Frequency Matters: Explaining Biases of Face Recognition in the Frequency Domain

TL;DR

This paper investigates how frequency domain features influence face recognition biases across demographic groups, revealing that different frequencies are crucial depending on ethnicity, which helps explain performance disparities.

Contribution

It introduces frequency-based analysis to explain demographic biases in face recognition models, highlighting the importance of frequency components in ethnicity-dependent performance.

Findings

Different frequencies are important for face recognition depending on ethnicity.

Frequency patterns significantly influence model biases.

Frequency-based explanations can help mitigate demographic disparities.

Abstract

Face recognition (FR) models are vulnerable to performance variations across demographic groups. The causes for these performance differences are unclear due to the highly complex deep learning-based structure of face recognition models. Several works aimed at exploring possible roots of gender and ethnicity bias, identifying semantic reasons such as hairstyle, make-up, or facial hair as possible sources. Motivated by recent discoveries of the importance of frequency patterns in convolutional neural networks, we explain bias in face recognition using state-of-the-art frequency-based explanations. Our extensive results show that different frequencies are important to FR models depending on the ethnicity of the samples.