cGANs with Projection Discriminator

TL;DR

This paper introduces a projection-based conditional discriminator for GANs, significantly improving class-conditional image generation quality and enabling advanced applications like super-resolution and category transformation.

Contribution

It proposes a novel projection discriminator that better incorporates conditional information, outperforming existing methods on ImageNet and extending to super-resolution tasks.

Findings

Achieved state-of-the-art class-conditional image generation on ImageNet.

Successfully applied to super-resolution, producing highly discriminative images.

Enabled high-quality category transformation via parametric conditional batch normalization.

Abstract

We propose a novel, projection based way to incorporate the conditional information into the discriminator of GANs that respects the role of the conditional information in the underlining probabilistic model. This approach is in contrast with most frameworks of conditional GANs used in application today, which use the conditional information by concatenating the (embedded) conditional vector to the feature vectors. With this modification, we were able to significantly improve the quality of the class conditional image generation on ILSVRC2012 (ImageNet) 1000-class image dataset from the current state-of-the-art result, and we achieved this with a single pair of a discriminator and a generator. We were also able to extend the application to super-resolution and succeeded in producing highly discriminative super-resolution images. This new structure also enabled high quality category transformation based on parametric functional transformation of conditional batch normalization layers in the generator.