I'm betting on a combination of this ICR technique with latent optimization (from deepmind's S.O.T.A LOGAN rejection at ICLR... can't believe it got rejected lol) to drop in the next few weeks. You heard it here first :)
Lol, it's definitely interesting to think of papers that fly under the radar due to which conferences they are published in. I saw a link to LOGAN while reading this paper, I think they hold the FID crown with the BigGAN-Deep model! Excited to look more into it!
1:25 Examples of Images
1:55 Consistency Regularization
2:52 Prior Knowledge in CR
3:22 CR-GAN
3:50 CR-GAN Artifacts
5:13 Balanced Consistency Regularization
5:55 Latent Consistency Regularization
7:00 StyleGAN2 Interpolation
7:50 Conditional GANs
8:58 Results
Good job
I'm betting on a combination of this ICR technique with latent optimization (from deepmind's S.O.T.A LOGAN rejection at ICLR... can't believe it got rejected lol) to drop in the next few weeks. You heard it here first :)
Lol, it's definitely interesting to think of papers that fly under the radar due to which conferences they are published in. I saw a link to LOGAN while reading this paper, I think they hold the FID crown with the BigGAN-Deep model! Excited to look more into it!