load geotifs from AWS S3

.gitignore

@@ -1,4 +1,6 @@
 hf_cache
 __pycache__
 .DS_Store
-data/
+data/*
+!data/dataset_info.json
+!data/chips_stats.yaml

data/chips_stats.yaml

@@ -0,0 +1,99 @@
+fold_0:
+  mean:
+  - 570.7305297851562
+  - 691.6322021484375
+  - 436.3498229980469
+  - 3222.44775390625
+  - 1939.10009765625
+  - 1180.9752197265625
+  n_chips: 72
+  std:
+  - 466.30303955078125
+  - 355.8695373535156
+  - 305.64422607421875
+  - 1125.038330078125
+  - 681.1266479492188
+  - 632.9395751953125
+fold_1:
+  mean:
+  - 578.840087890625
+  - 700.3932495117188
+  - 447.7603759765625
+  - 3214.770751953125
+  - 1935.7547607421875
+  - 1180.948974609375
+  n_chips: 72
+  std:
+  - 488.18951416015625
+  - 380.0794677734375
+  - 339.2040100097656
+  - 1149.1806640625
+  - 684.062255859375
+  - 638.1996459960938
+fold_2:
+  mean:
+  - 576.6995239257812
+  - 696.1321411132812
+  - 441.8569030761719
+  - 3244.72998046875
+  - 1951.02734375
+  - 1187.2139892578125
+  n_chips: 72
+  std:
+  - 479.5519104003906
+  - 371.16168212890625
+  - 329.15521240234375
+  - 1062.248779296875
+  - 642.8277587890625
+  - 624.2898559570312
+fold_3:
+  mean:
+  - 556.559814453125
+  - 678.816162109375
+  - 427.0155944824219
+  - 3229.951904296875
+  - 1929.81103515625
+  - 1163.3214111328125
+  n_chips: 71
+  std:
+  - 461.0555725097656
+  - 352.8609619140625
+  - 301.42449951171875
+  - 1098.100830078125
+  - 670.2280883789062
+  - 628.5005493164062
+fold_4:
+  mean:
+  - 565.2382202148438
+  - 683.7120361328125
+  - 431.3344421386719
+  - 3241.815185546875
+  - 1936.545654296875
+  - 1176.8934326171875
+  n_chips: 71
+  std:
+  - 470.9222412109375
+  - 360.7673645019531
+  - 309.9328918457031
+  - 1090.716796875
+  - 671.9968872070312
+  - 639.7266845703125
+fold_5:
+  mean:
+  - 568.527587890625
+  - 691.322021484375
+  - 444.148193359375
+  - 3116.252197265625
+  - 1885.5509033203125
+  - 1155.75927734375
+  n_chips: 71
+  std:
+  - 473.07781982421875
+  - 364.54290771484375
+  - 317.82159423828125
+  - 1228.0555419921875
+  - 732.2979736328125
+  - 648.6840209960938
+num_classes_tier1: 6
+num_classes_tier2: 17
+num_classes_tier3: 49

data/dataset_info.json

@@ -0,0 +1,182 @@
+{
+    "tier1": [
+        "Background",
+        "Grassland",
+        "Field crops",
+        "Orchards",
+        "Special crops",
+        "Forest"
+    ],
+    "tier2": [
+        "Background",
+        "Meadow",
+        "SmallGrainCereal",
+        "LargeGrainCereal",
+        "Pasture",
+        "BroadLeafRowCrop",
+        "VegetableCrop",
+        "OrchardCrop",
+        "Fallow",
+        "Hedge",
+        "Berries",
+        "TreeCrop",
+        "CropMix",
+        "BiodiversityArea",
+        "Forest",
+        "Multiple",
+        "Gardens"
+    ],
+    "tier3": [
+        "Background",
+        "Meadow",
+        "WinterWheat",
+        "Maize",
+        "Pasture",
+        "Sugar_beets",
+        "WinterBarley",
+        "WinterRapeseed",
+        "Vegetables",
+        "Potatoes",
+        "Wheat",
+        "Sunflowers",
+        "Vines",
+        "Spelt",
+        "Fallow",
+        "Hedge",
+        "Apples",
+        "Soy",
+        "Peas",
+        "Oat",
+        "Berries",
+        "EinkornWheat",
+        "Field bean",
+        "SummerWheat",
+        "Rye",
+        "TreeCrop",
+        "StoneFruit",
+        "MixedCrop",
+        "Sorghum",
+        "Grain",
+        "Chicory",
+        "Pears",
+        "SummerBarley",
+        "Biodiversity encouragement area",
+        "Forest",
+        "Linen",
+        "Legumes",
+        "Pumpkin",
+        "Tobacco",
+        "Buckwheat",
+        "Hemp",
+        "SummerRapeseed",
+        "Hops",
+        "Beets",
+        "Multiple",
+        "Lupine",
+        "Mustard",
+        "Gardens",
+        "Chestnut"
+    ],
+    "tier3_to_tier1": [
+        0,
+        1,
+        2,
+        2,
+        1,
+        2,
+        2,
+        2,
+        2,
+        2,
+        2,
+        2,
+        3,
+        2,
+        4,
+        4,
+        3,
+        2,
+        2,
+        2,
+        4,
+        2,
+        2,
+        2,
+        2,
+        3,
+        3,
+        2,
+        2,
+        2,
+        2,
+        3,
+        2,
+        1,
+        5,
+        2,
+        2,
+        2,
+        2,
+        2,
+        2,
+        2,
+        3,
+        2,
+        4,
+        2,
+        2,
+        4,
+        3
+    ],
+    "tier3_to_tier2": [
+        0,
+        1,
+        2,
+        3,
+        4,
+        5,
+        2,
+        5,
+        6,
+        5,
+        2,
+        5,
+        7,
+        2,
+        8,
+        9,
+        7,
+        5,
+        5,
+        2,
+        10,
+        2,
+        5,
+        2,
+        2,
+        11,
+        7,
+        12,
+        3,
+        2,
+        6,
+        7,
+        2,
+        13,
+        14,
+        5,
+        5,
+        6,
+        5,
+        2,
+        5,
+        5,
+        7,
+        5,
+        15,
+        5,
+        5,
+        16,
+        7
+    ]
+}

inference.py

@@ -14,6 +14,9 @@ import geopandas as gpd
 
 from messis.messis import LogConfusionMatrix
 
+import rasterio
+import rasterio.env
+
 class InferenceDataLoader:
     def __init__(self, features_path, labels_path, field_ids_path, stats_path, window_size=224, n_timesteps=3, fold_indices=None, debug=False):
         self.features_path = features_path
@@ -24,7 +27,7 @@ class InferenceDataLoader:
         self.n_timesteps = n_timesteps
         self.fold_indices = fold_indices if fold_indices is not None else []
         self.debug = debug
-        
+
         # Load normalization stats
         self.means, self.stds = self.load_stats()
 
@@ -69,20 +72,20 @@ class InferenceDataLoader:
             if self.debug:
                 print("Source Transform", src.transform)
                 print(f"UTM X: {utm_x}, UTM Y: {utm_y}")
-            
+
             try:
                 px, py = rowcol(src.transform, utm_x, utm_y)
             except ValueError:
                 raise ValueError("Coordinates out of bounds for this raster.")
-            
+
             if self.debug:
                 print(f"Row: {py}, Column: {px}")
-            
+
             half_window_size = self.window_size // 2
-            
+
             row_off = px - half_window_size
             col_off = py - half_window_size
-            
+
             if row_off < 0:
                 row_off = 0
             if col_off < 0:
@@ -91,7 +94,7 @@ class InferenceDataLoader:
                 row_off = src.width - self.window_size
             if col_off + self.window_size > src.height:
                 col_off = src.height - self.window_size
-            
+
             window = Window(col_off, row_off, self.window_size, self.window_size)
             window_transform = src.window_transform(window)
             if self.debug:
@@ -109,7 +112,7 @@ class InferenceDataLoader:
         if self.debug:
             print(f"Extracted window data from {path}")
             print(f"Min: {window_data.min()}, Max: {window_data.max()}")
-        
+
         return window_data
 
     def prepare_data_for_model(self, features_data):
@@ -199,10 +202,12 @@ def crop_predictions_to_gdf(field_ids, targets, predictions, transform, crs, cla
     return gdf
 
 def perform_inference(lon, lat, model, config, debug=False):
-    features_path = "./data/stacked_features.tif"
-    labels_path = "./data/labels.tif"
-    field_ids_path = "./data/field_ids.tif"
+    features_path = "s3://messis-demo/stacked_features_cog.tif"
+    labels_path = "s3://messis-demo/labels_cog.tif"
+    field_ids_path = "s3://messis-demo/field_ids_cog.tif"
+
     stats_path = "./data/chips_stats.yaml"
+    dataset_info_path = "./data/dataset_info.json"
 
     loader = InferenceDataLoader(features_path, labels_path, field_ids_path, stats_path, n_timesteps=9, fold_indices=[0], debug=True)
 
@@ -215,8 +220,9 @@ def perform_inference(lon, lat, model, config, debug=False):
         print(label_data.shape)
         print(field_ids_data.shape)
 
-    with open('./data/dataset_info.json', 'r') as file:
+    with open(dataset_info_path, 'r') as file:
         dataset_info = json.load(file)
+    
     class_names = dataset_info['tier3']
 
     tiers_dict = {k: v for k, v in config.hparams.get('heads_spec').items() if v.get('is_metrics_tier', False)}
@@ -236,6 +242,5 @@ def perform_inference(lon, lat, model, config, debug=False):
     # Simple GeoDataFrame with only the necessary columns
     gdf = gdf[['prediction_class', 'target_class', 'correct', 'geometry']]
     gdf.columns = ['Prediction', 'Target', 'Correct', 'geometry']
-    # gdf = gdf[gdf['Target'] != 'Background']
 
     return gdf

pages/2_Perform_Crop_Classification.py

@@ -91,7 +91,7 @@ def perform_inference_step():
     # Add COG
     m.add_cog_layer(
         url="https://messis-demo.s3.amazonaws.com/stacked_features_cog.tif",
-        name="AWS COG",
+        name="Sentinel-2 Satellite Imagery",
         bands=selected_bands,
         rescale=f"{vmin_vmax[selected_band][0]},{vmin_vmax[selected_band][1]}",
         zoom_to_layer=True