This is a WGAN-gp version of the LOGAN which I used a modified version of it for my work. It was really improving the result of my WGAN-gp, so I decided to implement it based on your WGAN-gp and share it. It is a very new method and easy to use. I hope it will be helpful for others. The original paper used it over BigGAN and CS-GAN. My next objective (also in my work) is to implement LOGAN over BigGAN which would be a great compliment to your awesome GAN PyTorch repository. Moreover, the default hyperparameters are checked for several datasets, they are fine.
Cheers,
1912.00953.pdf
This is a WGAN-gp version of the LOGAN which I used a modified version of it for my work. It was really improving the result of my WGAN-gp, so I decided to implement it based on your WGAN-gp and share it. It is a very new method and easy to use. I hope it will be helpful for others. The original paper used it over BigGAN and CS-GAN. My next objective (also in my work) is to implement LOGAN over BigGAN which would be a great compliment to your awesome GAN PyTorch repository. Moreover, the default hyperparameters are checked for several datasets, they are fine. Cheers, 1912.00953.pdf