8.27957764e-02
-3.49021941e-01 1.75429463e+00 -8.63549300e-03 -6.97706044e-02
9.02381539e-01 -2.86246330e-01 2.68611073e-01 4.93922010e-02
-3.98753673e-01 -5.93576580e-02 -1.78099684e-02 7.84769833e-01
5.56395948e-01 1.19756229e-01 -9.84591663e-01 5.89023650e-01
-2.29830995e-01 4.53912526e-01 3.88169736e-01 7.12627769e-01
-5.18367946e-01 -8.73603642e-01 -6.16784036e-01 7.88798809e-01
-6.64765596e-01 1.36373714e-01 -7.51893073e-02 1.11465931e+00
1.84806749e-01 6.84395015e-01 4.29723486e-02 -9.28883776e-02
5.38873374e-01 7.40631640e-01 1.08461581e-01 -5.91864362e-02
-5.95423341e-01 -2.08021160e-02 7.87334517e-02 -1.53316408e-01
-6.62328720e-01 -8.77700925e-01 -1.05751944e+00 -3.37000668e-01
1.06423430e-01 1.66394208e-02 6.20647192e-01 8.06378484e-01
2.30011433e-01 7.21514046e-01 7.59774923e-01 -4.87753302e-01
-6.21855140e-01 -1.24854302e+00 -9.58625615e-01 2.48753458e-01
4.46071029e-01 -3.57536733e-01 -7.62325406e-01 3.11400712e-01]
|
[14.063065528869629, 3.1151304244995117]
|
439274d6-e486-418d-a5c9-72fd1944d642
|
hamiltonian-prior-to-disentangle-content-and
|
2112.01641
| null |
https://arxiv.org/abs/2112.01641v4
|
https://arxiv.org/pdf/2112.01641v4.pdf
|
Hamiltonian latent operators for content and motion disentanglement in image sequences
|
We introduce \textit{HALO} -- a deep generative model utilising HAmiltonian Latent Operators to reliably disentangle content and motion information in image sequences. The \textit{content} represents summary statistics of a sequence, and \textit{motion} is a dynamic process that determines how information is expressed in any part of the sequence. By modelling the dynamics as a Hamiltonian motion, important desiderata are ensured: (1) the motion is reversible, (2) the symplectic, volume-preserving structure in phase space means paths are continuous and are not divergent in the latent space. Consequently, the nearness of sequence frames is realised by the nearness of their coordinates in the phase space, which proves valuable for disentanglement and long-term sequence generation. The sequence space is generally comprised of different types of dynamical motions. To ensure long-term separability and allow controlled generation, we associate every motion with a unique Hamiltonian that acts in its respective subspace. We demonstrate the utility of \textit{HALO} by swapping the motion of a pair of sequences, controlled generation, and image rotations.
|
['Amos Storkey', 'Asif Khan']
|
2021-12-02
| null | null | null | null |
['motion-disentanglement']
|
['computer-vision']
|
[ 3.26402277e-01 -6.23442791e-02 -1.57266304e-01 2.20014721e-01
-4.72520292e-02 -1.05641389e+00 1.18962383e+00 -3.25138897e-01
-2.24568456e-01 7.27561593e-01 3.23949158e-01 -5.01976945e-02
-2.24724472e-01 -6.71907008e-01 -4.22738463e-01 -1.38981891e+00
-3.06997865e-01 2.55424529e-01 -5.54543920e-02 -1.63919613e-01
1.25656962e-01 4.57813203e-01 -1.08529508e+00 -2.00782686e-01
4.21502382e-01 3.89845699e-01 2.43786708e-01 1.11794019e+00
3.22032273e-01 9.73488748e-01 -3.28161716e-01 -1.38304085e-01
4.53385264e-01 -9.95156407e-01 -8.20299625e-01 2.46302098e-01
1.75395086e-01 -1.62076965e-01 -6.82061315e-01 1.36038971e+00
1.35510430e-01 1.33858606e-01 8.82091165e-01 -1.41148090e+00
-5.29127181e-01 2.73273170e-01 -3.60413343e-01 1.59456596e-01
3.75222445e-01 3.98189455e-01 1.20510626e+00 -3.24728012e-01
1.18679500e+00 9.76900637e-01 2.48548105e-01 4.44736302e-01
-1.82352948e+00 -2.24338233e-01 -2.61319876e-01 -5.54719083e-02
-1.38063490e+00 -4.04486418e-01 7.82619178e-01 -8.41814041e-01
4.84604776e-01 7.75233388e-01 1.03828502e+00 1.10556340e+00
8.31503093e-01 4.78075862e-01 8.08260202e-01 -3.20872813e-01
3.30131710e-01 -3.56332481e-01 -4.20834981e-02 8.10832500e-01
9.01570246e-02 1.52503625e-01 -6.30940020e-01 -4.51278538e-02
1.14110553e+00 -2.57252395e-01 -7.17115760e-01 -8.26768994e-01
-1.87199378e+00 8.26610029e-01 -1.33761898e-01 3.73443484e-01
-1.43518001e-01 1.92061782e-01 2.63910711e-01 3.86701703e-01
-3.13125961e-02 4.12699014e-01 3.23500574e-01 -2.73851275e-01
-9.30115461e-01 4.27584112e-01 9.24545050e-01 8.79823804e-01
8.34854901e-01 -2.18077879e-02 -1.51301906e-01 4.17586276e-03
3.13510112e-02 6.45736217e-01 5.24439275e-01 -1.41148138e+00
1.64948702e-01 1.12335153e-01 1.05618559e-01 -1.19003987e+00
-1.41915604e-01 -1.40906051e-01 -1.18719983e+00 1.81960732e-01
3.90538633e-01 1.21336997e-01 -7.72259772e-01 2.18602514e+00
-1.48121445e-02 7.97748119e-02 -4.81835343e-02 6.57246292e-01
2.26257741e-01 8.37531865e-01 -1.53127044e-01 -7.64446557e-01
1.16717446e+00 -3.78960341e-01 -8.64426136e-01 2.55885482e-01
4.45316374e-01 -6.51178598e-01 7.43459105e-01 5.92505336e-02
-1.31024730e+00 -3.66416186e-01 -1.21250772e+00 1.64308175e-01
8.79865885e-02 -2.82615304e-01 -7.20618963e-02 3.11388850e-01
-1.18910968e+00 7.55408943e-01 -1.07563007e+00 -7.71149481e-03
-2.46492475e-01 3.06084245e-01 -6.69826686e-01 3.69100779e-01
-1.06606042e+00 5.74705839e-01 4.31455910e-01 -9.75738559e-03
-9.56083298e-01 -3.77155155e-01 -9.01010513e-01 -1.23932481e-01
-6.68584555e-02 -1.00215065e+00 9.16190863e-01 -7.68329859e-01
-1.32317793e+00 8.17518234e-01 -2.77787924e-01 -1.85364425e-01
8.55331659e-01 2.02006310e-01 -1.26598805e-01 4.61233228e-01
1.50218442e-01 6.69397473e-01 1.20884264e+00 -9.18411851e-01
-1.82680517e-01 -5.92728406e-02 -1.23405345e-01 4.13742840e-01
-4.61401641e-02 -2.55761415e-01 -2.13669896e-01 -7.09597707e-01
2.27174684e-01 -1.36368978e+00 -1.10191993e-01 -1.78176090e-01
-6.51547968e-01 2.34969631e-01 9.50154841e-01 -4.81105834e-01
1.37123728e+00 -2.25722241e+00 8.43985498e-01 2.82528605e-02
5.38766682e-01 -6.35930449e-02 -2.73054056e-02 6.44863009e-01
-2.82466918e-01 2.17281595e-01 -4.91923839e-01 -1.57566354e-01
-6.19753450e-02 4.83819991e-01 -3.65821600e-01 9.23318326e-01
1.81814209e-01 1.02300572e+00 -1.03239942e+00 -4.37386781e-01
2.17189476e-01 4.31259960e-01 -4.71313000e-01 4.10627425e-02
-2.43456513e-01 1.00713503e+00 -4.39428478e-01 -1.84272066e-01
4.01801497e-01 -1.98168740e-01 2.14640379e-01 -1.34060621e-01
-4.00208414e-01 1.37913942e-01 -1.22681224e+00 1.49722445e+00
7.96958804e-02 8.38570297e-01 -4.76064794e-02 -6.77684128e-01
4.40597802e-01 4.34894919e-01 7.48627126e-01 -4.15409327e-01
-2.01266054e-02 3.78787853e-02 2.67796904e-01 -5.02772689e-01
4.89672273e-01 -9.57461968e-02 -8.90190378e-02 6.66224599e-01
-6.15105815e-02 -1.89998105e-01 4.92334545e-01 3.74152899e-01
1.03252578e+00 1.05152152e-01 3.67821664e-01 -7.67209232e-01
4.43074018e-01 -2.14932844e-01 5.89720190e-01 6.28419578e-01
-3.46698016e-01 6.55187190e-01 8.62406790e-01 -2.02133060e-01
-1.55884290e+00 -1.19828713e+00 -1.30780548e-01 3.86393964e-01
3.01972568e-01 -3.84431392e-01 -8.17978263e-01 -7.67962560e-02
-4.14999843e-01 4.77625847e-01 -8.24241459e-01 -3.13020945e-01
-8.58939946e-01 -7.22489953e-01 3.54510903e-01 4.80889790e-02
4.96945441e-01 -9.73889887e-01 -1.00627232e+00 1.34372085e-01
-5.95619321e-01 -9.04754698e-01 -9.59807813e-01 -3.71085517e-02
-6.37256622e-01 -1.02751160e+00 -7.54681826e-01 -5.30437171e-01
6.48061574e-01 1.09526888e-01 9.04206932e-01 -3.51146817e-01
-2.77372539e-01 3.89782280e-01 2.16164775e-02 4.80708331e-01
-8.08633149e-01 -2.23257020e-01 3.28172520e-02 2.42025957e-01
-3.66433024e-01 -8.03502440e-01 -7.39703536e-01 2.83446610e-01
-1.22230887e+00 3.06112319e-01 -6.74858019e-02 8.49376738e-01
5.64553499e-01 1.95253998e-01 -2.27694631e-01 -3.80811960e-01
4.29281116e-01 -2.85445243e-01 -4.21169072e-01 1.38095573e-01
-1.33231282e-01 5.29256046e-01 2.18546927e-01 -6.26017809e-01
-7.96454072e-01 -4.49956767e-02 2.26980686e-01 -2.94684112e-01
1.27506360e-01 2.94532388e-01 -2.59449929e-01 2.24183515e-01
3.96548778e-01 5.81500649e-01 2.90097743e-01 6.08432665e-02
5.68536639e-01 1.01528853e-01 6.70897245e-01 -3.02148819e-01
6.83009267e-01 8.46515059e-01 5.58794618e-01 -1.20717657e+00
-2.89970905e-01 -2.39081264e-01 -1.07411480e+00 -2.42793724e-01
1.38054895e+00 -5.96118271e-01 -8.63265932e-01 6.36097074e-01
-1.07945418e+00 -3.74402434e-01 -5.24002552e-01 4.03305143e-01
-1.02956426e+00 6.75476313e-01 -5.53038359e-01 -7.96085715e-01
-2.53673662e-02 -1.38482332e+00 9.51845348e-01 -1.86832815e-01
-7.11817443e-01 -1.26090991e+00 4.51388359e-01 -1.99105769e-01
-5.42250164e-02 6.03899002e-01 1.05986130e+00 -1.10073306e-01
-8.56937051e-01 -1.73154175e-02 4.33518052e-01 1.69414505e-01
8.56014043e-02 3.35838377e-01 -4.71626580e-01 -3.99653673e-01
4.51531023e-01 1.86578512e-01 7.32578635e-01 3.92446131e-01
4.21344519e-01 -6.69569194e-01 -1.31787598e-01 6.78275943e-01
1.09762597e+00 3.57918680e-01 5.81192374e-01 1.66309044e-01
8.84112239e-01 5.15707970e-01 -1.02052890e-01 2.22118765e-01
9.44155231e-02 7.90756404e-01 2.32218862e-01 2.39769697e-01
-1.46231189e-01 -3.07945549e-01 6.53559029e-01 1.27820194e+00
-1.19027190e-01 -2.83519924e-01 -6.57339990e-01 4.36178952e-01
-1.86623943e+00 -1.45857418e+00 -2.69217968e-01 2.33489990e+00
8.38567019e-01 -4.06295992e-02 8.94171894e-02 1.92310005e-01
6.66517079e-01 6.13614261e-01 -4.46019888e-01 -1.23639524e-01
-4.57697839e-01 -8.66205469e-02 4.28174347e-01 8.01915824e-01
-9.44351256e-01 4.20765907e-01 6.67174387e+00 6.19574547e-01
-1.14859879e+00 -8.91693905e-02 3.15853924e-01 -6.15964755e-02
-6.28451467e-01 1.91882536e-01 -4.55987185e-01 6.74123287e-01
5.79939127e-01 -2.49490023e-01 2.92950004e-01 1.96387842e-01
3.26319426e-01 -2.17163891e-01 -1.22436738e+00 7.04038262e-01
-1.02447942e-01 -1.42274153e+00 2.29989916e-01 5.35884380e-01
8.09116483e-01 -3.66449088e-01 3.25821996e-01 -3.98576289e-01
2.20652729e-01 -8.38337481e-01 1.13574469e+00 7.26487339e-01
8.05518687e-01 -6.37713552e-01 1.81924060e-01 5.66044331e-01
-1.17570913e+00 2.27463171e-01 2.63468903e-02 7.75947422e-02
5.08855581e-01 4.00961995e-01 -2.93680251e-01 5.05978167e-01
3.04308478e-02 8.06496024e-01 -1.89341471e-01 4.85309303e-01
-2.66554415e-01 4.25542444e-01 -1.96104180e-02 2.55128950e-01
1.92558795e-01 -9.92782235e-01 1.20791483e+00 9.68570769e-01
3.38952750e-01 2.14018539e-01 1.88013270e-01 9.83117759e-01
2.91954339e-01 -4.87843186e-01 -7.31907725e-01 -3.30428183e-01
2.07271159e-01 8.82281423e-01 -8.60331416e-01 -1.00764677e-01
4.37750481e-03 1.18864608e+00 -2.31829718e-01 6.37843788e-01
-5.19649446e-01 -8.75766501e-02 1.00339949e+00 6.88203946e-02
3.24820429e-01 -8.35940182e-01 -1.95220768e-01 -1.34432352e+00
-9.12883654e-02 -6.24611914e-01 6.13984279e-02 -7.50643253e-01
-6.44538820e-01 5.08646786e-01 1.55316874e-01 -1.18845427e+00
-5.87179303e-01 -1.98064864e-01 -5.09955943e-01 1.00423777e+00
-7.40305364e-01 -8.29906642e-01 8.95328000e-02 6.95584714e-01
2.51414627e-01 1.12330489e-01 6.75423086e-01 -2.23805144e-01
-5.72821796e-01 1.82639256e-01 2.10298091e-01 7.48834237e-02
9.13946480e-02 -1.37310839e+00 4.70030636e-01 1.11967540e+00
3.36982340e-01 8.61326098e-01 9.99286354e-01 -8.54116321e-01
-1.44082856e+00 -8.29076529e-01 9.31231081e-01 -5.47373235e-01
9.66228127e-01 -4.61805314e-01 -6.94606066e-01 8.07748079e-01
3.40368688e-01 -2.27533698e-01 3.53271276e-01 -6.25268281e-01
-1.86035052e-01 4.17998135e-01 -6.46851242e-01 1.02360129e+00
1.12636960e+00 -8.16376269e-01 -3.03193748e-01 8.98106322e-02
6.84272170e-01 -2.48021215e-01 -7.40319908e-01 1.94135085e-01
5.87190986e-01 -1.13269365e+00 7.72361457e-01 -4.20343459e-01
5.52656054e-01 -4.32635009e-01 -7.68835098e-02 -1.00308645e+00
-5.52712083e-01 -1.26300752e+00 -2.99428821e-01 8.27790618e-01
-8.10728669e-02 -4.24727529e-01 5.69750786e-01 3.64972711e-01
3.09576571e-01 -3.22811902e-01 -9.65571880e-01 -8.73339176e-01
1.74525470e-01 6.88277483e-02 3.33076715e-01 9.52770293e-01
-4.39948626e-02 3.10897082e-01 -5.34066558e-01 -1.39746424e-02
6.30494297e-01 5.42196967e-02 4.43271965e-01 -8.36596251e-01
-3.24586093e-01 -6.11447036e-01 -7.42844939e-01 -1.15908444e+00
1.06749736e-01 -8.50024581e-01 2.18003970e-02 -1.11334181e+00
5.37137032e-01 1.89207438e-02 1.77418455e-01 -6.38776692e-03
6.83990866e-02 9.19814110e-02 3.93579751e-01 8.40251863e-01
-1.29984915e-01 6.96622074e-01 1.62270713e+00 1.79184139e-01
-1.41447440e-01 -2.04005718e-01 -1.04124151e-01 5.99931598e-01
4.30458665e-01 -1.76709384e-01 -5.62890470e-01 -1.25452280e-01
4.08049434e-01 4.54729199e-01 5.35630047e-01 -6.60313189e-01
4.59255986e-02 -2.57390022e-01 -3.26266028e-02 -3.01139861e-01
3.11911911e-01 -3.55997264e-01 7.56567538e-01 6.88236713e-01
-5.34143388e-01 2.49742851e-01 -3.60264540e-01 6.20133042e-01
-1.08691096e-01 -1.69795394e-01 7.67388642e-01 -1.90667734e-01
-3.08421910e-01 3.15451562e-01 -6.56048417e-01 1.14895031e-01
1.03376830e+00 -3.55953366e-01 -3.38111594e-02 -6.43034279e-01
-8.65702152e-01 -1.58684477e-01 8.38303566e-01 9.25588459e-02
1.81261688e-01 -1.48198879e+00 -5.26411116e-01 4.71378803e-01
-1.88695744e-01 -6.63693547e-02 1.92291722e-01 9.93536830e-01
-6.96276248e-01 4.26700562e-01 -3.15489233e-01 -8.16002846e-01
-1.17746007e+00 3.94242734e-01 3.53945196e-01 -3.36174309e-01
-7.98251510e-01 6.00256264e-01 5.58173895e-01 9.29211751e-02
-3.00340533e-01 -3.34598906e-02 -1.65497307e-02 1.95824295e-01
3.67279619e-01 4.15953726e-01 -3.22188407e-01 -1.23467290e+00
7.20295636e-03 5.44564366e-01 1.30869016e-01 -7.78965473e-01
8.67409110e-01 -3.47619176e-01 -4.44972068e-01 7.71877646e-01
1.45854509e+00 4.42053527e-02 -1.66657698e+00 2.32712597e-01
-1.89153478e-01 -2.32502535e-01 -1.85044974e-01 -1.72889426e-01
-7.21679807e-01 7.08562315e-01 1.71828091e-01 6.33835435e-01
9.21487212e-01 -1.42164722e-01 5.67528009e-01 -9.76308957e-02
2.39528790e-01 -6.87730014e-01 1.88281789e-01 5.94859898e-01
9.30688679e-01 -5.65879047e-01 -2.93499678e-01 -2.69659817e-01
-5.72651684e-01 1.08906376e+00 -2.65664041e-01 -1.70605049e-01
5.74137211e-01 1.67717546e-01 -2.56758481e-01 -1.43095255e-01
-5.61380684e-01 1.17720008e-01 3.75098169e-01 3.97916168e-01
3.20070624e-01 1.91395357e-01 -2.59956539e-01 -3.75467390e-01
-6.53852880e-01 -3.56536508e-01 7.38807857e-01 9.23014820e-01
-2.11253643e-01 -1.15256000e+00 -3.02653730e-01 -1.03523619e-01
-4.07746285e-02 7.49362409e-02 -2.17873529e-01 8.71508896e-01
4.11073044e-02 6.16052091e-01 2.89179564e-01 -1.48091093e-01
-1.59462482e-01 6.10881634e-02 6.98239148e-01 -2.80767530e-01
1.62315771e-01 2.58044004e-01 -2.30955005e-01 -6.02217317e-01
-4.40258324e-01 -1.17922747e+00 -1.04902697e+00 -4.44131702e-01
1.51870493e-02 1.33502096e-01 2.87499309e-01 9.46283460e-01
1.07991718e-01 4.41758037e-01 6.98106170e-01 -6.70933902e-01
-5.83007336e-01 -6.01863921e-01 -7.06621706e-01 5.34260452e-01
8.13987911e-01 -4.14794266e-01 -3.98198485e-01 7.35298932e-01]
|
[10.881400108337402, -0.7239387631416321]
|
8151ff1c-ca75-4ccf-98e0-18f24793a9b7
|
calibrate-the-inter-observer-segmentation
|
2208.03016
| null |
https://arxiv.org/abs/2208.03016v1
|
https://arxiv.org/pdf/2208.03016v1.pdf
|
Calibrate the inter-observer segmentation uncertainty via diagnosis-first principle
|
On the medical images, many of the tissues/lesions may be ambiguous. That is why the medical segmentation is typically annotated by a group of clinical experts to mitigate the personal bias. However, this clinical routine also brings new challenges to the application of machine learning algorithms. Without a definite ground-truth, it will be difficult to train and evaluate the deep learning models. When the annotations are collected from different graders, a common choice is majority vote. However such a strategy ignores the difference between the grader expertness. In this paper, we consider the task of predicting the segmentation with the calibrated inter-observer uncertainty. We note that in clinical practice, the medical image segmentation is usually used to assist the disease diagnosis. Inspired by this observation, we propose diagnosis-first principle, which is to take disease diagnosis as the criterion to calibrate the inter-observer segmentation uncertainty. Following this idea, a framework named Diagnosis First segmentation Framework (DiFF) is proposed to estimate diagnosis-first segmentation from the raw images.Specifically, DiFF will first learn to fuse the multi-rater segmentation labels to a single ground-truth which could maximize the disease diagnosis performance. We dubbed the fused ground-truth as Diagnosis First Ground-truth (DF-GT).Then, we further propose Take and Give Modelto segment DF-GT from the raw image. We verify the effectiveness of DiFF on three different medical segmentation tasks: OD/OC segmentation on fundus images, thyroid nodule segmentation on ultrasound images, and skin lesion segmentation on dermoscopic images. Experimental results show that the proposed DiFF is able to significantly facilitate the corresponding disease diagnosis, which outperforms previous state-of-the-art multi-rater learning methods.
|
['Yanwu Xu', 'Huiying Liu', 'Weihua Yang', 'Mingkui Tan', 'Lixin Duan', 'Hoayi Xiong', 'Huihui Fang', 'Junde Wu']
|
2022-08-05
| null | null | null | null |
['skin-lesion-segmentation']
|
['medical']
|
[ 3.91122431e-01 5.64252257e-01 -5.72334170e-01 -6.32722616e-01
-1.32672656e+00 -3.92103076e-01 1.42468914e-01 6.71771914e-02
-3.36294919e-01 7.15437353e-01 -1.19126610e-01 -3.51942420e-01
-2.65190005e-01 -5.32087743e-01 -4.45649743e-01 -1.14030933e+00
6.98092163e-01 8.57442737e-01 2.84256544e-02 4.51258540e-01
-1.10114627e-01 8.11269358e-02 -9.48240340e-01 3.98365915e-01
1.42469525e+00 1.12561643e+00 2.40632951e-01 5.35475910e-01
-6.24920763e-02 6.10629082e-01 -6.40641212e-01 -6.19888306e-01
1.46955863e-01 -5.83234251e-01 -1.00223720e+00 5.32526851e-01
9.26299617e-02 -4.44996715e-01 1.91154391e-01 1.41691434e+00
4.23705310e-01 -3.68433833e-01 8.88433576e-01 -8.75440359e-01
-5.46524405e-01 7.89414346e-01 -9.03452039e-01 -2.27779560e-02
-7.32555687e-02 3.61641496e-02 8.85335505e-01 -2.53835440e-01
4.27518457e-01 9.48669612e-01 4.97457474e-01 7.52507031e-01
-7.93892801e-01 -4.53911453e-01 2.22989582e-02 -9.05340612e-02
-1.41434431e+00 -3.37141403e-03 3.71792436e-01 -5.18274963e-01
-1.68072194e-01 2.93907166e-01 5.18898368e-01 1.03496599e+00
2.86463797e-01 9.55657780e-01 1.49433780e+00 -2.97191799e-01
2.64285743e-01 2.90118724e-01 1.95187792e-01 7.34909296e-01
3.61569464e-01 -1.13876751e-02 8.57789218e-02 -4.77049313e-02
7.26876974e-01 1.33748818e-02 -5.23190975e-01 -3.40323150e-02
-9.66206551e-01 7.03053296e-01 5.68019450e-01 3.40044826e-01
-3.04187000e-01 -5.35935871e-02 1.73262745e-01 -6.71382397e-02
4.95858163e-01 1.00642726e-01 -2.17027843e-01 3.20924461e-01
-1.23468232e+00 -1.02024876e-01 6.88673496e-01 4.97537434e-01
3.34150195e-01 -3.97915989e-01 -3.60402405e-01 8.34503949e-01
5.74096918e-01 5.23988962e-01 6.58038378e-01 -8.23113918e-01
1.20949969e-02 5.96371472e-01 7.14101940e-02 -6.20315611e-01
-3.49942327e-01 -6.89563453e-01 -1.07945204e+00 5.25699668e-02
6.38378441e-01 -1.84566885e-01 -1.50727630e+00 1.44236875e+00
5.94023347e-01 2.12070778e-01 4.47308868e-02 1.21994281e+00
9.11091268e-01 3.56344640e-01 1.39980495e-01 -4.40812320e-01
1.47069287e+00 -9.30265129e-01 -8.57502222e-01 3.26781012e-02
7.49454916e-01 -6.09364152e-01 6.39355302e-01 5.28148413e-01
-8.91880274e-01 -4.26038623e-01 -8.62242758e-01 3.38660926e-01
1.10071726e-01 5.99215269e-01 4.07643348e-01 7.91204572e-01
-8.77204359e-01 4.81920719e-01 -1.03022444e+00 -2.19598413e-01
5.95838606e-01 3.36097807e-01 -1.52653635e-01 -1.19462013e-01
-1.23276854e+00 8.84591401e-01 5.09067595e-01 3.21786135e-01
-1.04906058e+00 -7.12825716e-01 -4.97199833e-01 -4.78938937e-01
5.77247858e-01 -7.23063827e-01 1.45119429e+00 -1.28848875e+00
-1.43416405e+00 1.41067934e+00 -1.75594781e-02 -3.37191194e-01
9.35322046e-01 1.47232071e-01 -2.79048890e-01 1.99864894e-01
-1.10583836e-02 6.06206238e-01 8.69286180e-01 -1.53170478e+00
-8.21042359e-01 -3.82321119e-01 3.51320393e-02 2.38363907e-01
1.16384082e-01 -2.82748282e-01 -4.68723178e-01 -4.35502619e-01
1.26826152e-01 -1.08952081e+00 -4.31201041e-01 1.52230769e-01
-8.30745697e-01 -2.03796521e-01 3.66495430e-01 -8.20228338e-01
1.14304626e+00 -2.01928163e+00 5.44025414e-02 3.79924148e-01
5.85527897e-01 4.01600182e-01 2.63212383e-01 -5.76930940e-01
-4.66932915e-02 3.24755400e-01 -5.49016953e-01 -3.16412181e-01
-1.67668238e-01 3.42176139e-01 4.11964059e-02 6.37681663e-01
8.42328072e-02 8.16420257e-01 -8.73484075e-01 -9.81624186e-01
2.36999184e-01 3.14707637e-01 -3.18062544e-01 4.84267205e-01
-2.05689579e-01 9.11958396e-01 -6.50912821e-01 8.58587027e-01
7.35478044e-01 -7.39072025e-01 1.65810242e-01 -4.41291511e-01
3.24710637e-01 -2.10711494e-01 -1.09561908e+00 1.57269275e+00
-4.01015431e-01 2.93219760e-02 1.20017536e-01 -1.00155413e+00
7.91973650e-01 6.17655575e-01 6.85277522e-01 -1.89903393e-01
5.22848845e-01 6.34243786e-01 3.37038517e-01 -6.21954143e-01
-1.04134195e-01 -3.15666378e-01 5.54136895e-02 2.67066300e-01
-9.15089101e-02 -4.36751157e-01 -7.78119713e-02 -5.60251437e-02
7.44064569e-01 -1.55221805e-01 6.17073059e-01 -1.17554069e-01
6.98450506e-01 -4.15722840e-02 7.81678140e-01 6.34256899e-01
-6.08711600e-01 8.25490713e-01 5.81746399e-01 -2.28437051e-01
-6.66165471e-01 -1.15433419e+00 -6.00559175e-01 4.49166983e-01
3.77215624e-01 2.10218430e-01 -1.08879173e+00 -1.11675692e+00
-1.51386857e-01 3.90766501e-01 -7.01451838e-01 -1.24648875e-02
-1.58308625e-01 -1.16808820e+00 5.30114055e-01 3.19272727e-01
5.49801111e-01 -7.09469080e-01 -3.23066741e-01 -3.41071151e-02
-3.49277109e-01 -9.63934422e-01 -4.62374985e-01 -1.16582140e-01
-7.69852400e-01 -1.22235799e+00 -1.23998904e+00 -5.72485924e-01
9.23769832e-01 -2.00165153e-01 8.88005078e-01 8.12576190e-02
-2.62672156e-01 1.74980566e-01 -2.28345692e-01 -3.08649659e-01
-8.00949991e-01 1.09137602e-01 -2.99711496e-01 3.06304097e-01
2.55400032e-01 -2.10306898e-01 -7.99582839e-01 3.17493230e-01
-9.49736297e-01 2.27183938e-01 8.97691071e-01 9.04381692e-01
1.07772672e+00 -5.14718704e-03 4.95618075e-01 -1.36602974e+00
4.41945583e-01 -4.00128573e-01 -4.96851444e-01 6.89166784e-01
-6.87814415e-01 -9.05610472e-02 2.96924144e-01 -3.47694039e-01
-1.15945148e+00 7.34261796e-02 -2.25192696e-01 -4.72849339e-01
-2.10907027e-01 6.97723031e-01 -2.55323807e-03 -6.49934113e-02
5.80964983e-01 -1.92634478e-01 -1.44540876e-01 -9.52987671e-02
2.48844996e-01 1.04313779e+00 5.34087598e-01 -5.39851964e-01
3.19295704e-01 4.40370172e-01 -5.52433878e-02 -1.93445206e-01
-1.42807817e+00 -4.04452711e-01 -4.34973687e-01 -3.14811707e-01
1.12567472e+00 -8.44431221e-01 -5.09416819e-01 5.54576635e-01
-1.14317930e+00 -2.05995247e-01 -1.78578004e-01 5.42210221e-01
-4.25292224e-01 2.85037816e-01 -7.27661073e-01 -7.42647946e-01
-4.28048342e-01 -1.71671832e+00 1.26349723e+00 6.05917335e-01
-1.49711192e-01 -1.23903322e+00 -2.92993616e-02 6.90425396e-01
7.20233172e-02 4.23865467e-01 7.93867409e-01 -7.49519885e-01
-4.96463597e-01 -1.13455087e-01 -4.17373210e-01 4.37650025e-01
4.21626031e-01 1.45491004e-01 -1.09877682e+00 -8.43064860e-02
1.67245895e-01 -3.77687007e-01 8.59386504e-01 9.70247030e-01
1.53739321e+00 1.52431548e-01 -4.11402643e-01 6.63061857e-01
1.40619230e+00 1.51333034e-01 3.90883684e-01 -2.53336072e-01
8.39915454e-01 5.65280497e-01 6.53319538e-01 2.55718797e-01
4.39238071e-01 3.83568615e-01 3.74389619e-01 -4.21121627e-01
-1.50457799e-01 -4.00819117e-03 -9.81404483e-02 9.70974922e-01
-1.35613173e-01 -5.31893075e-01 -9.50217783e-01 6.37367845e-01
-1.73488986e+00 -2.19921440e-01 3.39397090e-03 2.15530753e+00
1.33854675e+00 2.29016468e-02 -1.99950710e-01 -7.68992156e-02
1.09990644e+00 -3.59339453e-02 -6.71712637e-01 -2.14763626e-01
1.57311499e-01 4.76318188e-02 6.62040293e-01 6.26671255e-01
-1.14846611e+00 5.62541366e-01 5.61242199e+00 1.27647495e+00
-1.26878119e+00 3.11816365e-01 1.24937177e+00 2.61682779e-01
-4.15455490e-01 -2.36559272e-01 -7.81838715e-01 5.17150939e-01
7.19267845e-01 -6.77427277e-02 1.02470197e-01 6.08967900e-01
1.22487068e-01 -2.72593647e-01 -1.11651003e+00 9.11155879e-01
-6.91975579e-02 -8.74997139e-01 7.30828568e-02 7.18273744e-02
1.04427302e+00 -3.99627775e-01 2.48641253e-01 3.88635211e-02
5.36432624e-01 -1.10447311e+00 2.58665651e-01 6.90638959e-01
1.10765076e+00 -4.35125232e-01 1.16592026e+00 3.13910335e-01
-8.02700460e-01 3.22652370e-01 -2.04194576e-01 7.66525686e-01
2.22245678e-01 1.07648790e+00 -9.44506168e-01 7.92476118e-01
2.77528375e-01 5.88777959e-01 -2.73473024e-01 1.13511646e+00
-4.44918334e-01 6.72966897e-01 -6.86183572e-02 3.54149044e-01
1.98430076e-01 -1.57257885e-01 2.38585323e-01 9.39379692e-01
3.99127215e-01 1.51500881e-01 3.01863194e-01 1.04567873e+00
-1.86767042e-01 -9.18004215e-02 -2.16708966e-02 -2.40935646e-02
2.40301400e-01 1.53205943e+00 -8.42451394e-01 -3.55779856e-01
-1.41898930e-01 9.08557296e-01 3.70688434e-03 2.83106297e-01
-1.09465444e+00 1.24131843e-01 1.30411342e-01 -2.03211188e-01
-2.08665296e-01 5.95890641e-01 -4.27934855e-01 -9.77123499e-01
-2.10224688e-01 -9.09751832e-01 5.57932377e-01 -6.29141271e-01
-1.45256698e+00 6.09437704e-01 -6.19208850e-02 -1.20167208e+00
-1.99243322e-01 -5.87983549e-01 -3.95040393e-01 8.01168561e-01
-1.67973542e+00 -1.22396541e+00 -4.50384617e-01 3.43627602e-01
2.44430363e-01 2.05638885e-01 5.67887485e-01 3.30985546e-01
-6.93410873e-01 8.19574416e-01 -6.72099665e-02 1.41700640e-01
1.04107547e+00 -1.50674903e+00 -3.23819131e-01 5.49638331e-01
-2.22373381e-01 2.92549253e-01 3.80676746e-01 -7.20615566e-01
-5.20797133e-01 -1.16761482e+00 5.43769419e-01 -2.08155483e-01
4.79714155e-01 3.32060933e-01 -1.01511073e+00 5.36779284e-01
-3.49097885e-02 3.53376478e-01 7.63228297e-01 -3.27067077e-01
3.83037962e-02 -2.70033449e-01 -1.53969419e+00 4.46852386e-01
5.81425130e-01 -4.11901444e-01 -4.50307399e-01 4.90539730e-01
8.06725085e-01 -9.62958217e-01 -1.15012658e+00 7.43383348e-01
4.46639299e-01 -7.04335153e-01 6.16336226e-01 -2.96617866e-01
5.54722071e-01 -3.48699957e-01 6.60605356e-02 -1.37860334e+00
1.19110674e-01 -2.53316343e-01 1.52174860e-01 1.17164600e+00
5.76185524e-01 -5.10648608e-01 7.82224357e-01 7.89958239e-01
-1.50853336e-01 -1.00476599e+00 -8.89717042e-01 -2.41293624e-01
8.32376853e-02 -1.73571691e-01 3.37031722e-01 1.02277744e+00
-3.77991259e-01 -2.93170094e-01 -2.14102849e-01 4.45324481e-01
6.72032118e-01 1.10132970e-01 1.60450488e-01 -1.18536651e+00
-4.20585722e-01 -3.17779899e-01 -1.58004150e-01 -8.60828459e-01
-1.72365271e-02 -9.28490639e-01 2.56244838e-01 -1.61889637e+00
4.70558256e-01 -6.41656637e-01 -5.93183935e-01 3.04464906e-01
-6.21267080e-01 1.19818687e-01 -1.94641784e-01 3.34377617e-01
-6.48086131e-01 1.69485718e-01 1.95319140e+00 -2.52660751e-01
-1.90812126e-02 4.12936062e-01 -7.82185853e-01 8.45866024e-01
6.46978080e-01 -6.23860657e-01 -3.61124426e-01 -2.67700732e-01
-6.59063607e-02 2.87582099e-01 2.90731698e-01 -7.28905618e-01
2.29131430e-01 -2.82202154e-01 9.94743779e-02 -4.54578876e-01
-1.48014510e-02 -7.85935700e-01 8.18209723e-02 6.36866212e-01
-5.10889649e-01 -6.32547975e-01 -1.79464549e-01 5.76853454e-01
-3.01748306e-01 -4.59703922e-01 1.12319505e+00 -3.13153476e-01
-2.72365838e-01 5.57657659e-01 -5.74856140e-02 1.25400633e-01
1.12968981e+00 -9.67461709e-03 -3.28911394e-01 -1.29669234e-01
-9.33474720e-01 3.74403983e-01 2.06540391e-01 -3.42201032e-02
3.81958097e-01 -1.13226283e+00 -8.61857295e-01 -2.21499085e-01
1.37068599e-01 4.57551122e-01 5.07718205e-01 1.25112629e+00
-5.38650393e-01 2.64300734e-01 2.35721454e-01 -1.05721295e+00
-1.03300154e+00 3.76057655e-01 8.18848073e-01 -7.20298171e-01
3.87995294e-03 9.70299304e-01 4.26350176e-01 -4.45300430e-01
1.14814162e-01 -7.05525041e-01 -2.43247405e-01 -5.24441265e-02
3.81931216e-01 8.74650627e-02 3.12017333e-02 -3.91732067e-01
-1.52174130e-01 8.50391209e-01 -4.34906900e-01 4.23436388e-02
6.78003073e-01 -2.34483853e-01 -1.38392463e-01 4.78345841e-01
8.45067918e-01 -2.64187843e-01 -1.13344574e+00 -3.67113769e-01
-1.94657132e-01 -2.60752559e-01 2.95715809e-01 -1.02811897e+00
-1.49573779e+00 9.31463540e-01 8.81953597e-01 1.70051560e-01
1.19394469e+00 -3.18040587e-02 7.53215849e-01 -6.38497919e-02
2.94042677e-01 -9.71397579e-01 -2.03829810e-01 -2.49249503e-01
5.03925264e-01 -1.68577111e+00 -2.58632917e-02 -7.07233667e-01
-8.85970891e-01 9.91125166e-01 6.11079276e-01 8.84226859e-02
6.51895642e-01 1.63144171e-01 3.63098055e-01 1.37733305e-02
-3.46705407e-01 -2.97724307e-01 5.78314304e-01 3.58980477e-01
4.20472652e-01 5.74057698e-01 -4.97441769e-01 8.78354967e-01
1.79260358e-01 3.24793965e-01 3.59646916e-01 3.65600079e-01
-3.27837169e-01 -1.14146650e+00 -3.38335693e-01 5.39647460e-01
-7.95300841e-01 1.70938641e-01 -1.74777299e-01 5.67714572e-01
4.86754626e-01 8.84771109e-01 -2.82226920e-01 -2.83040166e-01
-8.87278989e-02 -2.41770104e-01 4.01396960e-01 -6.36631250e-01
-4.90492612e-01 1.72078595e-01 -1.56654179e-01 -3.86404157e-01
-7.16979444e-01 -3.65036577e-01 -1.31857979e+00 1.00230463e-02
-6.12363815e-01 5.18559515e-02 5.18218040e-01 1.26747060e+00
-1.37287319e-01 8.21767211e-01 5.68891346e-01 -1.37316242e-01
-6.73124671e-01 -7.84856915e-01 -7.05948472e-01 3.77642781e-01
3.98808599e-01 -5.95841527e-01 -4.75627631e-01 1.61111563e-01]
|
[14.616496086120605, -2.086751699447632]
|
895c7470-dd9a-43dd-a410-99f2faaf6d8d
|
polycentric-clustering-and-structural
|
2210.07463
| null |
https://arxiv.org/abs/2210.07463v1
|
https://arxiv.org/pdf/2210.07463v1.pdf
|
Polycentric Clustering and Structural Regularization for Source-free Unsupervised Domain Adaptation
|
Source-Free Domain Adaptation (SFDA) aims to solve the domain adaptation problem by transferring the knowledge learned from a pre-trained source model to an unseen target domain. Most existing methods assign pseudo-labels to the target data by generating feature prototypes. However, due to the discrepancy in the data distribution between the source domain and the target domain and category imbalance in the target domain, there are severe class biases in the generated feature prototypes and noisy pseudo-labels. Besides, the data structure of the target domain is often ignored, which is crucial for clustering. In this paper, a novel framework named PCSR is proposed to tackle SFDA via a novel intra-class Polycentric Clustering and Structural Regularization strategy. Firstly, an inter-class balanced sampling strategy is proposed to generate representative feature prototypes for each class. Furthermore, k-means clustering is introduced to generate multiple clustering centers for each class in the target domain to obtain robust pseudo-labels. Finally, to enhance the model's generalization, structural regularization is introduced for the target domain. Extensive experiments on three UDA benchmark datasets show that our method performs better or similarly against the other state of the art methods, demonstrating our approach's superiority for visual domain adaptation problems.
|
['Huiyu Zhou', 'Ningzhong Liu', 'Han Sun', 'Xinyu Guan']
|
2022-10-14
| null | null | null | null |
['source-free-domain-adaptation']
|
['computer-vision']
|
[ 1.51873529e-01 -3.31248343e-01 -3.31797898e-01 -4.33738589e-01
-5.72370291e-01 -4.56144363e-01 4.20936882e-01 -7.79574141e-02
-1.87830970e-01 7.14207053e-01 8.06479007e-02 2.94647753e-01
-6.67944700e-02 -6.48367643e-01 -5.60993254e-01 -1.14705622e+00
6.45008266e-01 5.43613255e-01 2.11217105e-01 -6.16977271e-03
1.38455749e-01 3.44548464e-01 -1.49925280e+00 2.40025699e-01
1.20815825e+00 9.62554455e-01 4.52705860e-01 -1.71027318e-01
-4.16115344e-01 3.00368816e-01 -7.43468642e-01 1.11133792e-02
1.66709051e-01 -6.03443325e-01 -5.12392640e-01 7.27625787e-01
1.94662273e-01 4.66809096e-03 -4.76077721e-02 1.16763282e+00
4.01470900e-01 3.87800992e-01 1.13190663e+00 -1.28765213e+00
-9.06473637e-01 2.97819264e-02 -8.52331221e-01 -1.42210469e-01
-1.24813758e-01 -3.51576097e-02 6.05613947e-01 -1.05413556e+00
6.85645461e-01 1.20682967e+00 2.38607481e-01 8.23572516e-01
-1.45738542e+00 -9.28743660e-01 4.51421231e-01 2.16401011e-01
-1.48475051e+00 -2.40360633e-01 1.22097075e+00 -4.63620871e-01
1.72004759e-01 -1.74363941e-01 4.19263095e-01 1.31322742e+00
-4.06316787e-01 7.31548488e-01 9.83374357e-01 -4.71792459e-01
5.83020151e-01 6.96100295e-01 -5.29896794e-03 9.44976434e-02
3.19518931e-02 -3.67840379e-02 -1.67454705e-01 -1.94157809e-01
7.00646639e-01 5.87787442e-02 -1.54189736e-01 -1.19700789e+00
-1.05386257e+00 8.74564171e-01 5.44445038e-01 2.73971558e-01
-2.21349671e-01 -6.10682070e-01 5.03135741e-01 -6.21190853e-02
5.49439967e-01 1.15598239e-01 -3.95008117e-01 4.78899240e-01
-5.54634273e-01 1.28311161e-02 2.49547645e-01 1.17169595e+00
8.51757824e-01 1.07691146e-01 -1.99444413e-01 1.40371943e+00
3.38238239e-01 4.89049911e-01 8.31070423e-01 -7.15074003e-01
5.41728258e-01 9.51436281e-01 1.73014998e-01 -8.87484550e-01
-5.01347147e-02 -6.31447017e-01 -9.78690803e-01 1.53292611e-01
6.00580633e-01 8.42269976e-03 -9.94033575e-01 1.85353827e+00
7.32974887e-01 1.44713730e-01 2.80631185e-01 1.09797156e+00
6.71536446e-01 8.14679325e-01 3.12454164e-01 -3.65313470e-01
9.78750110e-01 -8.71747732e-01 -5.02485394e-01 -4.01747942e-01
4.63999331e-01 -6.33600652e-01 1.29604578e+00 2.79700398e-01
-4.56543744e-01 -8.80672991e-01 -9.32167351e-01 2.07492217e-01
-2.87386745e-01 5.76814532e-01 1.77111059e-01 3.76083821e-01
-2.82316387e-01 -7.92637840e-02 -2.95025915e-01 -3.17513645e-01
7.60415673e-01 3.81102040e-02 -4.24172580e-01 -4.85826910e-01
-9.41123188e-01 4.47717696e-01 7.85444915e-01 -2.49146461e-01
-8.24831009e-01 -5.69199920e-01 -7.84096360e-01 -1.34037748e-01
3.29801500e-01 -4.67300326e-01 8.23772788e-01 -1.52814114e+00
-1.32865202e+00 8.96769881e-01 -2.71118164e-01 8.96301940e-02
2.71440268e-01 2.66639620e-01 -5.03252387e-01 2.20302399e-02
2.15426162e-01 6.93973660e-01 1.23480809e+00 -1.71177840e+00
-7.32802570e-01 -6.38994455e-01 -4.60274488e-01 5.09397984e-01
-6.06144071e-01 -3.96995097e-01 -5.11156738e-01 -8.30620527e-01
2.82898396e-01 -7.50568509e-01 -1.38995245e-01 -9.61057544e-02
-1.59127891e-01 -4.39341754e-01 9.73989487e-01 -4.22718138e-01
1.08027506e+00 -2.66350579e+00 3.69046122e-01 4.15640652e-01
8.39960948e-02 4.77091670e-01 -2.63422698e-01 -8.29150602e-02
-3.02299023e-01 -3.69322866e-01 -3.97157848e-01 -8.96562114e-02
-1.43103868e-01 2.55334586e-01 -3.11736554e-01 3.77362072e-01
1.12818949e-01 4.23805505e-01 -8.78364444e-01 -5.86423337e-01
3.03813785e-01 4.51833874e-01 -3.50299895e-01 3.86455804e-01
-2.39753529e-01 7.37620234e-01 -5.64013660e-01 5.32107234e-01
8.99359643e-01 -2.07501218e-01 1.60715684e-01 -8.15551206e-02
3.16021621e-01 -1.73022091e-01 -1.30542755e+00 1.60579228e+00
-3.90946537e-01 1.39331415e-01 -1.40242362e-02 -1.32694304e+00
1.38193607e+00 8.60103518e-02 2.87858546e-01 -7.45083809e-01
1.72378451e-01 2.44671419e-01 -1.09940678e-01 -3.62727851e-01
8.84796381e-02 -2.95773685e-01 -4.06840630e-03 2.07119033e-01
3.27734537e-02 6.55575236e-03 7.24907890e-02 8.84812698e-03
3.23203743e-01 2.59546459e-01 2.97078431e-01 -1.12042338e-01
7.95733988e-01 2.37037659e-01 1.06705213e+00 2.06362545e-01
-3.17382455e-01 7.80599117e-01 2.10792825e-01 -3.05492431e-01
-1.16682279e+00 -1.19885504e+00 -2.56474137e-01 1.14728570e+00
3.22143674e-01 1.97244272e-01 -7.56372273e-01 -1.00226355e+00
-4.71051559e-02 8.18880379e-01 -6.52033865e-01 -5.19291222e-01
-2.92157680e-01 -7.64438033e-01 5.01385964e-02 6.89584374e-01
5.01608372e-01 -9.96669531e-01 -4.27548885e-02 2.90683597e-01
-3.21101815e-01 -8.26828420e-01 -5.20661414e-01 -5.31128272e-02
-9.78277385e-01 -1.04559445e+00 -1.10940218e+00 -1.29286182e+00
1.20551324e+00 4.20927048e-01 7.08378375e-01 -4.06463802e-01
-6.14171289e-02 8.84040445e-02 -5.31443834e-01 -2.68924385e-01
-4.20833856e-01 -5.72589748e-02 1.09297164e-01 4.19602305e-01
8.73381138e-01 -3.37779880e-01 -4.53778774e-01 5.98952472e-01
-9.00347948e-01 -1.31597012e-01 5.85162938e-01 1.01772594e+00
7.74637520e-01 3.00665617e-01 8.80802512e-01 -9.23830509e-01
4.78190571e-01 -6.73534930e-01 -5.02954900e-01 2.76278526e-01
-5.59786677e-01 -1.86305329e-01 9.22815859e-01 -9.67976391e-01
-1.48510671e+00 3.73398453e-01 3.96807075e-01 -6.45988345e-01
-6.56968355e-01 2.21100166e-01 -6.95092082e-01 2.90677965e-01
8.87434125e-01 4.75376636e-01 1.87010065e-01 -5.16872704e-01
2.85013556e-01 8.83849800e-01 5.01448929e-01 -6.79102361e-01
8.40868294e-01 5.26970923e-01 -3.32826316e-01 -6.72316849e-01
-7.68649638e-01 -6.48351192e-01 -8.75067115e-01 7.94831440e-02
5.94218254e-01 -1.10250568e+00 1.26821294e-01 5.99713147e-01
-8.25548947e-01 -1.31579638e-01 -4.49721187e-01 5.96306920e-01
-3.64660352e-01 3.89776945e-01 -3.51143838e-03 -6.38942361e-01
-1.75062958e-02 -1.04115844e+00 7.69324362e-01 5.15840948e-01
1.55392773e-02 -1.04358160e+00 -1.46663366e-02 4.06189293e-01
1.33217320e-01 5.36295995e-02 1.07638538e+00 -9.13839221e-01
-1.02848843e-01 -1.03550501e-01 -5.17015159e-01 7.57047594e-01
4.55963045e-01 -1.71546087e-01 -8.74936342e-01 -4.34647501e-01
1.93852335e-02 -3.95051479e-01 5.79597175e-01 3.71993661e-01
1.21273005e+00 -5.56052811e-02 -5.49904585e-01 4.13679570e-01
1.29272723e+00 4.43987906e-01 3.86167675e-01 3.02021921e-01
6.92156971e-01 8.10766280e-01 1.03769183e+00 5.03645778e-01
2.75043368e-01 6.17084980e-01 1.21755563e-01 -1.52548969e-01
-1.63456470e-01 -3.63480449e-01 1.91331133e-01 5.23992479e-01
3.66849273e-01 2.57503949e-02 -8.12309802e-01 8.80562961e-01
-1.68662727e+00 -5.69901168e-01 2.29269010e-03 2.39861560e+00
8.53479922e-01 -3.20541598e-02 2.68886715e-01 -9.22215544e-03
1.16273046e+00 -2.49221325e-01 -9.37323928e-01 3.11303977e-02
-2.45016560e-01 -1.55596420e-01 5.35802394e-02 1.75737627e-02
-1.20028996e+00 8.22160065e-01 4.89168167e+00 1.22989583e+00
-1.10143554e+00 -1.26073658e-02 4.89491075e-01 1.02094628e-01
-2.32364833e-02 -1.81937158e-01 -8.00340414e-01 6.57091737e-01
3.55744720e-01 -2.68666029e-01 2.94791728e-01 1.20968544e+00
4.31935266e-02 3.47288661e-02 -8.21556926e-01 1.07239151e+00
1.77558213e-01 -6.48900449e-01 1.74380288e-01 7.61027485e-02
9.12229955e-01 -4.02410090e-01 1.13880195e-01 4.13038284e-01
2.56908327e-01 -5.22441804e-01 2.91138262e-01 1.57239377e-01
1.01348495e+00 -1.11279547e+00 5.56532860e-01 5.19467592e-01
-1.02998316e+00 -2.71898270e-01 -8.08722734e-01 2.46763557e-01
-2.33809277e-01 5.88439226e-01 -7.89215982e-01 4.90735143e-01
7.27843940e-01 7.96055377e-01 -6.77456260e-01 1.10209930e+00
-2.20410705e-01 4.34466571e-01 -1.43946201e-01 3.53738427e-01
1.20127417e-01 -4.38045144e-01 4.91576701e-01 8.36510658e-01
3.15564424e-01 -8.61756504e-02 2.34923080e-01 8.29183340e-01
7.95667467e-04 3.92840326e-01 -6.25357866e-01 2.60237783e-01
6.81218922e-01 1.10611153e+00 -5.21491230e-01 -4.43658620e-01
-3.75748813e-01 9.67646539e-01 3.51617783e-01 6.85101211e-01
-7.16060281e-01 -5.25102019e-01 4.97065514e-01 2.87022069e-02
4.18831378e-01 1.14004634e-01 -3.43713760e-01 -1.22466469e+00
6.92040473e-02 -8.95300627e-01 6.54003024e-01 -6.16341412e-01
-1.71571410e+00 3.52550715e-01 9.62210074e-03 -1.75109494e+00
-9.22942981e-02 -3.39126259e-01 -4.64474618e-01 9.21534777e-01
-1.39778686e+00 -1.13098001e+00 -4.13657099e-01 9.01140988e-01
6.27767920e-01 -5.19564867e-01 6.73085570e-01 3.75776410e-01
-6.18218541e-01 7.12844193e-01 7.12007761e-01 1.50513694e-01
1.16374278e+00 -9.95013058e-01 -1.48037240e-01 6.92072451e-01
-2.32008368e-01 5.93315661e-01 3.98355573e-01 -5.66546977e-01
-6.54235482e-01 -1.40982819e+00 5.37118852e-01 -1.31657958e-01
3.26594651e-01 -3.96342099e-01 -1.39357996e+00 6.11658752e-01
-1.59916058e-01 1.47622332e-01 8.53615999e-01 -7.09678531e-02
-5.63765943e-01 -2.85153955e-01 -1.35936713e+00 4.45672899e-01
7.07924724e-01 -2.06523001e-01 -8.04461658e-01 2.54411608e-01
3.88245225e-01 -2.90557981e-01 -6.50465846e-01 3.52872819e-01
2.29410186e-01 -6.80647731e-01 9.77038443e-01 -5.03982008e-01
2.27528378e-01 -7.05535889e-01 -6.83523044e-02 -1.74543941e+00
-5.55741906e-01 2.70163298e-01 9.20468718e-02 1.66507518e+00
9.52652395e-02 -5.73556721e-01 8.01454425e-01 3.91905814e-01
-2.07295194e-02 -2.79753119e-01 -8.52555275e-01 -9.62587178e-01
3.03996146e-01 1.13337468e-02 5.31769037e-01 1.22110224e+00
-3.02093148e-01 5.23318708e-01 -1.07220456e-01 1.56452507e-01
8.20377111e-01 3.03412795e-01 8.32642198e-01 -1.39644313e+00
6.14044890e-02 -2.36144766e-01 -1.22869872e-01 -9.06740129e-01
3.27418000e-01 -8.32903802e-01 -2.05351207e-02 -1.28017819e+00
2.83587396e-01 -7.83902049e-01 -3.89541686e-01 3.22483748e-01
-2.34262198e-01 1.96780086e-01 -4.99876514e-02 4.53773588e-01
-4.95119482e-01 8.70420754e-01 1.40727878e+00 -1.95297420e-01
-4.91153061e-01 1.16916813e-01 -7.85502791e-01 6.50021970e-01
8.08811069e-01 -5.44228137e-01 -6.61744893e-01 -2.51487195e-01
-4.77517515e-01 -3.53297859e-01 2.83046514e-01 -9.26148832e-01
-2.08076332e-02 -3.87363583e-01 7.38476753e-01 -4.82259750e-01
1.65117517e-01 -1.12253642e+00 -4.64280546e-02 1.36081129e-01
-1.42082557e-01 -7.54528105e-01 1.65546417e-01 7.76838541e-01
-3.76114488e-01 -2.08076552e-01 1.34816408e+00 2.11785343e-02
-9.44487393e-01 1.18878782e-01 -6.69037849e-02 2.89800882e-01
1.37036216e+00 -3.81750077e-01 -2.09308386e-01 1.25626931e-02
-7.65295029e-01 3.16595286e-01 7.55284250e-01 6.18627667e-01
6.43665612e-01 -1.62838161e+00 -6.68395519e-01 4.23836768e-01
7.02806354e-01 3.71246636e-01 5.13688147e-01 4.14317280e-01
1.51618747e-02 1.84226438e-01 -4.09000039e-01 -7.13575363e-01
-1.10961664e+00 9.69415963e-01 1.64247364e-01 9.94974822e-02
-4.68483925e-01 6.43276095e-01 7.96554267e-01 -7.52076328e-01
2.20730901e-01 2.30901808e-01 -3.79362106e-01 2.72031784e-01
5.20755708e-01 3.54119807e-01 -2.13678375e-01 -8.62095237e-01
-2.87022144e-01 6.98672652e-01 -2.61274606e-01 3.40808690e-01
1.11373615e+00 -4.61807966e-01 1.32385075e-01 3.77767771e-01
1.17152190e+00 -2.50668913e-01 -1.55787492e+00 -7.03849375e-01
-1.66417018e-01 -5.02676785e-01 -2.43005544e-01 -7.92884946e-01
-1.02964759e+00 9.76598322e-01 7.74053276e-01 -3.49146008e-01
1.34842730e+00 5.25227711e-02 5.58388770e-01 2.49904171e-02
2.24036410e-01 -1.46105087e+00 2.20040917e-01 2.03304216e-01
7.85483539e-01 -1.34965992e+00 -2.25929737e-01 -4.16344076e-01
-1.09799242e+00 8.14393580e-01 1.03856158e+00 -1.03135020e-01
3.12484235e-01 -4.62581605e-01 2.56024092e-01 2.50618190e-01
-2.74873793e-01 -7.70349950e-02 3.27510536e-01 1.04877770e+00
-3.91068533e-02 8.54723807e-03 -1.01608135e-01 7.81380594e-01
3.24916661e-01 -1.76551212e-02 2.41341472e-01 6.65062189e-01
-3.86046499e-01 -1.22049904e+00 -7.28501558e-01 2.61975914e-01
-1.69532206e-02 2.95783371e-01 -3.28776330e-01 7.64907897e-01
3.00695688e-01 8.66967797e-01 1.46807376e-02 -2.21893013e-01
4.49327290e-01 3.24256569e-01 2.14392960e-01 -8.10629785e-01
-8.68476927e-02 3.79624784e-01 -5.30891240e-01 -5.78706339e-02
-3.40170354e-01 -5.79296947e-01 -1.30613041e+00 2.27701634e-01
-2.77254760e-01 3.20085168e-01 4.79048967e-01 6.89539552e-01
4.80492890e-01 3.39349091e-01 9.07709718e-01 -5.99815190e-01
-4.91889119e-01 -9.15656090e-01 -8.63762379e-01 8.47112715e-01
7.78821856e-02 -1.04519629e+00 -5.01740217e-01 2.31904402e-01]
|
[10.348424911499023, 3.082895278930664]
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.