Yining Li

We will close this PR for now. Please feel free to reopen it if there are any updates or questions.

Not sure if paddle tinypose supports multiple box-id HPE. In MMPose, maybe you can define the keypoint label as the union of human keypoints and kangaroo keypoints and for each...

Thanks for your help! I made a tiny change to add back the '.' before the extension.

I think the default arg is `m4a`: https://github.com/open-mmlab/mmaction2/blob/f189c04f4392c9acb79c596e6567a2ed296bc041/tools/data/build_audio_features.py#L293

Sorry for the late reply. `forward_dummy` simply performs network forward while `inference_detector` performs the whole inference pipeline including model forward and post-processings, and that is why they give different outputs....

Hi there, I hope everything is going well. Please let us know if there are any updates or if there is anything we can help with. Thank you very much!

There is no hurry :) BTW, we have pre-released [MMPose v1.0](https://github.com/open-mmlab/mmpose/tree/1.x) which involves a major refactoring and is in open beta now. Feel free to have a try if you...

Yes, the script `preprocessing_h36m.py` is used for processing Human3.6M data for VideoPose3D training. The raw data need to be downloaded from the official site.

Sorry for the late reply. Yes, this model only predicts the relative keypoint coordinates around the root.

Wow, that would be great! Thank you very much and we are looking forward to your PR.