facebookresearch / facebookresearch/SlowFast
Run demo on videos,the predicted bbox does not exactly match the person of each frame
Open
- Dominant language
- Python
- Stars
- 7.4k
- Forks
- 1.3k
- PR merge metrics
- No merged PRs in 30d
Description
Thanks for your excellent work!
When i run demo on video, I found the predicted bbox does not exactly match the person's position of each frame,It looks like the bounding box is jumping around person. I tried to read the visualization code, then i found each input clip only predicts the bbox of the middle frame,and this only one bbox is drawn on each frame of input clip.
Then i tried to modify the visualization code to predict the bounding box of each frame and display it, but I encountered a threading problem. Hope you can provide some help,thanks for your reply.
Contributor guide
Assessment
This issue has not been assessed yet.