A curated list of resources related to training of GANs
GAN Variants
Year | Conf | Code | Title |
---|---|---|---|
2014 | NIPS | Pt+Tf | Vanilla-GAN - Generative Adversarial Networks |
2017 | ICCV | Pt+Tf | LS-GAN - Least Squares Generative Adversarial Networks |
2017 | ICML | Pt+Tf | WGAN - Wasserstein GAN |
2017 | NIPS | Pt+Tf | WGAN-GP - Improved Training of Wasserstein GANs |
2019 | ICLR | Pt | The relativistic discriminator: a key element missing from standard gan |
Normalization, Network Architecture
Performance analysis and comparisons
Misc
Hacks
-
2016 - Deconvolution and Checkerboard Artifacts. Link
-
2018-ECCVW - ESRGAN: Enhanced Super-Resolution Generative Adversarial Networks. [Paper]
- First work to apply relativistic discriminator for image super-resolution.
- Batch normalization (BN) can generate artifacts when the generator goes deeper or there is an extra BN layer in HR (high-resolution) space.
-
How generative adversarial networks and their variants work: An overview
Other Resources