Add Files

alm_pipeline.py

@@ -26,8 +26,9 @@ def speech_to_text(audio):
 
 
 def detect_sound(audio):
-    waveform, sr = librosa.load(audio, sr=16000)
-    waveform = waveform.reshape(1, -1)
+    # Load mono waveform at 16 kHz as 1D float32 array, as expected by YAMNet
+    waveform, sr = librosa.load(audio, sr=16000, mono=True)
+    waveform = waveform.astype(np.float32)
     scores, embeddings, _ = yamnet(waveform)
     mean_scores = np.mean(scores.numpy(), axis=0)
     top_idx = np.argmax(mean_scores)