search

nghorbani / human_body_prior

VPoser: Variational Human Pose Prior
https://smpl-x.is.tue.mpg.de/
Other
798 stars 145 forks source link

Searching for 'github_data/dmpl_sample.npz' #57

Open skore11 opened 2 years ago

skore11 commented 2 years ago

Hello, I am working on bringing DMPL to a real time collision like seen in Unity. Are the contents of the dmpl_sample.npz file from the DMPL visualization .ipynb notebook the Dyna registrations? Meaning you need the eig vectors of the errors between the vertices of SMPL and Dyna correct? Also I did not see much of a difference between the animation with DMPLs and without. Is this because the number of DMPL blendshape principal components provided by SMPL are not enough? Keep up the good work! Thanks Abishek

MichaelJBlack commented 2 years ago

Are you getting the SMPL/DMPL parameters from AMASS? I only recently realized that the released AMASS fits are fits to virtual markers using MoSh and not fits to the scan data. The former seems not very good, while the latter is very detailed. I'm looking into it.

skore11 commented 2 years ago

Yessir I am using AMASS, the backend human body prior used the dmpls as well. so MoSH's extraction of DMPLs has 8 eig vectors (even with scaling) which is not that detailed. perhaps 8 eig vecs are not enough? Yes I believe Dyna would be far more detailed, do you happen to the DMPL extraction from Dyna as an .npz file?

MichaelJBlack commented 2 years ago

Right. With sparse markers, you can't get much soft tissue detail, hence, we use few DMPLs. I'm working on getting the more detailed Dyna/DFAUST fits on the AMASS site.

skore11 commented 2 years ago

That sounds good, Ill keep a lookout for the same, on a side note do you know if DMPLs will be included in the Unity build? I am trying to include the same with position based dynamics to integrate collision based blendshapes with SMPL and DMPL. I guess the big issue would be to redo lbs with the DMPL blendshapes.