Bug in inference or rendering

[ntujy]

Thanks for the excellent work!

I am following the guide to run inference on videos. While some of them return accurate results, some frames seem not to have results. One example frame is as follows:

The output of the side view and all view are as follows:

I tried to print the pred_keypoints_3d from the output, the coordinates are as follows:

So the model got some results but the rendering does not work. Any possible reason that could lead to this? Look forward to your reply.

[geopavlakos]

This example looks ok on my end. How are you doing inference on video? Maybe you edited the code and introduced a bug? If you use the single frame demo on this image, it should work ok.

[ntujy]

The problem is solved by re-installing the virtual environment. Close the issue. Thanks for your reply!