Update app.py

app.py

@@ -51,7 +51,7 @@ class HeartPredictor:
 
     def load_normalization_params(self):
         """Load the saved normalization parameters from training"""
-        norm_file = "heart_norm_params.pkl"
+        norm_file = "comprehensive_norm_params.pkl"
         if os.path.exists(norm_file):
             with open(norm_file, 'rb') as f:
                 self.norm_params = pickle.load(f)
@@ -233,59 +233,43 @@ class HeartPredictor:
     
     def _predict_segment(self, audio_segment, debug=False):
         """Predict on a single segment using training methodology"""
-        print("Starting segment prediction")
-        try:
-            # Apply same denoising as training
-            print("Applying denoising...")
-            denoised_audio = self.denoise_audio(audio_segment, TARGET_SR)
-            print(f"Denoised audio stats: mean={np.mean(denoised_audio):.6f}, std={np.std(denoised_audio):.6f}")
-            
-            # Extract features exactly as in training
-            print("Extracting features...")
-            features = self.extract_features(denoised_audio, TARGET_SR)
-            
-            # Prepare inputs exactly as in training  
-            print("Preparing inputs...")
-            X_mfcc, X_chroma, X_mspec = self.prepare_input(features)
+        # Apply same denoising as training
+        denoised_audio = self.denoise_audio(audio_segment, TARGET_SR)
+        
+        # Extract features exactly as in training
+        features = self.extract_features(denoised_audio, TARGET_SR)
+        
+        # Prepare inputs exactly as in training  
+        X_mfcc, X_chroma, X_mspec = self.prepare_input(features)
 
-            # FIXED: Use training normalization approach
-            print("Normalizing features...")
-            X_mfcc_norm = self.normalize_with_training_stats(X_mfcc, 'mfcc')
-            X_chroma_norm = self.normalize_with_training_stats(X_chroma, 'chroma') 
-            X_mspec_norm = self.normalize_with_training_stats(X_mspec, 'mspec')
+        # FIXED: Use training normalization approach
+        X_mfcc_norm = self.normalize_with_training_stats(X_mfcc, 'mfcc')
+        X_chroma_norm = self.normalize_with_training_stats(X_chroma, 'chroma') 
+        X_mspec_norm = self.normalize_with_training_stats(X_mspec, 'mspec')
 
-            print("Running model prediction...")
-            # Get model prediction
-            raw_prediction = self.model.predict([X_mfcc_norm, X_chroma_norm, X_mspec_norm], verbose=0)
-            pred_prob = float(raw_prediction[0][0])
-            print(f"Model raw output: {pred_prob}")
-            
-            result = {
-                "raw_output": pred_prob,
-                "audio_stats": {
-                    "mean": float(np.mean(denoised_audio)),
-                    "std": float(np.std(denoised_audio))
-                } if debug else {}
+        # Get model prediction
+        raw_prediction = self.model.predict([X_mfcc_norm, X_chroma_norm, X_mspec_norm], verbose=0)
+        pred_prob = float(raw_prediction[0][0])
+        
+        result = {
+            "raw_output": pred_prob,
+            "audio_stats": {
+                "mean": float(np.mean(denoised_audio)),
+                "std": float(np.std(denoised_audio))
+            } if debug else {}
+        }
+        
+        if debug:
+            result["normalization_check"] = {
+                "mfcc_norm_mean": float(np.mean(X_mfcc_norm)),
+                "mfcc_norm_std": float(np.std(X_mfcc_norm)),
+                "chroma_norm_mean": float(np.mean(X_chroma_norm)),
+                "chroma_norm_std": float(np.std(X_chroma_norm)),
+                "mspec_norm_mean": float(np.mean(X_mspec_norm)),
+                "mspec_norm_std": float(np.std(X_mspec_norm))
             }
-            
-            if debug:
-                result["normalization_check"] = {
-                    "mfcc_norm_mean": float(np.mean(X_mfcc_norm)),
-                    "mfcc_norm_std": float(np.std(X_mfcc_norm)),
-                    "chroma_norm_mean": float(np.mean(X_chroma_norm)),
-                    "chroma_norm_std": float(np.std(X_chroma_norm)),
-                    "mspec_norm_mean": float(np.mean(X_mspec_norm)),
-                    "mspec_norm_std": float(np.std(X_mspec_norm))
-                }
-            
-            print("Segment prediction completed successfully")
-            return result
-            
-        except Exception as e:
-            print(f"SEGMENT PROCESSING ERROR: {str(e)}")
-            import traceback
-            traceback.print_exc()
-            raise
+        
+        return result
 
 # FastAPI app
 app = FastAPI(title="Heart Sound Classifier API", version="1.0.0")