Problem with target_path parameter in process_video function.

[hoang-nam-2004]

Search before asking

• I have searched the Supervision issues and found no similar feature requests.

Question

Hi, I am trying to follow the documentation: https://supervision.roboflow.com/utils/video/#process_video

I ran the pipeline perfectly in here: https://www.kaggle.com/code/namlh2004/yolov8-video-inference

But I could not find the result video in the target_path

Can anyone have a look at this for me please ?

Thank you!

Additional

No response

[hoang-nam-2004]

Update: It just showed me a result.mp4 file, but it's an empty file.

[onuralpszr]

Hello @hoang-nam-2004 sorry late answer,

I checked your kaggle and I found that you were missing "return" I also did some small changes like you don't need to "re-init" box_annotator or add labels since you also added.

Here is my kaggle version from your original kaggle notebook : https://www.kaggle.com/code/onuralpsezer/yolov8-video-inference

Better Documentation with example

https://supervision.roboflow.com/latest/how_to/track_objects/#run-inference

I hope that helps.

[onuralpszr]

Converting to Q&A

[Shahram-73]

Empty result.mp4 almost always comes from the callback returning the wrong thing (or nothing), or returning frames with the wrong shape/dtype. process_video() expects your callback to return a valid np.ndarray frame every time.
supervision.roboflow.com
+1

Quick checklist that usually fixes it:

Make sure you return frame at the end of the callback (even if you don’t draw anything).
GitHub

Ensure the returned frame is H×W×3, dtype uint8, and same size each frame.

Use an absolute target_path and make sure the folder exists.

Minimal working example:

import os
import cv2
import supervision as sv

def callback(frame, index):
# ... your detection + drawing ...
return frame # <-- important

os.makedirs("out", exist_ok=True)
sv.process_video(
source_path="input.mp4",
target_path=os.path.abspath("out/result.mp4"),
callback=callback,
)