Toward responsible face datasets: modeling the distribution of a disentangled latent space for sampling face images from demographic groups

TL;DR

This paper proposes a method to generate balanced, bias-free synthetic face datasets by modeling and sampling a disentangled StyleGAN latent space, aiming to improve fairness in facial recognition systems.

Contribution

It introduces a simple approach for modeling and sampling a disentangled latent space to generate diverse demographic face images, addressing dataset bias issues.

Findings

Effective synthesis of demographic-specific face images

Generated identities differ from original datasets

Source code is publicly available

Abstract

Recently, it has been exposed that some modern facial recognition systems could discriminate specific demographic groups and may lead to unfair attention with respect to various facial attributes such as gender and origin. The main reason are the biases inside datasets, unbalanced demographics, used to train theses models. Unfortunately, collecting a large-scale balanced dataset with respect to various demographics is impracticable. In this paper, we investigate as an alternative the generation of a balanced and possibly bias-free synthetic dataset that could be used to train, to regularize or to evaluate deep learning-based facial recognition models. We propose to use a simple method for modeling and sampling a disentangled projection of a StyleGAN latent space to generate any combination of demographic groups (e.g. $hispanic-female$). Our experiments show that we can synthesis any combination of demographic groups effectively and the identities are different from the original training dataset. We also released the source code.