MagFace: A Universal Representation for Face Recognition and Quality Assessment

TL;DR

MagFace introduces a universal face embedding method where the feature magnitude indicates face quality, improving recognition accuracy and robustness by adaptively handling varying face qualities and preventing overfitting on noisy data.

Contribution

The paper proposes MagFace, a novel loss function that learns face embeddings with magnitudes reflecting quality, enhancing recognition and quality assessment simultaneously.

Findings

Outperforms state-of-the-art methods in face recognition accuracy.

Effectively measures face quality through embedding magnitude.

Improves clustering and robustness in face recognition tasks.

Abstract

The performance of face recognition system degrades when the variability of the acquired faces increases. Prior work alleviates this issue by either monitoring the face quality in pre-processing or predicting the data uncertainty along with the face feature. This paper proposes MagFace, a category of losses that learn a universal feature embedding whose magnitude can measure the quality of the given face. Under the new loss, it can be proven that the magnitude of the feature embedding monotonically increases if the subject is more likely to be recognized. In addition, MagFace introduces an adaptive mechanism to learn a wellstructured within-class feature distributions by pulling easy samples to class centers while pushing hard samples away. This prevents models from overfitting on noisy low-quality samples and improves face recognition in the wild. Extensive experiments conducted on face recognition, quality assessments as well as clustering demonstrate its superiority over state-of-the-arts. The code is available at https://github.com/IrvingMeng/MagFace.