fix

app/main.py

@@ -27,16 +27,16 @@ default_sample_rate=22050
 def load(file_name, skip_seconds=0):
     return librosa.load(file_name, sr=None, res_type='kaiser_fast')
 
-# def preprocess_audio(audio_data, rate):
-#     # Apply preprocessing steps
-#     audio_data = nr.reduce_noise(y=audio_data, sr=rate)
-#     audio_data = librosa.util.normalize(audio_data)
-#     audio_data, _ = librosa.effects.trim(audio_data)
-#     audio_data = librosa.resample(audio_data, orig_sr=rate, target_sr=default_sample_rate)
-# #     audio_data = fix_length(audio_data)
-#     rate = default_sample_rate
+def preprocess_audio(audio_data, rate):
+    # Apply preprocessing steps
+    audio_data = nr.reduce_noise(y=audio_data, sr=rate)
+    audio_data = librosa.util.normalize(audio_data)
+    audio_data, _ = librosa.effects.trim(audio_data)
+    audio_data = librosa.resample(audio_data, orig_sr=rate, target_sr=default_sample_rate)
+#     audio_data = fix_length(audio_data)
+    rate = default_sample_rate
 
-#     return audio_data, rate
+    return audio_data, rate
 
 def extract_features(X, sample_rate):
     # Generate Mel-frequency cepstral coefficients (MFCCs) from a time series
@@ -187,22 +187,22 @@ pca = joblib.load('app/pca.pkl')
 scaler = joblib.load('app/1713696947.894978_scaler.joblib')
 label_encoder = joblib.load('app/1713696954.9487948_label_encoder.joblib')
 
-def preprocess_audio(audio_data, rate):
-    # Resample first if the target rate is lower to reduce data size for subsequent operations
-    if rate > default_sample_rate:
-        audio_data = librosa.resample(audio_data, orig_sr=rate, target_sr=default_sample_rate)
-        rate = default_sample_rate
+# def preprocess_audio(audio_data, rate):
+#     # Resample first if the target rate is lower to reduce data size for subsequent operations
+#     if rate > default_sample_rate:
+#         audio_data = librosa.resample(audio_data, orig_sr=rate, target_sr=default_sample_rate)
+#         rate = default_sample_rate
 
-    # Trim silence before applying computationally expensive noise reduction
-    audio_data, _ = librosa.effects.trim(audio_data)
+#     # Trim silence before applying computationally expensive noise reduction
+#     audio_data, _ = librosa.effects.trim(audio_data)
 
-    # Normalize the audio data
-    audio_data = librosa.util.normalize(audio_data)
+#     # Normalize the audio data
+#     audio_data = librosa.util.normalize(audio_data)
 
-    # Apply noise reduction
-    audio_data = nr.reduce_noise(y=audio_data, sr=rate)
+#     # Apply noise reduction
+#     audio_data = nr.reduce_noise(y=audio_data, sr=rate)
 
-    return audio_data, rate
+#     return audio_data, rate
 
 # def preprocess_audio(audio_data, rate):
 #     audio_data = nr.reduce_noise(y=audio_data, sr=rate)