Distill and De-bias: Mitigating Bias in Face Verification using Knowledge Distillation
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:Q43CMMPQrecord.jsonopen to challenge →
read the original abstract
Face recognition networks generally demonstrate bias with respect to sensitive attributes like gender, skintone etc. For gender and skintone, we observe that the regions of the face that a network attends to vary by the category of an attribute. This might contribute to bias. Building on this intuition, we propose a novel distillation-based approach called Distill and De-bias (D&D) to enforce a network to attend to similar face regions, irrespective of the attribute category. In D&D, we train a teacher network on images from one category of an attribute; e.g. light skintone. Then distilling information from the teacher, we train a student network on images of the remaining category; e.g., dark skintone. A feature-level distillation loss constrains the student network to generate teacher-like representations. This allows the student network to attend to similar face regions for all attribute categories and enables it to reduce bias. We also propose a second distillation step on top of D&D, called D&D++. Here, we distill the `un-biasedness' of the D&D network into a new student network, the D&D++ network, while training this new network on all attribute categories; e.g., both light and dark skintones. This helps us train a network that is less biased for an attribute, while obtaining higher face verification performance than D&D. We show that D&D++ outperforms existing baselines in reducing gender and skintone bias on the IJB-C dataset, while obtaining higher face verification performance than existing adversarial de-biasing methods. We evaluate the effectiveness of our proposed methods on two state-of-the-art face recognition networks: ArcFace and Crystalface.
This paper has not been read by Pith yet.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.