fix

README.md

@@ -45,10 +45,10 @@ dataset_info:
         dtype: string
   splits:
   - name: train
-    num_bytes: 34542
+    num_bytes: 18890
     num_examples: 52
   download_size: 313334253
-  dataset_size: 34542
+  dataset_size: 18890
 - config_name: video_02
   features:
   - name: id
@@ -85,10 +85,10 @@ dataset_info:
         dtype: string
   splits:
   - name: train
-    num_bytes: 38528
+    num_bytes: 21070
     num_examples: 58
   download_size: 67368145
-  dataset_size: 38528
+  dataset_size: 21070
 - config_name: video_03
   features:
   - name: id
@@ -125,10 +125,10 @@ dataset_info:
         dtype: string
   splits:
   - name: train
-    num_bytes: 32550
+    num_bytes: 17801
     num_examples: 49
   download_size: 148418007
-  dataset_size: 32550
+  dataset_size: 17801
 ---
 
 # Dogs Video Object Tracking Dataset

dogs-video-object-tracking-dataset.py

@@ -81,7 +81,7 @@ class DogsVideoObjectTrackingDataset(datasets.GeneratorBasedBuilder):
     def _split_generators(self, dl_manager):
         images = dl_manager.download_and_extract(f"{_DATA}{self.config.name}.zip")
         annotations = dl_manager.download(f"{_DATA}{self.config.name}.xml")
-        images = dl_manager.iter_files(images)
+        # images = dl_manager.iter_files(images)
         return [
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
@@ -156,20 +156,17 @@ class DogsVideoObjectTrackingDataset(datasets.GeneratorBasedBuilder):
         tree = ET.parse(annotations)
         root = tree.getroot()
 
-        for idx, image in enumerate(images):
-            if "images" in image:
-                i = image.split("_")[-1][:2]
-                image_name = image.split("/")[3]
-                img = self.extract_shapes_from_tracks(root, image_name, i)
-
-                image_id = img.get("id")
-                name = img.get("name")
-                shapes = [self.parse_shape(shape) for shape in img]
-
-                yield idx, {
-                    "id": image_id,
-                    "name": image,
-                    "image": image,
-                    "mask": image,
-                    "shapes": shapes,
-                }
+        for idx, file in enumerate(sorted(os.listdir(f"{images}/images"))):
+            img = self.extract_shapes_from_tracks(root, file, idx)
+
+            image_id = img.get("id")
+            name = img.get("name")
+            shapes = [self.parse_shape(shape) for shape in img]
+
+            yield idx, {
+                "id": image_id,
+                "name": name,
+                "image": f"{images}/images/{file}",
+                "mask": f"{images}/masks/{file}",
+                "shapes": shapes,
+            }