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
3.43k stars 344 forks source link

Include ECGAN into StudioGAN #107

Open sian-chen opened 3 years ago

sian-chen commented 3 years ago

Dear Authors,

Thank you for your great work. We are happy to share with you that our paper A Unified View of cGANs with and without Classifiers is accepted by NeurIPS 2021. We develop a new cGAN architecture called ECGAN, which achieves state-of-the-art results compared with current cGANs. Our implementation is based on StudioGAN, so it can be easily merged into your repository. We appreciate your great work that allows us to develop new architecture much easier. It would be my honor if ECGAN can be included in StudioGAN. If you are interested, I can prepare a pull request. What do you think?

Best Regards, Si-An

mingukkang commented 3 years ago

Hi Si-An,

Congratulations on your great work being published in Neurips 2021.

I saw your paper (ECGAN) published in arXiv yesterday, and I am so happy since I remember you emailed me to ask questions regarding ContraGAN (publishing paper in arXiv or Neurips means you achieve significant improvement over baselines).

It is also a great honor for us if you make a pull request to include your wonderful research output into the StudioGAN library.

Unfortunately, StudioGAN has undergone a big version change, and almost all codes are different from the previous implementations. Thus, if you are willing to add your implementation into StudioGAN, we are welcome your contribution.

Thank you.

Best,

Minguk

sian-chen commented 3 years ago

So glad to know that you remember me. I see the differences. It may take some time but I'm still willing to contribute to the library. I'll let you know once the PR is ready. Thanks!

mingukkang commented 3 years ago

Thank you so much.

I hope to see you at an offline conference someday.