%0 Journal Article
%T 深度人脸识别方法综述<br>A Survey of Deep Face Recognition
%A 王海勇
%A 潘海涛
%J Software Engineering and Applications
%P 609-619
%@ 2325-2278
%D 2023
%I Hans Publishing
%R 10.12677/SEA.2023.124059
%X 深度人脸识别通过大规模数据集训练卷积神经网络获取更鲁棒的人脸表示,极大的提升了人脸识别性能。文中总结了深度人脸识别方法的发展脉络，首先根据卷积神经网络的不同发展阶段回顾了现有的深度人脸识别方法，其次对基于欧几里得距离以及基于角余弦裕度的损失函数进行了回顾，同时总结了一些针对特定任务的人脸识别方法。然后总结了现有的人脸识别数据集以及人脸识别性能的评价指标，并对主流深度人脸识别方法进行了比较。最后总结了人脸识别当前面临的挑战和未来的发展趋势。<br />
Deep face recognition greatly improves the performance of face recognition by training convolutional neural networks on large-scale data sets to obtain more robust face representation. This paper summarizes the development of depth face recognition methods. First, the existing depth face recognition methods are reviewed according to the different development stages of convolutional neural networks. Secondly, the loss functions based on Euclidean distance and angular cosine margin are reviewed, and some task-specific face recognition methods are summarized. Then, the existing data sets and the evaluation indicators of face recognition performance are summarized, and the mainstream depth face recognition methods are compared. Finally, the current challenges and future trends of face recognition are summarized.
%K 人脸识别，卷积神经网络，人脸表示，损失函数<br>Face Recognition
%K Convolutional Neural Networks
%K Face Representation
%K Loss Function
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=70813