take every n frame

app.py

@@ -77,19 +77,21 @@ def process_video(
     # cleanup of old video files
     remove_files_older_than(RESULTS, 30)
 
+    OUTPUT_LENGTH = 4
+
     video_info = sv.VideoInfo.from_video_path(input_video)
-    video_info.fps = video_info.fps // 2
-    total = calculate_end_frame_index(input_video)
+    video_info.fps = video_info.fps // OUTPUT_LENGTH
+    total = calculate_end_frame_index(input_video, OUTPUT_LENGTH)
     frame_generator = sv.get_video_frames_generator(
         source_path=input_video,
         end=total,
-        stride=2
+        stride=OUTPUT_LENGTH
     )
     result_file_name = generate_file_name(extension="mp4")
     result_file_path = os.path.join(RESULTS, result_file_name)
     TRACKER.reset()
     with sv.VideoSink(result_file_path, video_info=video_info) as sink:
-        for _ in tqdm(range(total // 2), desc="Processing video..."):
+        for _ in tqdm(range(total // OUTPUT_LENGTH), desc="Processing video..."):
             frame = next(frame_generator)
             caption = run_captioning(
                 model=MODEL,

utils/video.py

@@ -47,11 +47,14 @@ def remove_files_older_than(directory: str, diff_minutes: int) -> None:
     )
 
 
-def calculate_end_frame_index(source_video_path: str) -> int:
+def calculate_end_frame_index(
+    source_video_path: str,
+    max_video_length_sec: int = MAX_VIDEO_LENGTH_SEC
+) -> int:
     video_info = sv.VideoInfo.from_video_path(source_video_path)
     return min(
         video_info.total_frames,
-        video_info.fps * MAX_VIDEO_LENGTH_SEC
+        video_info.fps * max_video_length_sec
     )