search

genforce / idinvert_pytorch

[ECCV 2020] In-Domain GAN Inversion for Real Image Editing (PyTorch code)
https://genforce.github.io/idinvert/
MIT License
407 stars 54 forks source link

Convert pretrained encoders trained using TF repository of idinvert. #3

Closed Caselles closed 3 years ago

Caselles commented 3 years ago

Hello,

First of all, thanks for releasing the PyTorch version of the code for your paper. It's pretty useful !

I have several questions :

1) regarding the pretrained encoder provided in the repo : is it the same model as the pretrained encoder provided in the TF repository ? If yes, do you have the code needed for converting the TF model in PyTorch ? It seems that you are using the same architecture of models as in InterfaceGAN, which provided code for TF->PyTorch conversion (for stylegan generator for instance). Having the code for converting encoders would be nice for training other encoders in TF and then doing inference in PyTorch.

2) are you considering releasing code for training encoders also in PyTorch ?

Thank you so much again for all the work you put in (paper and code release).

ShenYujun commented 3 years ago

Yes, we are currently planning to release a full-function PyTorch codebase. It should be released in around 1-1.5 months. Please be patient! Thanks for liking our work ;)

Caselles commented 3 years ago

Thank you very much!

wataoka commented 3 years ago

I'm also waiting for the release. Thank you for your work!