bryandlee/FreezeG
A transfer learning approach that freezes early generator layers for pseudo image-to-image translation using StyleGAN2.
Not currently ranked — collecting fresh signals.
star history
FreezeG applies transfer learning by freezing the early layers of a pre-trained generator to reuse high-level features for image-to-image translation. The method projects input images into the learned latent space then propagates them through the generator to produce target images. It demonstrates style manipulation and domain adaptation on datasets like AFHQ, FFHQ, and various face-to-art translation tasks.
Frequently asked
- What is bryandlee/FreezeG?
- A transfer learning approach that freezes early generator layers for pseudo image-to-image translation using StyleGAN2.
- Is FreezeG open source?
- Yes — bryandlee/FreezeG is open source, released under the MIT license.
- What language is FreezeG written in?
- bryandlee/FreezeG is primarily written in Python.
- How popular is FreezeG?
- bryandlee/FreezeG has 473 stars on GitHub.
- Where can I find FreezeG?
- bryandlee/FreezeG is on GitHub at https://github.com/bryandlee/FreezeG.