SphereFace: Deep Hypersphere Embedding for Face Recognition

TLDR

Deep face recognition under an open‑set protocol requires features whose maximal intra‑class distance is smaller than the minimal inter‑class distance, yet few existing algorithms satisfy this criterion. To address this, we introduce the angular softmax (A‑Softmax) loss, enabling convolutional neural networks to learn angularly discriminative features. A‑Softmax imposes discriminative constraints on a hypersphere manifold that aligns with the prior that faces lie on such a manifold, and its angular margin m can be quantitatively adjusted and derived to approximate the ideal feature criterion. Extensive analysis and experiments on LFW, YTF, and MegaFace Challenge 1 demonstrate that A‑Softmax loss outperforms existing methods in face recognition tasks.

Abstract

This paper addresses deep face recognition (FR) problem under open-set protocol, where ideal face features are expected to have smaller maximal intra-class distance than minimal inter-class distance under a suitably chosen metric space. However, few existing algorithms can effectively achieve this criterion. To this end, we propose the angular softmax (A-Softmax) loss that enables convolutional neural networks (CNNs) to learn angularly discriminative features. Geometrically, A-Softmax loss can be viewed as imposing discriminative constraints on a hypersphere manifold, which intrinsically matches the prior that faces also lie on a manifold. Moreover, the size of angular margin can be quantitatively adjusted by a parameter m. We further derive specific m to approximate the ideal feature criterion. Extensive analysis and experiments on Labeled Face in the Wild (LFW), Youtube Faces (YTF) and MegaFace Challenge 1 show the superiority of A-Softmax loss in FR tasks.