vighneshb

3 ms·

1. 1.▲
   by vighneshb 2y ago

   This is very interesting. Unlike us (who focus on the decoder) they focus on changing the representation itself so that they can achieve better generation. Thanks for the link.
2. 2.▲
   Sample what you can't compress; image auto-encoders wihtout GANs(arxiv.org)

   19 points by vighneshb 2y ago|4 comments
3. 3.▲
   by vighneshb 2y ago

   Hi In our latest paper we shoa that a GAN loss (used by almost all latent diffusion models) to train their autoencoders is not required and instead can be replaced with a diffusion loss. Our auto-encoder is trained end-to-end and achieves h