FunFace: Feature Utility and Norm Estimation for Face Recognition

2026-04-29

AI Summary

This work addresses the challenge that existing face recognition methods struggle to effectively model the biometric utility of low-quality samples due to their overreliance on image perceptual quality or feature norm. To overcome this limitation, we propose FunFace, which, for the first time, explicitly incorporates biometric utility—estimated via a Certainty Ratio—into an adaptive margin-based loss function, thereby transcending conventional design constraints. By integrating AdaFace’s margin modulation mechanism, FunFace dynamically optimizes discriminative capacity for both high- and low-quality facial images. Extensive experiments demonstrate that FunFace achieves state-of-the-art performance on high-quality benchmarks and significantly outperforms current best methods on low-quality evaluation protocols.

Abstract

Face Recognition (FR) is used in a variety of application domains, from entertainment and banking to security and surveillance. Such applications rely on the FR model to be robust and perform well in a variety of settings. To achieve this, state-of-the-art FR models typically use expressive adaptive margin loss functions, which tie the feature norm to concepts related to sample quality, such as recognizability and perceptual image quality. Recently, through the development of Face Image Quality Assessment (FIQA) techniques, biometric utility has become the preferred measure of face-image quality and has been shown to be a better predictor of the usefulness of samples for face recognition compared to more human-centric aspects, such as resolution, blur, and lighting, tied to general image quality. While image quality expressed through feature norms exhibits a certain level of correlation with biometric utility, it does not fully encapsulate all aspects of utility. To address this point, we propose a new adaptive margin loss, FunFace (Face Recognition Through Utility and Norm Estimation), which incorporates biometric utility, estimated by the Certainty Ratio, into the adaptive margin, taking inspiration from AdaFace. We show that FunFace (when used to train a face recognition model) achieves competitive results to other state-of-the-art FR models on benchmarks containing high-quality samples, while surpassing them on low quality benchmarks.

Problem

• Face Recognition
• Biometric Utility
• Feature Norm
• Image Quality Assessment
• Adaptive Margin Loss

Innovation

• adaptive margin loss
• biometric utility
• face recognition
• feature norm
• Certainty Ratio