FairStyle: Debiasing StyleGAN2 with Style Channel Manipulations

TL;DR

FairStyle introduces a quick and effective method to debias StyleGAN2 models by manipulating style channels, enabling the generation of more balanced and fair images without retraining the entire model.

Contribution

It presents a novel, model-agnostic approach to debiasing StyleGAN2 through style channel manipulations, avoiding additional training and maintaining image quality.

Findings

Successfully debiases GANs within minutes

Maintains high image quality after debiasing

Applicable to multiple attributes simultaneously

Abstract

Recent advances in generative adversarial networks have shown that it is possible to generate high-resolution and hyperrealistic images. However, the images produced by GANs are only as fair and representative as the datasets on which they are trained. In this paper, we propose a method for directly modifying a pre-trained StyleGAN2 model that can be used to generate a balanced set of images with respect to one (e.g., eyeglasses) or more attributes (e.g., gender and eyeglasses). Our method takes advantage of the style space of the StyleGAN2 model to perform disentangled control of the target attributes to be debiased. Our method does not require training additional models and directly debiases the GAN model, paving the way for its use in various downstream applications. Our experiments show that our method successfully debiases the GAN model within a few minutes without compromising the quality of the generated images. To promote fair generative models, we share the code and debiased models at http://catlab-team.github.io/fairstyle.