Update

app.py

@@ -160,16 +160,16 @@ def extract_images_and_embeds(video_id, video_path, subtitles, output, expanded=
                 'frame_no': frame_no
             })
 
+            encoding = processor(image, text, return_tensors="pt").to(device)
+            encoding['text'] = text
+            encoding['image_filepath'] = img_fpath
+            encoding['start_time'] = caption.start
+            
+            batch_list.append(encoding)
+
         else:
             break
         
-        encoding = processor(image, text, return_tensors="pt").to(device)
-        encoding['text'] = text
-        encoding['image_filepath'] = img_fpath
-        encoding['start_time'] = caption.start
-        
-        batch_list.append(encoding)
-                
         if len(batch_list) == batch_size:
             batch = collate_fn(batch_list)
             with torch.no_grad():