search

yuval-alaluf / hyperstyle

Official Implementation for "HyperStyle: StyleGAN Inversion with HyperNetworks for Real Image Editing" (CVPR 2022) https://arxiv.org/abs/2111.15666
https://yuval-alaluf.github.io/hyperstyle/
MIT License
1.01k stars 115 forks source link

About the Faces W-Encoder #71

Closed shenhaiyoualn closed 1 year ago

shenhaiyoualn commented 1 year ago

I train a e4e model on my own dataset, because i want to replace the [Faces W-Encoder] with it. But i find the model does not match, I want to know how you got the [Faces W-Encoder]？ Thanks！！

shenhaiyoualn commented 1 year ago

Perhaps I should ask how to obtain an initial inversion into the W latent space.

yuval-alaluf commented 1 year ago

The initial inversion into the W latent space was done by training the WEncoder using the e4e codebase. We simply change the encoder to use our WEncoder class and change the number of latents from 18 to 1. I believe everything else was kept as default.