taconite / IntrinsicAvatar

[CVPR 2024] IntrinsicAvatar: Physically Based Inverse Rendering of Dynamic Humans from Monocular Videos via Explicit Ray Tracing
https://neuralbodies.github.io/IntrinsicAvatar/
MIT License
87 stars 5 forks source link

Novel Poses Process #4

Closed zhao-yiqun closed 6 months ago

zhao-yiqun commented 6 months ago

Hi

Thanks for your amazing project! I want to render the novel poses beyond aist_demo.npz like other actions such as kick, punch(from AMASS dataset) etc as in your project page IntrinsicAvatar. I wonder how should I process the poses from AMASS to make it suitable for the dataloader you provided datasets/animation.py. Could you please share your pose processing scripts?

Thanks in advance.

Bests

taconite commented 6 months ago

Hi, Yiqun

Thanks for your interest in our work :)

Actually, the kick and punch poses in our video are from the CAPE dataset. I've updated my preprocessing script for CAPE poses to the repo, though in order to work with AMASS you may need to remove hand poses of AMASS since it uses SMPL+H model.

Hope this helps.

Best

zhao-yiqun commented 6 months ago

Thanks fot the sharing!