how to inferencce on video？

[FRunyang]

No description provided.

[FRunyang]

hello，I want to inference my videos， what should i do？

[FRunyang]

Oh， I do。我做好了，我发现有的帧腿就没了

[FRunyang]

[hellojialee]

The image sample you tested above seems easy, what scale configurations did you use?
For single scale inference, 768 length is better than 512:
https://github.com/jialee93/Improved-Body-Parts/blob/316e71fa93e1dc444b1cfd4fc312c21c13bfe93f/utils/config#L36

And lower threshold thre1=thre2=0.08 is better for more keypoint detection if the background is not very complexed:
https://github.com/jialee93/Improved-Body-Parts/blob/316e71fa93e1dc444b1cfd4fc312c21c13bfe93f/utils/config#L16

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.