search

xuchen-ethz / gdna

Official code release for CVPR 2022 paper gDNA: Towards Generative Detailed Neural Avatars
MIT License
181 stars 14 forks source link

SMPL fitting on 3D scan #4

Open bjkim95 opened 2 years ago

bjkim95 commented 2 years ago

Hi, thank you for sharing a great work.

I have a question about SMPL parameters for training data. I've purchased 3D scan data from RenderPeople and I'm trying to train gDNA from scratch. How did you get SMPL parameters of RenderPeople dataset? I can think of the way getting a 3D pose from triangulation of 2D poses from rendered images of multiple views at first, and optimize SMPL parameters of which joints agree with those from the 3D pose. Is there any better way to get SMPL parameters of 3D scans?

Thank you in advance.

Xuanmeng-Zhang commented 2 years ago

@bjkim95 hello, I also want to know how to get SMPL parameters of RenderPeople dataset, have you solved this problem?

bjkim95 commented 2 years ago

@Xuanmeng-Zhang Hi, if you're using the same RenderPeople dataset the authors are providing(renderpeople.csv), their smpl parameters are given in AGORA dataset. But I have different lists, so I'm implementing the code to fit SMPL to 3D mesh, yet not completed.