theEricMa / OTAvatar

This is the official repository for OTAvatar: One-shot Talking Face Avatar with Controllable Tri-plane Rendering [CVPR2023].
306 stars 37 forks source link

Difference between code and paper #12

Closed ldz666666 closed 1 year ago

ldz666666 commented 1 year ago

Hi, thanks for your great work ! In line 730 of decouple_by_invert.py, the parameter of the motion controller is updated together with the eg3d generator. However, in the algorithm of the appendix, the parameters of the motion controller is not updated while finetuning theta_eg

截屏2023-05-15 16 50 00

By the way, in the 16th line of the algorithm, Lt is not mentioned in the original paper, is there a mistake?

theEricMa commented 1 year ago

Thanks for reminding us the typo! We did update \theta_c$ while finetuning $theta_eg$, and the Lt should be Ld. Please be subject to the code, it shows what we want to express in the paper.