search

POSTECH-CVLab / PyTorch-StudioGAN

StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.
https://github.com/MINGUKKANG
Other
3.41k stars 339 forks source link

[CVPR 2023] Adding NoisyTwins Regularization to StyleGAN2 #188

Open rangwani-harsh opened 1 year ago

rangwani-harsh commented 1 year ago

Hi MingKug and team,

Thanks for your great work in maintaining the repository. We just released our new work NoisyTwins [CVPR 2023], a regularizer for the latent space of GAN. The regularizer improves performance on GAN in terms of class consistency and diversity. As we implemented our work in StudioGAN, we wanted to check if it could be integrated into StudioGAN. The details about the work are presented in the following links:

Paper: https://arxiv.org/abs/2304.05866 Project: https://rangwani-harsh.github.io/NoisyTwins/ Code: https://github.com/val-iisc/NoisyTwins

Thanks for your fantastic contribution.

Best, Harsh

rangwani-harsh commented 1 year ago

We will submit the pull request with the required code changes in case you all agree for it to be integrated in StudioGAN.

alex4727 commented 1 year ago

We welcome new implementations that contribute to the GAN community. After carefully reviewing changes with Minguk, we will reflect the pull request unless it requires big changes to fundamental structure of StudioGAN. Thanks for using our StudioGAN and your contribution!