lucidrains / big-sleep

A simple command line tool for text to image generation, using OpenAI's CLIP and a BigGAN. Technique was originally created by https://twitter.com/advadnoun
MIT License
2.57k stars 305 forks source link

Implementing StyleGAN2 #6

Open ShivanshuPurohit opened 3 years ago

ShivanshuPurohit commented 3 years ago

How should I go about to replace BigGAN with your StyleGAN2 package?

lucidrains commented 3 years ago

what is needed is pretrained weights on a diverse set of 512x512 images (maybe Imagenet) for pytorch. If you can find that, it is doable!

molo32 commented 3 years ago

how to use with the stylegan2-ffhq-config-f.pkl? just download the pkl and replace it? if so, what line should I change to change biggan with stylegan2-ffhq-config-f.pkl?

walmsley commented 3 years ago

Found this ImageNet-trained StyleGAN2, but unfortunately not for PyTorch: https://battle.shawwn.com/sdc/stylegan2-imagenet-512/model.ckpt-533504.pkl

Found via this tweet: https://twitter.com/theshawwn/status/1244081581347598341

Seems to be created using this NVIDIA-based TensorFlow repo: https://github.com/shawwn/stylegan2/tree/working

illtellyoulater commented 2 years ago

any news on this?