基于深度学习的人脸分析研究进展

(厦门大学 信息科学与技术学院,福建省智慧城市感知与计算重点实验室,福建 厦门 361005)

深度学习; 卷积神经网络; 人脸数据库; 人脸识别; 人脸分析;

Recent Advances on Deep-Learning-Based Face Analysis
YAN Yan,CHEN Riwei,WANG Hanzi*

(Fujian Key Laboratory of Sensing and Computing for Smart City,School of Information Science and Engineering,Xiamen University,Xiamen 361005,China)

DOI: 10.6043/j.issn.0438-0479.201609024

备注

近年来,基于深度学习的人脸分析取得了巨大的进步,成为计算机视觉领域最为活跃的研究方向之一.为了进一步推动深度学习和人脸分析的研究,结合近年已发表的相关文献,对基于深度学习的人脸分析技术进行综述.首先,简要概述深度学习及其发展历史,并分析深度学习有效性原因.然后,按照任务目的 的不同,将人脸分析分成了人脸检测、人脸关键点检测、人脸识别、人脸属性识别等任务进行详细的介绍和讨论,重点分析各种任务现阶段存在的主要问题.接着,介绍人脸分析中常用的人脸数据库.最后,讨论深度学习和人脸分析面临的主要挑战,并给出结论.

In recent years,the face analysis based on deep learning has made great progress,and has become one of the most active research areas in the field of computer vision.In order to further promote the study of deep learning and face analysis,this paper overviews recent advances on the deep-learning-based face analysis techniques in the literature.First,a brief overview of deep learning and its history are given and reasons for the effectiveness of deep learning are also analyzed.Then,according to different objectives,four face analysis tasks,i.e.,face detection,facial key-point detection,face recognition,face attribute recognition,are introduced and discussed in detail,and the key problems existing in these tasks at present are analyzed.After that,the commonly used face databases in the face analysis are described.At last,main challenges of face analysis based deep learning are shown and the conclusion is presented.