Update app

app.py

@@ -1322,18 +1322,18 @@ def process_video_syncoffset(video_path, num_avg_frames, apply_preprocess):
 			msg = "Error: The input video is too short. Please use a longer input video."
 			return None, msg
 
-		if apply_preprocess:
-			# Load keypoints and check if gestures are visible
-			kp_dict, status = get_keypoints(frames)
-			if status != "success":
-				return None, status
-			print("Successfully extracted the keypoints: ", len(kp_dict), len(kp_dict["kps"]))
+		# if apply_preprocess:
+		# Load keypoints and check if gestures are visible
+		kp_dict, status = get_keypoints(frames)
+		if status != "success":
+			return None, status
+		print("Successfully extracted the keypoints: ", len(kp_dict), len(kp_dict["kps"]))
 
-			status = check_visible_gestures(kp_dict)
-			if status != "success":
-				return None, status
-		else:
-			kp_dict = None
+		status = check_visible_gestures(kp_dict)
+		if status != "success":
+			return None, status
+		# else:
+		# 	kp_dict = None
 
 		# Load RGB frames
 		rgb_frames, num_frames, orig_masked_frames, status = load_rgb_masked_frames(frames, kp_dict, asd=False, window_frames=25, width=480, height=270)
@@ -1520,11 +1520,8 @@ def process_video_activespeaker(video_path, global_speaker, num_avg_frames):
 
 					if idx==0:
 						video_emb, audio_emb = get_embeddings(video_sequences, audio_sequences, model, asd=True, calc_aud_emb=True)
-						video_emb = torch.cat(video_emb, dim=0)
-						audio_emb = torch.cat(audio_emb, dim=0)
 					else:
 						video_emb = get_embeddings(video_sequences, audio_sequences, model, asd=True, calc_aud_emb=False)
-						video_emb = torch.cat(video_emb, dim=0)
 					all_video_embs.append(video_emb)
 			print("Successfully extracted GestSync embeddings")