Upload k600/train/train/generate_json_dataset.py with huggingface_hub

k600/train/train/generate_json_dataset.py

@@ -6,36 +6,29 @@ import random
 def generate_dataset_json(root_dir='.', output_file='dataset.json'):
     dataset = []
     
-    # Get all directories in root_dir (each represents an action)
     actions = [d for d in os.listdir(root_dir) if os.path.isdir(os.path.join(root_dir, d))]
     
     for action in actions:
         action_dir = os.path.join(root_dir, action)
         
-        # Find all mp4 files in this action directory
         video_files = glob(os.path.join(action_dir, '*.mp4'))
         
         for video_path in video_files:
-            # Create relative path from root_dir
             rel_path = os.path.relpath(video_path, root_dir)
             
-            # Create caption by capitalizing the action name
             caption = action.capitalize()
             
             dataset.append({
-                "video": rel_path.replace('\\', '/'),  # Use forward slashes for consistency
+                "video": rel_path.replace('\\', '/'),
                 "caption": caption
             })
     
-    # Shuffle the dataset with seed 42 for reproducibility
     random.seed(42)
     random.shuffle(dataset)
     
-    # Write to JSON file
     with open(output_file, 'w') as f:
         json.dump(dataset, f, indent=2)
     
     print(f"Dataset JSON file created at {output_file} with {len(dataset)} entries.")
 
-# Example usage:
-generate_dataset_json(root_dir='./', output_file='kinetics.json')
+generate_dataset_json(root_dir='./', output_file='kinetics.json')