search

tianxintao / Online-Motion-Style-Transfer

Code for the CVPR 2022 Paper - Style-ERD: Responsive and Coherent Online Motion Style Transfer
https://tianxintao.github.io/Online-Motion-Style-Transfer/
41 stars 12 forks source link

using of FID #2

Closed zzgg1 closed 2 years ago

zzgg1 commented 2 years ago

Thanks for your amazing code! I am wondering if it could be possible that you provide pretrained model for FID, and elaborate on the detailed training procedures of FID

tianxintao commented 2 years ago

Most of the training code should be available here: fid.py Basically, a denoising autoencoder is first trained with reconstruction loss, which is then used as features for FID. The computation of FID follows the standard FID formula.