# L2-constrained Softmax Loss for Discriminative Face Verification

arXiv: https://arxiv.org/abs/1703.09507 · HTML: https://mlsanity.com/paper/1703.09507/

## Key facts

- Cited by 96 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 28 Mar 2017.

## Built on by (most cited first, 30 of 96)

- [Open-Set Recognition: a Good Closed-Set Classifier is All You Need?](https://mlsanity.com/paper/2110.06207/) (2021)
- [MagFace: A Universal Representation for Face Recognition and Quality Assessment](https://mlsanity.com/paper/2103.06627/) (2021)
- [Mitigating Neural Network Overconfidence with Logit Normalization](https://mlsanity.com/paper/2205.09310/) (2022)
- [Are we really making much progress? Revisiting, benchmarking, and refining heterogeneous graph neural networks](https://mlsanity.com/paper/2112.14936/) (2021)
- [Deep Gait Recognition: A Survey](https://mlsanity.com/paper/2102.09546/) (2021)
- [Neural Collapse with Normalized Features: A Geometric Analysis over the Riemannian Manifold](https://mlsanity.com/paper/2209.09211/) (2022)
- [von Mises-Fisher Loss: An Exploration of Embedding Geometries for Supervised Learning](https://mlsanity.com/paper/2103.15718/) (2021)
- [SphereFace2: Binary Classification is All You Need for Deep Face Recognition](https://mlsanity.com/paper/2108.01513/) (2021)
- [Learning Compatible Embeddings](https://mlsanity.com/paper/2108.01958/) (2021)
- [Domain Generalization through Audio-Visual Relative Norm Alignment in First Person Action Recognition](https://mlsanity.com/paper/2110.10101/) (2021)
- [HINormer: Representation Learning On Heterogeneous Information Networks with Graph Transformer](https://mlsanity.com/paper/2302.11329/) (2023)
- [Block Selection Method for Using Feature Norm in Out-of-distribution Detection](https://mlsanity.com/paper/2212.02295/) (2022)
- [Variational Graph Normalized Auto-Encoders](https://mlsanity.com/paper/2108.08046/) (2021)
- [A review of ensemble learning and data augmentation models for class imbalanced problems: combination, implementation and evaluation](https://mlsanity.com/paper/2304.02858/) (2023)
- [Regular Polytope Networks](https://mlsanity.com/paper/2103.15632/) (2021)
- [SphereFace Revived: Unifying Hyperspherical Face Recognition](https://mlsanity.com/paper/2109.05565/) (2021)
- [Rethinking Common Assumptions to Mitigate Racial Bias in Face Recognition Datasets](https://mlsanity.com/paper/2109.03229/) (2021)
- [Cluster and Aggregate: Face Recognition with Large Probe Set](https://mlsanity.com/paper/2210.10864/) (2022)
- [SFace: Sigmoid-Constrained Hypersphere Loss for Robust Face Recognition](https://mlsanity.com/paper/2205.12010/) (2022)
- [A Non-isotropic Probabilistic Take on Proxy-based Deep Metric Learning](https://mlsanity.com/paper/2207.03784/) (2022)
- [Improving Federated Learning Face Recognition via Privacy-Agnostic Clusters](https://mlsanity.com/paper/2201.12467/) (2022)
- [50 Years of Automated Face Recognition](https://mlsanity.com/paper/2505.24247/) (2025)
- [Guiding Neural Collapse: Optimising Towards the Nearest Simplex Equiangular Tight Frame](https://mlsanity.com/paper/2411.01248/) (2024)
- [Deep transfer learning for image classification: a survey](https://mlsanity.com/paper/2205.09904/) (2022)
- [Dynamic Metric Learning: Towards a Scalable Metric Space to Accommodate Multiple Semantic Scales](https://mlsanity.com/paper/2103.11781/) (2021)
- [IronMask: Modular Architecture for Protecting Deep Face Template](https://mlsanity.com/paper/2104.02239/) (2021)
- [Weak Adaptation Learning -- Addressing Cross-domain Data Insufficiency with Weak Annotator](https://mlsanity.com/paper/2102.07358/) (2021)
- [Similarity Distribution based Membership Inference Attack on Person Re-identification](https://mlsanity.com/paper/2211.15918/) (2022)
- [MANO: Exploiting Matrix Norm for Unsupervised Accuracy Estimation Under Distribution Shifts](https://mlsanity.com/paper/2405.18979/) (2024)
- [Harnessing Unrecognizable Faces for Improving Face Recognition](https://mlsanity.com/paper/2106.04112/) (2021)

## Abstract (arXiv)

In recent years, the performance of face verification systems has significantly improved using deep convolutional neural networks (DCNNs). A typical pipeline for face verification includes training a deep network for subject classification with softmax loss, using the penultimate layer output as the feature descriptor, and generating a cosine similarity score given a pair of face images. The softmax loss function does not optimize the features to have higher similarity score for positive pairs and lower similarity score for negative pairs, which leads to a performance gap. In this paper, we add an L2-constraint to the feature descriptors which restricts them to lie on a hypersphere of a fixed radius. This module can be easily implemented using existing deep learning frameworks. We show that integrating this simple step in the training pipeline significantly boosts the performance of face verification. Specifically, we achieve state-of-the-art results on the challenging IJB-A dataset, achieving True Accept Rate of 0.909 at False Accept Rate 0.0001 on the face verification protocol. Additionally, we achieve state-of-the-art performance on LFW dataset with an accuracy of 99.78%, and competing performance on YTF dataset with accuracy of 96.08%.