Upload folder using huggingface_hub

.gitattributes

@@ -57,3 +57,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+train.jsonl filter=lfs diff=lfs merge=lfs -text

images.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fd1a4cae5a787c50460cd085dac5995d92ac87e5dcbc54e0b61b3276fb3b6c1
+size 1407451163

train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c92b25a69c9f6d616b6f5d53b335e3560a3cc3f9f2339d279a93a0b4a5401b0c
+size 645695224

visual_dataset.py

@@ -0,0 +1,115 @@
+#!/usr/bin/env python3
+
+import os
+import json
+import base64
+import argparse
+from datetime import datetime
+from pathlib import Path
+from PIL import Image
+from io import BytesIO
+from tqdm import tqdm
+
+def get_output_filename():
+    """Generate unique output filename with timestamp"""
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    return f"output_images_{timestamp}.jsonl"
+
+def encode_image_base64(pil_image):
+    """Convert PIL Image to base64 string"""
+    buffered = BytesIO()
+    # Convert to RGB if RGBA
+    if pil_image.mode == 'RGBA':
+        pil_image = pil_image.convert('RGB')
+    pil_image.save(buffered, format="JPEG", quality=95)
+    return base64.b64encode(buffered.getvalue()).decode('utf-8')
+
+def process_image(image_path, output_file):
+    """Process single image and write to JSONL with conditional resizing"""
+    MAX_WIDTH = 800
+    MAX_HEIGHT = 1000
+    
+    try:
+        image = Image.open(image_path)
+        width, height = image.size
+        
+        # Only resize if image exceeds max dimensions
+        if width > MAX_WIDTH or height > MAX_HEIGHT:
+            # Calculate aspect ratio preserving dimensions
+            ratio = min(MAX_WIDTH/width, MAX_HEIGHT/height)
+            new_size = (int(width * ratio), int(height * ratio))
+            processed_image = image.resize(new_size, Image.Resampling.LANCZOS)
+        else:
+            processed_image = image
+            new_size = (width, height)
+        
+        entry = {
+            "source": str(image_path),
+            "timestamp": datetime.now().isoformat(),
+            "width": new_size[0],
+            "height": new_size[1],
+            "original_width": width,
+            "original_height": height,
+            "was_resized": width > MAX_WIDTH or height > MAX_HEIGHT,
+            "image": encode_image_base64(processed_image)
+        }
+        
+        with open(output_file, 'a') as f:
+            f.write(json.dumps(entry) + '\n')
+            
+        return True
+    
+    except Exception as e:
+        print(f"Error processing {image_path}: {str(e)}")
+        return False
+
+def find_image_files(directory):
+    """Recursively find all image files in directory"""
+    image_extensions = {'.jpg', '.jpeg', '.png'}
+    image_files = []
+    
+    for path in Path(directory).rglob('*'):
+        if path.suffix.lower() in image_extensions:
+            image_files.append(path)
+    
+    return image_files
+
+def main():
+    parser = argparse.ArgumentParser(description='Convert images to JSONL format')
+    parser.add_argument('--input', default='./images',
+                        help='Input directory containing images (default: ./images)')
+    parser.add_argument('--output', default=None,
+                        help='Output JSONL file (default: output_images_<timestamp>.jsonl)')
+    
+    args = parser.parse_args()
+    
+    # Setup paths
+    input_dir = Path(args.input)
+    output_file = args.output or get_output_filename()
+    
+    if not input_dir.exists():
+        raise FileNotFoundError(f"Input directory {input_dir} not found")
+    
+    # Get list of image files
+    image_files = find_image_files(input_dir)
+    
+    if not image_files:
+        print(f"No image files found in {input_dir}")
+        return
+    
+    print(f"Found {len(image_files)} image files")
+    print(f"Output will be written to: {output_file}")
+    
+    # Process each image
+    successful = 0
+    for image_file in tqdm(image_files, desc="Processing images"):
+        if process_image(image_file, output_file):
+            successful += 1
+        # Print progress after each file
+        tqdm.write(f"Processed {image_file}")
+    
+    print(f"\nComplete! Successfully processed {successful} out of {len(image_files)} files")
+    print(f"Output written to: {output_file}")
+
+if __name__ == "__main__":
+    main()