Dataset Viewer (First 5GB)
Auto-converted to Parquet Duplicate
id
int64
0
29.4k
image
imagewidth (px)
259
6k
image_id
stringlengths
16
16
question_id
stringlengths
1
5
question
stringlengths
10
113
answers
listlengths
10
10
answer
stringlengths
1
411
image_emb
list
question_emb
list
ocr_tokens
listlengths
0
100
image_classes
listlengths
1
61
set_name
stringclasses
1 value
0
0054c91397f2fe05
0
what is the brand of phone?
[ "nokia", "nokia", "nokia", "nokia", "toshiba", "nokia", "nokia", "nokia", "nokia", "nokia" ]
nokia
[ 0.0195159912109375, -0.050384521484375, -0.0007681846618652344, -0.01210784912109375, 0.032470703125, -0.035003662109375, -0.01739501953125, 0.03240966796875, -0.003498077392578125, -0.045501708984375, 0.009185791015625, -0.01641845703125, -0.008514404296875, 0.069580078125, 0.0196533203...
[ -0.02545166015625, -0.039031982421875, -0.05108642578125, -0.000946044921875, -0.0210723876953125, -0.0010499954223632812, 0.00933837890625, -0.0609130859375, -0.0169677734375, -0.0294036865234375, -0.05181884765625, -0.05889892578125, -0.0018205642700195312, 0.00894927978515625, -0.0075...
[ "MIA", "NOKIA" ]
[ "Belt", "Headphones", "Goggles", "Scale", "Bottle opener", "Mobile phone", "Mirror", "Digital clock", "Television", "Telephone", "Tool", "Wheel", "Camera", "Watch", "Glasses", "Aircraft" ]
train
1
005635e119b9f32f
1
what type of plane is this?
[ "lape", "cargo", "ec-agg", "lape", "lape", "lape", "lape", "lape", "lape", "airplane" ]
lape
[ -0.0169219970703125, -0.07025146484375, -0.10357666015625, 0.07586669921875, -0.01450347900390625, 0.044677734375, 0.0242919921875, -0.01123046875, 0.0048065185546875, -0.0121612548828125, -0.00865936279296875, 0.00982666015625, 0.031829833984375, -0.005779266357421875, 0.020462036132812...
[ 0.013214111328125, -0.038787841796875, -0.03497314453125, -0.0303192138671875, -0.048919677734375, 0.027984619140625, -0.04022216796875, -0.0301666259765625, -0.0021152496337890625, 0.01177215576171875, 0.0014858245849609375, -0.01209259033203125, 0.039306640625, 0.0199432373046875, -0.0...
[ "GOELANI", "LAPE", "EC-AGG", "EC", "AGG" ]
[ "Vehicle", "Helicopter", "Airplane", "Bomb", "Aircraft" ]
train
2
005635e119b9f32f
2
what are the letters on the tail section of the plane?
[ "ec agg", "ec-agg", "ec", "ec-agg", "ec", "ec", "ec", "ec", "ec goeland", "ec" ]
ec agg
[ -0.0169219970703125, -0.07025146484375, -0.10357666015625, 0.07586669921875, -0.01450347900390625, 0.044677734375, 0.0242919921875, -0.01123046875, 0.0048065185546875, -0.0121612548828125, -0.00865936279296875, 0.00982666015625, 0.031829833984375, -0.005779266357421875, 0.020462036132812...
[ -0.05810546875, -0.04046630859375, 0.0005159378051757812, -0.00008803606033325195, -0.038330078125, -0.005764007568359375, -0.03216552734375, -0.03289794921875, 0.008392333984375, -0.0161590576171875, 0.01226043701171875, 0.00423431396484375, 0.042144775390625, 0.0304412841796875, 0.0178...
[ "GOELANI", "LAPE", "EC-AGG", "EC", "AGG" ]
[ "Vehicle", "Helicopter", "Airplane", "Bomb", "Aircraft" ]
train
3
00685bc495504d61
3
who is this copyrighted by?
[ "simon clancy", "simon ciancy", "simon clancy", "simon clancy", "the brand is bayard", "simon clancy", "simon clancy", "simon clancy", "simon clancy", "simon clancy" ]
simon clancy
[ -0.042266845703125, -0.0989990234375, -0.0233612060546875, -0.0172271728515625, -0.0304718017578125, -0.037994384765625, -0.029144287109375, -0.01496124267578125, 0.0239715576171875, -0.0250244140625, 0.04290771484375, -0.01398468017578125, 0.01055908203125, 0.0153045654296875, -0.000226...
[ 0.01406097412109375, -0.03057861328125, -0.035614013671875, -0.00931549072265625, -0.020538330078125, -0.024566650390625, -0.00481414794921875, -0.0017261505126953125, -0.01702880859375, 0.0225067138671875, 0.030517578125, -0.005001068115234375, 0.0092620849609375, -0.05206298828125, 0.0...
[ "Vingin", "SONOOOOL", "VH-VXU", "Copyright", "Simon", "Clancy", "RWY34.COM" ]
[ "Vehicle", "Tower", "Airplane", "Aircraft" ]
train
4
00685bc495504d61
4
what brand is on the plane?
[ "virgin is the brand on the plane.", "virgin mobile", "virgin", "virgin", "virgin", "virgin", "virgin", "virgin", "virgin", "virgin" ]
virgin is the brand on the plane.
[ -0.042266845703125, -0.0989990234375, -0.0233612060546875, -0.0172271728515625, -0.0304718017578125, -0.037994384765625, -0.029144287109375, -0.01496124267578125, 0.0239715576171875, -0.0250244140625, 0.04290771484375, -0.01398468017578125, 0.01055908203125, 0.0153045654296875, -0.000226...
[ -0.04248046875, -0.0216064453125, 0.01262664794921875, -0.0247344970703125, -0.007167816162109375, 0.008148193359375, -0.01470184326171875, -0.0300140380859375, -0.0007414817810058594, -0.05413818359375, -0.034271240234375, -0.008026123046875, 0.038665771484375, 0.00653839111328125, -0.0...
[ "Vingin", "SONOOOOL", "VH-VXU", "Copyright", "Simon", "Clancy", "RWY34.COM" ]
[ "Vehicle", "Tower", "Airplane", "Aircraft" ]
train
5
006d10667d17b924
5
what year is shown in the photo?
[ "2011", "2011", "2011", "2011", "2011", "2011", "2011", "2011", "2011", "2011" ]
2011
[ 0.0134429931640625, -0.08416748046875, -0.1583251953125, 0.00963592529296875, -0.054443359375, 0.01212310791015625, 0.0215911865234375, -0.02734375, 0.0020923614501953125, 0.0119781494140625, -0.0190277099609375, 0.01258087158203125, 0.098388671875, -0.06298828125, -0.023529052734375, ...
[ -0.0079345703125, 0.0087738037109375, -0.043914794921875, -0.00974273681640625, -0.04254150390625, 0.0005168914794921875, 0.01506805419921875, -0.050079345703125, -0.0205230712890625, -0.0197906494140625, 0.0018157958984375, -0.0123138427734375, -0.0111541748046875, -0.00010645389556884766...
[ "meeny", "rimini", "DIVENTA", "UNA", "IM", "H.G.", "G.ARMIAH", "VGISTEIZA", "21.27", "AGOSTO", "2011", "meeting" ]
[ "Person", "Woman", "Man", "Tree", "Clothing", "Airplane", "Human face", "Aircraft" ]
train
6
006d10667d17b924
6
what type of meeting is it?
[ "rimini", "royal theater", "rimini", "rimini", "rimini meeting", "rimini", "rimini meeting", "rimini", "rimini", "rimini" ]
rimini
[ 0.0134429931640625, -0.08416748046875, -0.1583251953125, 0.00963592529296875, -0.054443359375, 0.01212310791015625, 0.0215911865234375, -0.02734375, 0.0020923614501953125, 0.0119781494140625, -0.0190277099609375, 0.01258087158203125, 0.098388671875, -0.06298828125, -0.023529052734375, ...
[ -0.019622802734375, 0.003787994384765625, -0.005146026611328125, 0.035400390625, -0.04034423828125, -0.0240020751953125, 0.02783203125, -0.01367950439453125, 0.0030117034912109375, -0.04425048828125, 0.00428009033203125, -0.00679779052734375, 0.0268096923828125, 0.001270294189453125, 0.0...
[ "meeny", "rimini", "DIVENTA", "UNA", "IM", "H.G.", "G.ARMIAH", "VGISTEIZA", "21.27", "AGOSTO", "2011", "meeting" ]
[ "Person", "Woman", "Man", "Tree", "Clothing", "Airplane", "Human face", "Aircraft" ]
train
7
00c359f294f7dcd9
7
what is the name of this plane?
[ "g-atco", "g-atco", "g-atco", "g-atco", "g-atco", "g atco", "g-atco", "g-atco", "g-atco", "g-atco" ]
g-atco
[ -0.036468505859375, 0.0255889892578125, -0.0357666015625, 0.0010290145874023438, -0.054779052734375, -0.0153961181640625, -0.00824737548828125, 0.007740020751953125, -0.005214691162109375, 0.04840087890625, -0.0157318115234375, -0.0019779205322265625, 0.0292205810546875, 0.0087814331054687...
[ -0.008819580078125, -0.049896240234375, -0.034912109375, -0.0364990234375, -0.04132080078125, 0.036346435546875, -0.046539306640625, -0.033782958984375, 0.00879669189453125, 0.004383087158203125, 0.01171875, -0.004207611083984375, 0.0419921875, 0.016265869140625, -0.004245758056640625, ...
[ "G-ATCO" ]
[ "Vehicle", "Helicopter", "Airplane", "Aircraft" ]
train
8
00c359f294f7dcd9
8
what is the plane's call sign?
[ "g-atco", "g-atco", "g-atco", "g-atco", "g-atco", "g-atco", "g-atco", "g-atco", "g-atco", "g-atco" ]
g-atco
[ -0.036468505859375, 0.0255889892578125, -0.0357666015625, 0.0010290145874023438, -0.054779052734375, -0.0153961181640625, -0.00824737548828125, 0.007740020751953125, -0.005214691162109375, 0.04840087890625, -0.0157318115234375, -0.0019779205322265625, 0.0292205810546875, 0.0087814331054687...
[ -0.036529541015625, -0.0401611328125, -0.0169525146484375, 0.00612640380859375, -0.04241943359375, 0.01300048828125, -0.042327880859375, -0.054534912109375, 0.0156097412109375, -0.0109710693359375, -0.0072784423828125, 0.00799560546875, 0.031494140625, 0.0213165283203125, 0.0056266784667...
[ "G-ATCO" ]
[ "Vehicle", "Helicopter", "Airplane", "Aircraft" ]
train
9
0122c5279a501df2
9
what letter is on the plane's tail?
[ "f", "f", "f", "f", "f", "f", "f", "f", "f", "f" ]
f
[ -0.03997802734375, -0.253173828125, -0.08935546875, -0.0006532669067382812, -0.042022705078125, -0.0249176025390625, -0.0841064453125, 0.040435791015625, -0.0301513671875, 0.039459228515625, -0.0173797607421875, 0.0247802734375, 0.01088714599609375, 0.0238189697265625, -0.023208618164062...
[ -0.033111572265625, -0.0228729248046875, 0.002918243408203125, -0.0028095245361328125, -0.058837890625, 0.0081939697265625, -0.014190673828125, -0.033050537109375, 0.0184326171875, -0.037017822265625, 0.0085296630859375, 0.0087890625, 0.056610107421875, 0.00811767578125, 0.00487518310546...
[]
[ "Vehicle", "Airplane", "Aircraft" ]
train
End of preview. Expand in Data Studio

TextVQA (Lance Format)

Lance-formatted version of TextVQA — VQA where the question requires reading text in the image — sourced from lmms-lab/textvqa.

Each row carries the image bytes, the question, the 10 reference answers, the OCR tokens detected by the dataset's pre-processing, and CLIP image + question embeddings.

Splits

Split Rows
validation.lance 5,000
train.lance 34,602

Schema

Column Type Notes
id int64 Row index within split
image large_binary Inline JPEG bytes
image_id string? TextVQA image id
question_id string? TextVQA question id
question string The question text
answers list<string> 10 annotator answers
answer string First answer — used as canonical / FTS target
ocr_tokens list<string> OCR tokens detected on the image
image_classes list<string> OpenImages-style scene tags from the source
set_name string? Source partition (train, val)
image_emb fixed_size_list<float32, 512> OpenCLIP image embedding (cosine-normalized)
question_emb fixed_size_list<float32, 512> OpenCLIP text embedding of the question

Pre-built indices

  • IVF_PQ on image_emb and question_embmetric=cosine
  • INVERTED (FTS) on question and answer
  • BTREE on image_id, question_id, set_name

Quick start

import lance
ds = lance.dataset("hf://datasets/lance-format/textvqa-lance/data/validation.lance")
print(ds.count_rows(), ds.schema.names, ds.list_indices())

Load with LanceDB

These tables can also be consumed by LanceDB, the multimodal lakehouse and embedded search library built on top of Lance, for simplified vector search and other queries.

import lancedb

db = lancedb.connect("hf://datasets/lance-format/textvqa-lance/data")
tbl = db.open_table("validation")
print(f"LanceDB table opened with {len(tbl)} image-question pairs")

Cross-modal text→image search

import lance, pyarrow as pa, open_clip, torch

model, _, _ = open_clip.create_model_and_transforms("ViT-B-32", pretrained="laion2b_s34b_b79k")
tokenizer = open_clip.get_tokenizer("ViT-B-32")
model = model.eval().cuda().half()
with torch.no_grad():
    q = model.encode_text(tokenizer(["what brand is on this billboard?"]).cuda())
    q = (q / q.norm(dim=-1, keepdim=True)).float().cpu().numpy()[0]

ds = lance.dataset("hf://datasets/lance-format/textvqa-lance/data/validation.lance")
emb_field = ds.schema.field("image_emb")
hits = ds.scanner(
    nearest={"column": "image_emb", "q": pa.array([q.tolist()], type=emb_field.type)[0], "k": 10},
    columns=["question", "answer", "ocr_tokens"],
).to_table().to_pylist()

LanceDB cross-modal text→image search

import lancedb, open_clip, torch

model, _, _ = open_clip.create_model_and_transforms("ViT-B-32", pretrained="laion2b_s34b_b79k")
tokenizer = open_clip.get_tokenizer("ViT-B-32")
model = model.eval().cuda().half()
with torch.no_grad():
    q = model.encode_text(tokenizer(["what brand is on this billboard?"]).cuda())
    q = (q / q.norm(dim=-1, keepdim=True)).float().cpu().numpy()[0]

db = lancedb.connect("hf://datasets/lance-format/textvqa-lance/data")
tbl = db.open_table("validation")

results = (
    tbl.search(q.tolist(), vector_column_name="image_emb")
    .metric("cosine")
    .select(["question", "answer", "ocr_tokens"])
    .limit(10)
    .to_list()
)

LanceDB full-text search

import lancedb

db = lancedb.connect("hf://datasets/lance-format/textvqa-lance/data")
tbl = db.open_table("validation")

results = (
    tbl.search("brand name")
    .select(["question", "answer"])
    .limit(10)
    .to_list()
)

Why Lance?

  • One dataset for images + questions + answers + OCR + dual embeddings + indices — no JSON/feature folders.
  • Cross-modal search and OCR-text filtering work on the same dataset on the Hub.
  • Schema evolution: add columns (alternate OCR systems, model predictions) without rewriting the data.

Source & license

Converted from lmms-lab/textvqa. TextVQA is released under CC BY 4.0 by Singh et al. (Facebook AI Research).

Citation

@inproceedings{singh2019towards,
  title={Towards VQA models that can read},
  author={Singh, Amanpreet and Natarajan, Vivek and Shah, Meet and Jiang, Yu and Chen, Xinlei and Batra, Dhruv and Parikh, Devi and Rohrbach, Marcus},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2019}
}
Downloads last month
76