fix

services/extractor.py

@@ -47,6 +47,8 @@ def extract_frames(video_path):
     indices = []
     total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
     # total_frames = 300 # TEMP
+    print(f"Total frames in video: {total_frames}")
+    print(f"Extracting frames at every {FRAME_RATE} frames...")
 
     for idx in tqdm(range(0, total_frames, FRAME_RATE)):
         cap.set(cv2.CAP_PROP_POS_FRAMES, idx)
@@ -56,11 +58,16 @@ def extract_frames(video_path):
         frames.append(Image.fromarray(frame))
         indices.append(idx)
 
+    print(f"Indices of extracted frames: {indices}")
+    print(f"Total frames extracted: {len(frames)}")
+
     cap.release()
     return frames, indices
 
 def extract_features(frames):
     features = [transform(frame) for frame in frames]
     features = torch.stack(features).to(DEVICE)
+    print("Features before GoogleNet extraction:", features.shape)
     features = feature_extractor(features)
+    print("Features after GoogleNet extraction:", features.shape)
     return features

services/summarizer.py

@@ -13,10 +13,16 @@ def get_scores(features):
     # features.dtype: torch.float32
     # features.device: cpu
     with torch.no_grad():
+        print("Features before model inference:", features.shape)
         scores, _ = model(features)
-    return scores.squeeze().cpu().numpy()
+        scores = scores.squeeze().cpu().numpy()
+        print("Features after model inference:", features.shape)
+    return scores
 
 def get_selected_indices(scores, picks, threshold=SCORE_THRESHOLD):
+    print("Threshold for selection:", threshold)
+    print("Scores:", scores.shape, scores)
+    print("Picks:", picks.shape, picks)
     return [picks[i] for i, score in enumerate(scores) if score >= threshold]
 
 import subprocess