Update app.py

app.py

@@ -34,7 +34,7 @@ CLASS_COLORS = {
     2: (65, 105, 225)    # #4169E1 - Royal blue
 }
 
-# Actual model performance metrics (trained model)
+# Actual model performance metrics
 MODEL_METRICS = {
     'segmentation': {
         'dice': 0.6212,
@@ -53,22 +53,13 @@ MODEL_METRICS = {
 }
 
 
-def create_detection_gauges(num_detections: int, avg_confidence: float) -> go.Figure:
-    """Create radial gauge charts for detection results"""
-    from plotly.subplots import make_subplots
-    
-    fig = make_subplots(
-        rows=1, cols=2,
-        specs=[[{'type': 'indicator'}, {'type': 'indicator'}]],
-        horizontal_spacing=0.15
-    )
-    
-    # Confidence gauge
-    fig.add_trace(go.Indicator(
+def create_gauge_chart(value: float, title: str, max_value: float = 1.0) -> go.Figure:
+    """Create a subtle radial gauge chart"""
+    fig = go.Figure(go.Indicator(
         mode="gauge+number",
-        value=avg_confidence * 100,
+        value=value * 100,
         domain={'x': [0, 1], 'y': [0, 1]},
-        title={'text': 'Avg Confidence', 'font': {'size': 14}},
+        title={'text': title, 'font': {'size': 14}},
         number={'suffix': '%', 'font': {'size': 24}},
         gauge={
             'axis': {'range': [0, 100], 'tickwidth': 1},
@@ -76,36 +67,17 @@ def create_detection_gauges(num_detections: int, avg_confidence: float) -> go.Fi
             'bgcolor': 'rgba(0,0,0,0)',
             'borderwidth': 0,
             'steps': [
-                {'range': [0, 60], 'color': 'rgba(217, 83, 79, 0.1)'},
-                {'range': [60, 80], 'color': 'rgba(240, 173, 78, 0.1)'},
-                {'range': [80, 100], 'color': 'rgba(92, 184, 92, 0.1)'}
-            ]
-        }
-    ), row=1, col=1)
-    
-    # Detections count gauge
-    fig.add_trace(go.Indicator(
-        mode="gauge+number",
-        value=num_detections,
-        domain={'x': [0, 1], 'y': [0, 1]},
-        title={'text': 'Regions Detected', 'font': {'size': 14}},
-        number={'font': {'size': 24}},
-        gauge={
-            'axis': {'range': [0, 10], 'tickwidth': 1},
-            'bar': {'color': '#d9534f' if num_detections > 0 else '#5cb85c', 'thickness': 0.7},
-            'bgcolor': 'rgba(0,0,0,0)',
-            'borderwidth': 0,
-            'steps': [
-                {'range': [0, 1], 'color': 'rgba(92, 184, 92, 0.1)'},
-                {'range': [1, 10], 'color': 'rgba(217, 83, 79, 0.1)'}
+                {'range': [0, 50], 'color': 'rgba(217, 83, 79, 0.1)'},
+                {'range': [50, 75], 'color': 'rgba(240, 173, 78, 0.1)'},
+                {'range': [75, 100], 'color': 'rgba(92, 184, 92, 0.1)'}
             ]
         }
-    ), row=1, col=2)
+    ))
     
     fig.update_layout(
         paper_bgcolor='rgba(0,0,0,0)',
         plot_bgcolor='rgba(0,0,0,0)',
-        height=250,
+        height=200,
         margin=dict(l=20, r=20, t=40, b=20)
     )
     
@@ -148,7 +120,8 @@ class ForgeryDetector:
         Returns:
             original_image: Original uploaded image
             overlay_image: Image with detection overlay
-            detection_gauges: Plotly figure with detection metrics
+            gauge_dice: Dice score gauge
+            gauge_accuracy: Accuracy gauge
             results_html: Detection results as HTML
         """
         # Handle PDF files
@@ -232,17 +205,14 @@ class ForgeryDetector:
         # Create visualization
         overlay = self._create_overlay(original_image, results)
         
-        # Calculate metrics for gauges
-        num_detections = len(results)
-        avg_confidence = sum(r['confidence'] for r in results) / num_detections if num_detections > 0 else 0
-        
-        # Create detection gauge charts
-        detection_gauges = create_detection_gauges(num_detections, avg_confidence)
+        # Create gauge charts
+        gauge_dice = create_gauge_chart(MODEL_METRICS['segmentation']['dice'], 'Segmentation Dice')
+        gauge_accuracy = create_gauge_chart(MODEL_METRICS['classification']['overall_accuracy'], 'Classification Accuracy')
         
         # Create HTML response
         results_html = self._create_html_report(results)
         
-        return original_image, overlay, detection_gauges, results_html
+        return original_image, overlay, gauge_dice, gauge_accuracy, results_html
     
     def _create_overlay(self, image, results):
         """Create overlay visualization"""
@@ -336,19 +306,19 @@ def detect_forgery(file):
     try:
         if file is None:
             empty_html = "<div style='padding:12px; border:1px solid #d9534f; border-radius:8px;'>❌ <b>No file uploaded.</b></div>"
-            return None, None, None, empty_html
+            return None, None, None, None, empty_html
         
         # Get file path
         file_path = file.name if hasattr(file, 'name') else file
         
         # Check if PDF
         if file_path.lower().endswith('.pdf'):
-            original, overlay, gauges, results_html = detector.detect(file_path)
+            original, overlay, gauge_dice, gauge_acc, results_html = detector.detect(file_path)
         else:
             image = Image.open(file_path)
-            original, overlay, gauges, results_html = detector.detect(image)
+            original, overlay, gauge_dice, gauge_acc, results_html = detector.detect(image)
         
-        return original, overlay, gauges, results_html
+        return original, overlay, gauge_dice, gauge_acc, results_html
     
     except Exception as e:
         import traceback
@@ -359,7 +329,7 @@ def detect_forgery(file):
             ❌ <b>Error:</b> {str(e)}
         </div>
         """
-        return None, None, None, error_html
+        return None, None, None, None, error_html
 
 
 # Custom CSS - subtle styling
@@ -385,7 +355,6 @@ with gr.Blocks(css=custom_css) as demo:
     )
     
     with gr.Row():
-        # Left column: Upload
         with gr.Column(scale=1):
             gr.Markdown("### Upload Document")
             input_file = gr.File(
@@ -411,18 +380,17 @@ with gr.Blocks(css=custom_css) as demo:
                 """
             )
         
-        # Right column: Results (Original and Overlay side-by-side)
         with gr.Column(scale=2):
             gr.Markdown("### Detection Results")
             with gr.Row():
                 original_image = gr.Image(label="Original Document", type="numpy")
                 output_image = gr.Image(label="Detected Forgeries", type="numpy")
     
-    # Bottom section: Gauges and Report
     with gr.Row():
         with gr.Column(scale=1):
-            gr.Markdown("### Detection Metrics")
-            detection_gauges = gr.Plot(label="Detection Results")
+            gr.Markdown("### Model Performance")
+            gauge_dice = gr.Plot(label="Segmentation Dice Score")
+            gauge_accuracy = gr.Plot(label="Classification Accuracy")
         
         with gr.Column(scale=1):
             gr.Markdown("### Analysis Report")
@@ -431,16 +399,12 @@ with gr.Blocks(css=custom_css) as demo:
             )
     
     gr.Markdown(
-        f"""
+        """
         ---
-        **Trained Model Performance:**
-        - Segmentation Dice: {MODEL_METRICS['segmentation']['dice']*100:.1f}% | IoU: {MODEL_METRICS['segmentation']['iou']*100:.1f}%
-        - Classification Accuracy: {MODEL_METRICS['classification']['overall_accuracy']*100:.1f}%
-        
         **Model Architecture:**
-        - Localization: MobileNetV3-Small + UNet
-        - Classification: LightGBM with 526 hybrid features
-        - Training: 140K samples from DocTamper dataset
+        - **Localization:** MobileNetV3-Small + UNet (Dice: 62.1%, IoU: 45.1%)
+        - **Classification:** LightGBM with 526 features (Accuracy: 88.97%)
+        - **Training:** 140K samples (DocTamper + SCD + FCD datasets)
         """
     )
     
@@ -448,13 +412,13 @@ with gr.Blocks(css=custom_css) as demo:
     analyze_btn.click(
         fn=detect_forgery,
         inputs=[input_file],
-        outputs=[original_image, output_image, detection_gauges, output_html]
+        outputs=[original_image, output_image, gauge_dice, gauge_accuracy, output_html]
     )
     
     clear_btn.click(
-        fn=lambda: (None, None, None, None, "<i>No analysis yet. Upload a document and click Analyze.</i>"),
+        fn=lambda: (None, None, None, None, None, "<i>No analysis yet. Upload a document and click Analyze.</i>"),
         inputs=None,
-        outputs=[input_file, original_image, output_image, detection_gauges, output_html]
+        outputs=[input_file, original_image, output_image, gauge_dice, gauge_accuracy, output_html]
     )