search

eth-siplab / AvatarPoser

Official Code for ECCV 2022 paper "AvatarPoser: Articulated Full-Body Pose Tracking from Sparse Motion Sensing"
https://siplab.org/projects/AvatarPoser
MIT License
277 stars 46 forks source link

Evaluation protocol (clarification) #18

Open sadegh-aa opened 1 year ago

sadegh-aa commented 1 year ago

Hi,

I'd like to know how the body shape (the beta parameters) is considered in the evaluations? For the quantitative results in the paper, does the model uses default shape parameters (torch.zeros(B, num_shape)) or does it utilize the GT beta values?

Thank you very much