search

akanimax / BMSG-GAN

[MSG-GAN] Any body can GAN! Highly stable and robust architecture. Requires little to no hyperparameter tuning. Pytorch Implementation
MIT License
629 stars 105 forks source link

Reproduction on another dataset #23

Closed TropComplique closed 5 years ago

TropComplique commented 5 years ago

Hi! I like your paper.

Here is my version of your algorithm trained to generate dresses: https://github.com/TropComplique/multi-scale-gradient-gan.

But I made some modifications:

  1. I use self modulation (stylegan-like generator),
  2. I use R1 penalty for discriminator,
  3. I use PacGAN discriminator.
akanimax commented 5 years ago

@TropComplique! Thanks a lot for sharing your implementation :smile:. I have starred :star: your repo! Your results on the dresses dataset are nice. I find it cool that you used the styleGAN architecture with MSG-GAN.

Even I am working on something very similar ... :smile:. Here is a small teaser of what is coming ...

celebahq_towers_1 ffhq_towers_4

Will be releasing this code soon.

Cheers :beers:! @akanimax

TropComplique commented 5 years ago

The teaser looks very delightful. And thanks for your good work.