Awesome Pretrained StyleGAN3
A collection of pretrained models for StyleGAN3 (and some links to encoding methods).
If you'd like to share a model please submit an issue.
For StyleGAN2 models please see: https://github.com/justinpinkney/awesome-pretrained-stylegan2
WikiArt 1024
- Name: WikiArt-1024
- Download URL: https://drive.google.com/file/d/18MOpwTMJsl_Z17q-wQVnaRLCUFZYSNkj/view?usp=sharing
- Author: Justin Pinkney/LambdaLabs
- Author URL: https://www.justinpinkney.com/ https://lambdalabs.com/
- Dataset: WikiArt
- Source URL: https://lambdalabs.com/blog/stylegan-3/
- Resolution: 1024x1024
- Config: T
- Notes: FID=8.1, trained 17.1 Mimgs, gamma=32, mapping layers=2
Landscapes 256
- Name: LHQ-256
- Download URL: https://drive.google.com/file/d/14UGDDOusZ9TMb-pOrF0PAjMGVWLSAii1/view?usp=sharing
- Author: Justin Pinkney/LambdaLabs
- Author URL: https://www.justinpinkney.com/ https://lambdalabs.com/
- Dataset: LHQ
- Source URL: https://twitter.com/Buntworthy/status/1460980442409185287?s=20
- Resolution: 256x256
- Config: T
- Notes: FID=2.31, trained 25 Mimgs, gamma=2, mapping layers=2
Various Wombo Dream models
- Name: Lucid Layers Stylegan3 Datasets and Models collection
- Source URL: https://github.com/edstoica/lucid_stylegan3_datasets_models/blob/main/README.md
- Resolution: 256-512
- Config: T
- Notes: default settings, no mirror
Benches 512
- Name: Memorial Benches
- Source URL: https://shkspr.mobi/blog/2022/04/this-bench-does-not-exist/
- Resolution: 512
- Config: ?
- Notes: none