The GAN is dead; long live the GAN! A Modern GAN Baseline
This provides a more stable and simplified baseline for GAN training, benefiting researchers and practitioners in generative modeling.
The authors tackled the perception that GANs are difficult to train by developing a principled regularized relativistic GAN loss that addresses mode dropping and non-convergence issues, resulting in R3GAN which surpasses StyleGAN2 and competes with state-of-the-art models across multiple datasets.
There is a widely-spread claim that GANs are difficult to train, and GAN architectures in the literature are littered with empirical tricks. We provide evidence against this claim and build a modern GAN baseline in a more principled manner. First, we derive a well-behaved regularized relativistic GAN loss that addresses issues of mode dropping and non-convergence that were previously tackled via a bag of ad-hoc tricks. We analyze our loss mathematically and prove that it admits local convergence guarantees, unlike most existing relativistic losses. Second, our new loss allows us to discard all ad-hoc tricks and replace outdated backbones used in common GANs with modern architectures. Using StyleGAN2 as an example, we present a roadmap of simplification and modernization that results in a new minimalist baseline -- R3GAN. Despite being simple, our approach surpasses StyleGAN2 on FFHQ, ImageNet, CIFAR, and Stacked MNIST datasets, and compares favorably against state-of-the-art GANs and diffusion models.