Update app.py

app.py

@@ -39,6 +39,7 @@ def overlay_masks(image: Image.Image, masks: torch.Tensor) -> Image.Image:
     return image
 
 spaces.GPU()
+
 def segment(image: Image.Image, text: str, threshold: float, mask_threshold: float):
     """
     Perform promptable concept segmentation using SAM3.
@@ -73,6 +74,15 @@ def segment(image: Image.Image, text: str, threshold: float, mask_threshold: flo
     except Exception as e:
         return image, f"❌ Error during segmentation: {str(e)}"
 
+def clear_all():
+    """Clear all inputs and outputs"""
+    return None, "", None, 0.5, 0.5
+
+def segment_example(image_path: str, prompt: str):
+    """Handle example clicks"""
+    image = Image.open(image_path) if image_path else None
+    return segment(image, prompt, 0.5, 0.5)
+
 # Gradio Interface
 with gr.Blocks(
     theme=gr.themes.Soft(),
@@ -98,7 +108,6 @@ with gr.Blocks(
             label="Input Image",
             type="pil",
             height=400,
-            sources=["upload", "url"],
         )
         image_output = gr.Image(
             label="Output (Segmented Image)",
@@ -112,9 +121,7 @@ with gr.Blocks(
             placeholder="e.g., a person, ear, cat, bicycle...",
             scale=3
         )
-        gr.Button("🔍 Clear", size="sm", variant="secondary").click(
-            fn=lambda: (None, "", None, 0.5, 0.5), outputs=[image_output, text_input, image_input, thresh_slider, mask_thresh_slider]
-        )
+        clear_btn = gr.Button("🔍 Clear", size="sm", variant="secondary")
     
     with gr.Row():
         thresh_slider = gr.Slider(
@@ -141,14 +148,19 @@ with gr.Blocks(
     
     segment_btn = gr.Button("🎯 Segment", variant="primary", size="lg")
     
-    # Event
+    # Clear button handler
+    clear_btn.click(
+        fn=clear_all,
+        outputs=[image_input, text_input, image_output, thresh_slider, mask_thresh_slider]
+    )
+    
+    # Segment button handler
     segment_btn.click(
         fn=segment,
         inputs=[image_input, text_input, thresh_slider, mask_thresh_slider],
         outputs=[image_output, info_output]
     ).then(
-        fn=lambda: gr.Info("Segmentation complete!"),
-        _js="() => {}"
+        fn=lambda: None,
     )
     
     # Examples
@@ -178,7 +190,7 @@ with gr.Blocks(
     gr.Examples(
         examples=examples,
         inputs=[image_input, text_input],
-        fn=segment,
+        fn=segment_example,
         outputs=[image_output, info_output],
         cache_examples=True,
         examples_per_page=10,
@@ -197,3 +209,108 @@ with gr.Blocks(
 
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0", server_port=7860, share=False, debug=True)
+```
+
+=== utils.py ===
+```python
+import torch
+import numpy as np
+from PIL import Image
+import matplotlib
+import requests
+from io import BytesIO
+
+def load_image_from_url(url: str) -> Image.Image:
+    """
+    Load an image from a URL.
+    
+    Args:
+        url: Image URL
+        
+    Returns:
+        PIL Image object
+    """
+    try:
+        response = requests.get(url, timeout=10)
+        response.raise_for_status()
+        image = Image.open(BytesIO(response.content))
+        return image.convert("RGB")
+    except Exception as e:
+        raise ValueError(f"Could not load image from URL: {str(e)}")
+
+def validate_image(image: Image.Image) -> bool:
+    """
+    Validate if the image is suitable for processing.
+    
+    Args:
+        image: PIL Image object
+        
+    Returns:
+        True if valid, False otherwise
+    """
+    if image is None:
+        return False
+    
+    if image.size[0] <= 0 or image.size[1] <= 0:
+        return False
+        
+    return True
+
+def resize_for_processing(image: Image.Image, max_size: int = 1024) -> Image.Image:
+    """
+    Resize image for processing while maintaining aspect ratio.
+    
+    Args:
+        image: Input PIL Image
+        max_size: Maximum size for the longer dimension
+        
+    Returns:
+        Resized PIL Image
+    """
+    width, height = image.size
+    if max(width, height) <= max_size:
+        return image
+    
+    if width > height:
+        new_width = max_size
+        new_height = int(height * max_size / width)
+    else:
+        new_height = max_size
+        new_width = int(width * max_size / height)
+    
+    return image.resize((new_width, new_height), Image.Resampling.LANCZOS)
+
+def overlay_masks_advanced(image: Image.Image, masks: torch.Tensor, alpha: float = 0.5) -> Image.Image:
+    """
+    Advanced overlay function with customizable alpha.
+    
+    Args:
+        image: Input PIL Image
+        masks: Segmentation masks tensor
+        alpha: Overlay transparency (0-1)
+        
+    Returns:
+        Overlaid PIL Image
+    """
+    image = image.convert("RGBA")
+    masks = 255 * masks.cpu().numpy().astype(np.uint8)
+    
+    n_masks = masks.shape[0]
+    if n_masks == 0:
+        return image.convert("RGB")
+    
+    # Use a good colormap
+    cmap = matplotlib.colormaps.get_cmap("tab10").resampled(n_masks)
+    colors = [
+        tuple(int(c * 255) for c in cmap(i)[:3])
+        for i in range(n_masks)
+    ]
+
+    for mask, color in zip(masks, colors):
+        mask_img = Image.fromarray(mask)
+        overlay = Image.new("RGBA", image.size, color + (0,))
+        alpha_map = mask_img.point(lambda v: int(v * alpha * 255))
+        overlay.putalpha(alpha_map)
+        image = Image.alpha_composite(image, overlay)
+    
+    return image