Fairer Analysis and Demographically Balanced Face Generation for Fairer Face Verification

TL;DR

This paper presents a new face generation pipeline that enhances fairness in face verification tasks by reducing biases and improving demographic balance, using advanced statistical analysis and the DCFace framework.

Contribution

It introduces a controlled generation pipeline that outperforms existing bias mitigation methods in fairness, with minimal impact on accuracy, based on comprehensive statistical evaluation.

Findings

Improved fairness metrics over baseline methods

Slight increase in raw face verification performance

Enhanced demographic balance in generated faces

Abstract

Face recognition and verification are two computer vision tasks whose performances have advanced with the introduction of deep representations. However, ethical, legal, and technical challenges due to the sensitive nature of face data and biases in real-world training datasets hinder their development. Generative AI addresses privacy by creating fictitious identities, but fairness problems remain. Using the existing DCFace SOTA framework, we introduce a new controlled generation pipeline that improves fairness. Through classical fairness metrics and a proposed in-depth statistical analysis based on logit models and ANOVA, we show that our generation pipeline improves fairness more than other bias mitigation approaches while slightly improving raw performance.