Asymmetric Rejection Loss for Fairer Face Recognition
This addresses fairness issues in face recognition for under-represented ethnic groups, representing an incremental improvement over existing semi-supervision methods.
The paper tackles racial bias in face recognition models caused by imbalanced training datasets, proposing an Asymmetric Rejection Loss that leverages unlabeled images of under-represented ethnic groups to reduce bias, with experiments showing it outperforms state-of-the-art semi-supervision methods and improves performance on under-represented groups while maintaining performance on well-represented groups.
Face recognition performance has seen a tremendous gain in recent years, mostly due to the availability of large-scale face images dataset that can be exploited by deep neural networks to learn powerful face representations. However, recent research has shown differences in face recognition performance across different ethnic groups mostly due to the racial imbalance in the training datasets where Caucasian identities largely dominate other ethnicities. This is actually symptomatic of the under-representation of non-Caucasian ethnic groups in the celebdom from which face datasets are usually gathered, rendering the acquisition of labeled data of the under-represented groups challenging. In this paper, we propose an Asymmetric Rejection Loss, which aims at making full use of unlabeled images of those under-represented groups, to reduce the racial bias of face recognition models. We view each unlabeled image as a unique class, however as we cannot guarantee that two unlabeled samples are from a distinct class we exploit both labeled and unlabeled data in an asymmetric manner in our loss formalism. Extensive experiments show our method's strength in mitigating racial bias, outperforming state-of-the-art semi-supervision methods. Performance on the under-represented ethnicity groups increases while that on the well-represented group is nearly unchanged.