Test on lengths

app.py

@@ -162,9 +162,9 @@ def extract_resampled_audio(video_path, target_sr=16000):
     torchaudio.save(resampled_audio_path, waveform, sample_rate=target_sr)
     return resampled_audio_path
 
-@spaces.GPU
-def yolo_detection(frame, verbose=False):
-    return model(frame, verbose=verbose)[0]
+#@spaces.GPU
+#def yolo_detection(frame, verbose=False):
+#    return model(frame, verbose=verbose)[0]
 
 @spaces.GPU
 def extract_faces(video_file):
@@ -178,8 +178,8 @@ def extract_faces(video_file):
             break
 
         # Inference
-        #results = model(frame, verbose=False)[0]
-        results = yolo_detection(frame, verbose=False)
+        results = model(frame, verbose=False)[0]
+        #results = yolo_detection(frame, verbose=False)
         for box in results.boxes:
             # version 1
             # x1, y1, x2, y2 = map(int, box.xyxy[0])