基于Caffe框架的人脸定位与识别系统的设计  

Design of face location and recognition system based on Caffe framework

作　　者：黄琳[1] 蒋为 杨铁军[1] Huang Lin;Jiang Wei;Yang Tiejun(Guilin University of Technology,Guilin,Guangxi 541004,China)

机构地区：[1]桂林理工大学信息学院

出　　处：《计算机时代》2019年第6期56-58,62,共4页Computer Era

基　　金：广西教育厅项目“基于卷积神经网络的色素性皮损图像语义分割方法研究”(2017KY0247);广西自然科学基金“融入深度卷积神经网络的色素性皮损图像水平集语义分割方法研究”(2018GXNSFBA281081)

摘　　要：人脸识别是生物特征识别的重要内容,其应用范围也越来越广,如手机的人脸解锁,支付宝的刷脸支付和公共区域的犯罪分子检测等。文章提出了一种基于Caffe框架的人脸定位与识别系统的设计方法。该方法中的人脸定位采用多任务卷积神经网络(Multi-task convolutional neural network,简称MTCNN),并基于Caffe框架实现卷积神经网络(Convolutional Neural Networks,简称CNN)来对人脸进行分类与识别,采用YouTube以及Labeled Faces数据集作为实验数据集,实现了人脸定位和人脸识别功能,取得了较好的定位与识别效果。Face recognition is an important part of biometric recognition,and its application range becomes wider and wider,such as the face unlock of mobile phones,Alipay's face payment and criminal detection in public areas.This paper proposes a design method of face location and recognition system based on Caffe framework.The multi-task convolutional neural network (MTCNN) is used for face localization,and the Caffe framework based convolutional neural network is used to classify and identify the face.YouTube and Labeled Faces datasets are used in experiment to realize the face location and recognition,and a good effect of localization and recognition is achived.

关 键 词：Caffe MTCNN 卷积神经网络 人脸定位 人脸识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]