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
999 stars 115 forks source link

Training for different domain #67

Closed zhihongp closed 1 year ago

zhihongp commented 1 year ago

It is a follow up question to #53. To train one for the whole AFHQ dataset (not only AFHQ wild like in pre-trained), it should start from training a domain-specific StyleGAN first, then an e4e WEncoder, and lastly your HyperStyle model, right?

yuval-alaluf commented 1 year ago

Correct :)

zhihongp commented 1 year ago

Thanks. I used pre-trained one from NVIDIA so that helps