search

ZPdesu / Barbershop

Barbershop: GAN-based Image Compositing using Segmentation Masks (SIGGRAPH Asia 2021)
https://zpdesu.github.io/Barbershop/
MIT License
1.32k stars 264 forks source link

re-train stylegan2 #25

Open wftang2020 opened 2 years ago

wftang2020 commented 2 years ago

We found that the your stylegan2 network parameter name is inconsistent with the original stylegan2. If we change the network structure of the original Stylegan2 to be the same as yours, and the other functions remain the same as the original Stylegan2. Is it possible to map the trained parameters as the inference· parameters of your Stylegan2 network? thanks for your reply

fingerella2000 commented 1 year ago

Curious how much time needed to re-train and how much time needed to run main.py and get a result image? Thanks.