Final commit

app.py

@@ -68,12 +68,13 @@ def process_frame(frame):
     return prediction, output.item(), visualization
 
 @spaces.GPU
-def analyze_video(video_path, sample_rate=30, top_n=5):
+def analyze_video(video_path, sample_rate=30, top_n=5, detection_threshold=0.5):
     cap = cv2.VideoCapture(video_path)
     frame_count = 0
     fake_count = 0
     total_processed = 0
     frames_info = []
+    confidence_scores = []
 
     while cap.isOpened():
         ret, frame = cap.read()
@@ -86,6 +87,7 @@ def analyze_video(video_path, sample_rate=30, top_n=5):
 
             if prediction is not None:
                 total_processed += 1
+                confidence_scores.append(confidence)
                 if prediction == "fake":
                     fake_count += 1
 
@@ -102,12 +104,23 @@ def analyze_video(video_path, sample_rate=30, top_n=5):
 
     if total_processed > 0:
         fake_percentage = (fake_count / total_processed) * 100
+        average_confidence = sum(confidence_scores) / len(confidence_scores)
+        model_confidence = 1 - (sum((score - average_confidence) ** 2 for score in confidence_scores) / len(confidence_scores))
+        
         frames_info.sort(key=lambda x: x['confidence'], reverse=True)
         top_frames = frames_info[:top_n]
 
-        return fake_percentage, top_frames
+        return {
+            'fake_percentage': fake_percentage,
+            'is_likely_deepfake': fake_percentage >= 60,
+            'top_frames': top_frames,
+            'model_confidence': model_confidence,
+            'total_frames_analyzed': total_processed,
+            'average_confidence_score': average_confidence,
+            'detection_threshold': detection_threshold
+        }
     else:
-        return 0, []
+        return None
 
 @app.route('/analyze', methods=['POST'])
 def analyze_video_api():
@@ -125,20 +138,17 @@ def analyze_video_api():
         file.save(filepath)
 
         try:
-            fake_percentage, top_frames = analyze_video(filepath)
+            result = analyze_video(filepath)
             os.remove(filepath)  # Remove the file after analysis
 
-            # Convert numpy arrays to base64 encoded strings
-            for frame in top_frames:
-                frame['visualization'] = base64.b64encode(cv2.imencode('.png', frame['visualization'])[1]).decode('utf-8')
+            if result:
+                # Convert numpy arrays to base64 encoded strings
+                for frame in result['top_frames']:
+                    frame['visualization'] = base64.b64encode(cv2.imencode('.png', frame['visualization'])[1]).decode('utf-8')
 
-            result = {
-                'fake_percentage': round(fake_percentage, 2),
-                'is_likely_deepfake': fake_percentage >= 60,
-                'top_frames': top_frames
-            }
-
-            return jsonify(result), 200
+                return jsonify(result), 200
+            else:
+                return jsonify({'error': 'No frames could be processed'}), 400
         except Exception as e:
             os.remove(filepath)  # Remove the file if an error occurs
             return jsonify({'error': str(e)}), 500