Update app.py

app.py

@@ -84,6 +84,82 @@ def create_gauge_chart(value: float, title: str, max_value: float = 1.0) -> go.F
     return fig
 
 
+def create_detection_metrics_gauge(avg_confidence: float, iou: float, precision: float, recall: float, num_detections: int) -> go.Figure:
+    """Create concentric radial gauge showing actual detection metrics"""
+    
+    # Calculate percentages for display
+    confidence_pct = avg_confidence * 100 if num_detections > 0 else 0
+    iou_pct = iou * 100
+    precision_pct = precision * 100
+    recall_pct = recall * 100
+    
+    fig = go.Figure()
+    
+    # Add concentric gauge rings (from outer to inner)
+    fig.add_trace(go.Indicator(
+        mode="gauge",
+        value=confidence_pct,
+        domain={'x': [0, 1], 'y': [0, 1]},
+        title={'text': f"Confidence: {confidence_pct:.1f}%", 'font': {'size': 11}},
+        gauge={
+            'axis': {'range': [0, 100], 'visible': False},
+            'bar': {'color': '#4169E1', 'thickness': 0.15},
+            'bgcolor': 'rgba(0,0,0,0.05)',
+            'borderwidth': 0,
+        }
+    ))
+    
+    fig.add_trace(go.Indicator(
+        mode="gauge",
+        value=precision_pct,
+        domain={'x': [0.1, 0.9], 'y': [0.1, 0.9]},
+        title={'text': f"Precision: {precision_pct:.1f}%", 'font': {'size': 10}},
+        gauge={
+            'axis': {'range': [0, 100], 'visible': False},
+            'bar': {'color': '#5cb85c', 'thickness': 0.15},
+            'bgcolor': 'rgba(0,0,0,0.05)',
+            'borderwidth': 0,
+        }
+    ))
+    
+    fig.add_trace(go.Indicator(
+        mode="gauge",
+        value=recall_pct,
+        domain={'x': [0.2, 0.8], 'y': [0.2, 0.8]},
+        title={'text': f"Recall: {recall_pct:.1f}%", 'font': {'size': 9}},
+        gauge={
+            'axis': {'range': [0, 100], 'visible': False},
+            'bar': {'color': '#f0ad4e', 'thickness': 0.15},
+            'bgcolor': 'rgba(0,0,0,0.05)',
+            'borderwidth': 0,
+        }
+    ))
+    
+    fig.add_trace(go.Indicator(
+        mode="gauge+number",
+        value=iou_pct,
+        domain={'x': [0.3, 0.7], 'y': [0.3, 0.7]},
+        title={'text': "IoU", 'font': {'size': 10}},
+        number={'suffix': '%', 'font': {'size': 16}},
+        gauge={
+            'axis': {'range': [0, 100], 'visible': False},
+            'bar': {'color': '#d9534f', 'thickness': 0.2},
+            'bgcolor': 'rgba(0,0,0,0.05)',
+            'borderwidth': 0,
+        }
+    ))
+    
+    fig.update_layout(
+        paper_bgcolor='rgba(0,0,0,0)',
+        plot_bgcolor='rgba(0,0,0,0)',
+        height=300,
+        margin=dict(l=20, r=20, t=60, b=20),
+        showlegend=False
+    )
+    
+    return fig
+
+
 class ForgeryDetector:
     """Main forgery detection pipeline"""
     
@@ -211,10 +287,38 @@ class ForgeryDetector:
         # Create visualization
         overlay = self._create_overlay(original_image, results)
         
+        # Calculate actual detection metrics from probability map and mask
+        num_detections = len(results)
+        avg_confidence = sum(r['confidence'] for r in results) / num_detections if num_detections > 0 else 0
+        
+        # Calculate IoU, Precision, Recall from the refined mask and probability map
+        if num_detections > 0:
+            # Use high-confidence areas from probability map as "predicted positive"
+            high_conf_mask = (prob_map > 0.7).astype(np.uint8)
+            predicted_positive = np.sum(refined_mask > 0)
+            high_conf_positive = np.sum(high_conf_mask > 0)
+            
+            # Calculate intersection and union
+            intersection = np.sum((refined_mask > 0) & (high_conf_mask > 0))
+            union = np.sum((refined_mask > 0) | (high_conf_mask > 0))
+            
+            # Calculate metrics
+            iou = intersection / union if union > 0 else 0
+            precision = intersection / predicted_positive if predicted_positive > 0 else 0
+            recall = intersection / high_conf_positive if high_conf_positive > 0 else 0
+        else:
+            # No detections - use zeros
+            iou = 0
+            precision = 0
+            recall = 0
+        
+        # Create detection metrics gauge with actual values
+        metrics_gauge = create_detection_metrics_gauge(avg_confidence, iou, precision, recall, num_detections)
+        
         # Create HTML response
         results_html = self._create_html_report(results)
         
-        return overlay, results_html
+        return overlay, metrics_gauge, results_html
     
     def _create_overlay(self, image, results):
         """Create overlay visualization"""
@@ -308,15 +412,15 @@ 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, empty_html
+            return None, None, empty_html
         
         # Get file path
         file_path = file if isinstance(file, str) else file
         
         # Detect forgeries
-        overlay, results_html = detector.detect(file_path)
+        overlay, metrics_gauge, results_html = detector.detect(file_path)
         
-        return overlay, results_html
+        return overlay, metrics_gauge, results_html
     
     except Exception as e:
         import traceback
@@ -327,7 +431,7 @@ def detect_forgery(file):
             ❌ <b>Error:</b> {str(e)}
         </div>
         """
-        return None, error_html
+        return None, None, error_html
 
 
 # Custom CSS - subtle styling
@@ -397,6 +501,10 @@ with gr.Blocks(css=custom_css) as demo:
             output_html = gr.HTML(
                 value="<i>No analysis yet. Upload a document and click Analyze.</i>"
             )
+        
+        with gr.Column(scale=1):
+            gr.Markdown("### Detection Metrics")
+            metrics_gauge = gr.Plot(label="Concentric Metrics Gauge")
     
     gr.Markdown("---")
     
@@ -441,13 +549,13 @@ with gr.Blocks(css=custom_css) as demo:
     analyze_btn.click(
         fn=detect_forgery,
         inputs=[input_file],
-        outputs=[output_image, output_html]
+        outputs=[output_image, metrics_gauge, output_html]
     )
     
     clear_btn.click(
-        fn=lambda: (None, None, "<i>No analysis yet. Upload a document and click Analyze.</i>"),
+        fn=lambda: (None, None, None, "<i>No analysis yet. Upload a document and click Analyze.</i>"),
         inputs=None,
-        outputs=[input_file, output_image, output_html]
+        outputs=[input_file, output_image, metrics_gauge, output_html]
     )