Leon Li

so, How to train on multiple gpus?

Hi, When do you plan to release the annotation preparation script for self-supervised training? Thank you

Only 3D datasets, i.e., h36m has the 3D keypoint gt, which are available to generate cam_trans gt. I observed that the root position is extremely stable in the demo video....

converge not well when training using BEAT dataset

1

converge not well when training using BEAT dataset. May I ask what could be the reason?

新手疑惑

1

为什么不直接使用Android SDK中的MediaRecorder？这种实现方式有什么优势？

about position embedding

1

How does token merging process position embedding? I am wondering whether token merging can adapt relative position embedding like RoPE? Look forward to ur reply