update app.py

app.py

@@ -67,7 +67,6 @@ except Exception as e:
     print(f"Failed to load model: {e}")
     import traceback
     traceback.print_exc()
-    # Create a dummy model as fallback
     model = get_model(NUM_CLASSES).to(DEVICE)
     model.eval()
     print("Using untrained model as fallback")
@@ -79,7 +78,6 @@ class GradCAM:
         self.target_layer = target_layer
         self.gradients = None
         self.activations = None
-        self.hook_handles = []
         
     def save_activation(self, module, input, output):
         self.activations = output.detach()
@@ -88,33 +86,23 @@ class GradCAM:
         self.gradients = grad_output[0].detach()
     
     def generate_cam(self, input_image, target_class=None):
-        # Register hooks
         forward_handle = self.target_layer.register_forward_hook(self.save_activation)
         backward_handle = self.target_layer.register_full_backward_hook(self.save_gradient)
         
         try:
-            # Forward pass
             model_output = self.model(input_image)
             
             if target_class is None:
                 target_class = model_output.argmax(dim=1).item()
             
-            # Backward pass
             self.model.zero_grad()
             class_score = model_output[0, target_class]
             class_score.backward(retain_graph=False)
             
-            if self.gradients is None or self.activations is None:
-                return np.zeros((7, 7))  # Default size for ResNet layer4
-            
             gradients = self.gradients[0]
             activations = self.activations[0]
-            
-            # Global average pooling of gradients
             weights = gradients.mean(dim=(1, 2), keepdim=True)
             cam = (weights * activations).sum(dim=0)
-            
-            # Apply ReLU and normalize
             cam = F.relu(cam)
             cam = cam - cam.min()
             if cam.max() > 0:
@@ -122,176 +110,113 @@ class GradCAM:
             
             return cam.detach().cpu().numpy()
         finally:
-            # Remove hooks
             forward_handle.remove()
             backward_handle.remove()
             self.gradients = None
             self.activations = None
 
 def overlay_heatmap(image, heatmap, alpha=0.4):
-    """Overlay heatmap on original image"""
     heatmap_resized = cv2.resize(heatmap, (image.shape[1], image.shape[0]))
     heatmap_colored = cv2.applyColorMap(np.uint8(255 * heatmap_resized), cv2.COLORMAP_JET)
-    output = cv2.addWeighted(image, 1 - alpha, heatmap_colored, alpha, 0)
-    return output
+    return cv2.addWeighted(image, 1 - alpha, heatmap_colored, alpha, 0)
 
 def predict_galaxy(image):
-    """Predict galaxy morphology and generate Grad-CAM"""
     if image is None:
         return None, "Please upload an image."
     
     if model is None:
         return None, "Error: Model not loaded. Please check the logs."
     
-    try:
-        # Ensure model is in eval mode
-        model.eval()
-        
-        # Convert image to PIL if it's not already
-        if isinstance(image, np.ndarray):
-            image = Image.fromarray(image.astype('uint8'))
-        elif not isinstance(image, Image.Image):
-            image = Image.open(image) if hasattr(image, 'read') else Image.fromarray(np.array(image))
-        
-        # Ensure image is RGB
-        if image.mode != 'RGB':
-            image = image.convert('RGB')
-        
-        # Preprocess image
-        img_tensor = preprocess(image).unsqueeze(0).to(DEVICE)
-        img_tensor.requires_grad = True
-        
-        # Get prediction
-        with torch.set_grad_enabled(True):
-            outputs = model(img_tensor)
-            probs = F.softmax(outputs, dim=1)
-            pred_class = outputs.argmax(dim=1).item()
-            confidence = probs[0][pred_class].item()
-            
-            # Generate Grad-CAM
-            try:
-                gradcam = GradCAM(model, model.layer4)
-                cam = gradcam.generate_cam(img_tensor, pred_class)
-            except Exception as cam_error:
-                print(f"Grad-CAM error: {cam_error}")
-                import traceback
-                traceback.print_exc()
-                # If Grad-CAM fails, just return the original image
-                cam = None
-        
-        # Prepare original image for overlay
-        img_np = np.array(image)
-        img_resized = cv2.resize(img_np, (224, 224))
-        
-        # Create overlay if Grad-CAM succeeded
-        if cam is not None:
-            try:
-                overlay = overlay_heatmap(img_resized, cam)
-                overlay_rgb = cv2.cvtColor(overlay, cv2.COLOR_BGR2RGB)
-                overlay_pil = Image.fromarray(overlay_rgb)
-            except Exception as overlay_error:
-                print(f"Overlay error: {overlay_error}")
-                overlay_pil = image.resize((224, 224))
-        else:
-            overlay_pil = image.resize((224, 224))
-        
-        # Format results
-        result_text = f"Predicted Class: {CLASS_NAMES[pred_class]}\nConfidence: {confidence:.2%}"
-        
-        # Ensure we return PIL Image
-        if not isinstance(overlay_pil, Image.Image):
-            overlay_pil = Image.fromarray(np.array(overlay_pil))
-        
-        return overlay_pil, str(result_text)
-    except Exception as e:
-        import traceback
-        error_msg = f"Error: {str(e)}\n{traceback.format_exc()}"
-        print(error_msg)  # Print for debugging
-        return None, f"Error: {str(e)}"
+    model.eval()
+    
+    if isinstance(image, np.ndarray):
+        image = Image.fromarray(image.astype("uint8"))
+    elif not isinstance(image, Image.Image):
+        image = Image.open(image)
+    
+    if image.mode != "RGB":
+        image = image.convert("RGB")
+    
+    img_tensor = preprocess(image).unsqueeze(0).to(DEVICE)
+    img_tensor.requires_grad = True
+    
+    outputs = model(img_tensor)
+    probs = F.softmax(outputs, dim=1)
+    pred_class = outputs.argmax(dim=1).item()
+    confidence = probs[0][pred_class].item()
+    
+    gradcam = GradCAM(model, model.layer4)
+    cam = gradcam.generate_cam(img_tensor, pred_class)
+    
+    img_np = np.array(image.resize((224, 224)))
+    overlay = overlay_heatmap(img_np, cam)
+    overlay_rgb = cv2.cvtColor(overlay, cv2.COLOR_BGR2RGB)
+    overlay_pil = Image.fromarray(overlay_rgb)
+    
+    result_text = f"Predicted Class: {CLASS_NAMES[pred_class]}\nConfidence: {confidence:.2%}"
+    
+    return overlay_pil, result_text
 
-# Custom CSS for black background and white text
+# =========================
+# Custom CSS
+# =========================
 custom_css = """
-    .gradio-container {
-        background-color: #000000 !important;
-        color: #ffffff !important;
-    }
-    body {
-        background-color: #000000 !important;
-        color: #ffffff !important;
-    }
-    .gradio-container * {
-        color: #ffffff !important;
-    }
-    h1, h2, h3, h4, p, label, span, div {
-        color: #ffffff !important;
-    }
-    .gr-markdown, .gr-markdown * {
-        color: #ffffff !important;
-    }
-    .gr-button {
-        background-color: #333333 !important;
-        color: #ffffff !important;
-        border: 1px solid #555555 !important;
-    }
-    .gr-button:hover {
-        background-color: #555555 !important;
-    }
-    .gr-textbox, .gr-textbox input, .gr-textbox textarea {
-        background-color: #1a1a1a !important;
-        color: #ffffff !important;
-        border: 1px solid #555555 !important;
-    }
-    .gr-image {
-        background-color: #000000 !important;
-        border: none !important;
-        padding: 0 !important;
-        margin: 0 !important;
-    }
-    .gr-image img {
-        border: none !important;
-        box-shadow: none !important;
-        background-color: #000000 !important;
-    }
-    .gr-image-container, .image-container, .image-wrapper {
-        border: none !important;
-        background-color: #000000 !important;
-        padding: 0 !important;
-        margin: 0 !important;
-    }
-    .gr-image .toolbar, .gr-image .image-controls {
-        display: none !important;
-    }
-    .gr-image label, .gr-image .label-wrap {
-        display: none !important;
-    }
-    .gr-box {
-        border: none !important;
-        background-color: #000000 !important;
-    }
-    .panel, .panel-header {
-        background-color: #000000 !important;
-        border: none !important;
-    }
+.gradio-container {
+    background-color: #000000 !important;
+    color: #ffffff !important;
+}
+body {
+    background-color: #000000 !important;
+    color: #ffffff !important;
+}
+
+/* 🔴 FIX 1: REMOVED unsafe global selector */
+/* .gradio-container * { color: #ffffff !important; } */
+
+h1, h2, h3, h4, p, label, span, div {
+    color: #ffffff !important;
+}
+.gr-markdown, .gr-markdown * {
+    color: #ffffff !important;
+}
+.gr-button {
+    background-color: #333333 !important;
+    color: #ffffff !important;
+    border: 1px solid #555555 !important;
+}
+.gr-button:hover {
+    background-color: #555555 !important;
+}
+.gr-textbox, .gr-textbox input, .gr-textbox textarea {
+    background-color: #1a1a1a !important;
+    color: #ffffff !important;
+    border: 1px solid #555555 !important;
+}
+.gr-image {
+    background-color: #000000 !important;
+    border: none !important;
+}
+.gr-image img {
+    background-color: #000000 !important;
+}
 """
 
-# Create Gradio interface
-# Note: There's a known Gradio bug with API schema generation that causes errors
-# The app will still work for classification, but API endpoints may fail
+# =========================
+# UI
+# =========================
 with gr.Blocks(css=custom_css) as demo:
-    # Landing Section
+
     with gr.Column():
-        landing_img = gr.Image(value="landing.jpg", height=500, show_label=False, container=False)
-        landing_text = gr.Markdown("""
-        <div style="text-align: center; padding: 30px; color: white; background-color: #000000; width: 100%; display: flex; flex-direction: column; align-items: center; justify-content: center;">
-            <h1 style="font-size: 96px; font-weight: bold; margin: 0 auto 30px auto; text-align: center; width: 100%;">Galaxy Morphology AI</h1>
-            <p style="font-size: 56px; font-weight: normal; margin: 0 auto; text-align: center; width: 100%;">Classify galaxies with state-of-the-art deep learning</p>
+        gr.Image(value="landing.jpg", height=500, show_label=False, container=False)
+        gr.Markdown("""
+        <div style="text-align: center; padding: 30px;">
+            <h1 style="font-size: 96px; font-weight: bold;">Galaxy Morphology AI</h1>
+            <p style="font-size: 56px;">Classify galaxies with state-of-the-art deep learning</p>
         </div>
         """)
-    
-    # Spacing between sections
+
     gr.Markdown("<div style='height: 60px;'></div>")
-    
-    # How Astrophysicists Use This Section
+
     with gr.Row():
         with gr.Column(scale=1):
             gr.Markdown("""
@@ -310,16 +235,14 @@ with gr.Blocks(css=custom_css) as demo:
             extragalactic astronomy.
             """)
         with gr.Column(scale=1):
-            astro_img = gr.Image(value="astro.jpg", show_label=False, container=False, height=400)
-            gr.Markdown("<p style='text-align: center; color: white; margin-top: 10px;'>Astrophysics Research</p>")
-    
-    # Spacing between sections
+            gr.Image(value="astro.jpg", show_label=False, container=False, height=400)
+            gr.Markdown("<p style='text-align: center;'>Astrophysics Research</p>")
+
     gr.Markdown("<div style='height: 60px;'></div>")
-    
-    # Classification Section
+
     gr.Markdown("# Galaxy Morphology Classification")
     gr.Markdown("Upload a galaxy image to classify its morphology and visualize the model's attention using Grad-CAM.")
-    
+
     with gr.Row():
         with gr.Column():
             input_image = gr.Image(label="Upload Galaxy Image")
@@ -327,21 +250,17 @@ with gr.Blocks(css=custom_css) as demo:
         
         with gr.Column():
             output_image = gr.Image(label="Grad-CAM Visualization")
-            result_text = gr.Textbox(label="Classification Result")
-    
-    # Register the classification function
-    # Disable API to avoid Gradio schema generation bug
+            result_text = gr.Markdown()  # 🔴 FIX 2: Textbox → Markdown (read-only)
+
     classify_btn.click(
         fn=predict_galaxy,
         inputs=[input_image],
         outputs=[output_image, result_text],
         api_name=False
     )
-    
-    # Spacing between sections
+
     gr.Markdown("<div style='height: 60px;'></div>")
-    
-    # Dark Energy Section
+
     gr.Markdown("""
     # Understanding Dark Energy Through Galaxy Morphology
     
@@ -364,13 +283,9 @@ with gr.Blocks(css=custom_css) as demo:
     its role in the fate of the universe.
     """)
 
-# Launch the demo
-# For Hugging Face Spaces, Gradio will automatically detect and launch the demo
-# The API error is a known Gradio bug - the app will still work for classification
+# Launch
 if __name__ == "__main__":
     try:
         demo.launch(show_api=False)
-    except Exception as e:
-        # If launch fails, try without API
-        print(f"Launch error (non-critical): {e}")
+    except Exception:
         demo.launch()