基于深度学习的人脸识别系统

Face Recognition System Based on Deep Learning

以深度学习为基础,将人脸检测方法(MTCNN)、OpenCV仿射变换方法、人脸识别方法(FaceNet)、相似性查找方法(HNSW)和人脸属性识别方法(FAN)有机结合,构建了一个人脸识别系统。在深度学习开发框架(Caffe)上进行模型训练与测试,并将预训练模型应用于人脸注册、人脸检索和人脸属性识别任务,通过实例测试表明在现实场景中具有一定的鲁棒性。该系统在课堂考勤、身份认证、门禁系统、登录解锁和社交媒体平台等方面也具有巨大的应用潜力。并将该研究成果向学生进行专题讲解,以期开拓学生的视野,增加学生学习科研兴趣,提高学生学术水平,锻炼动手能力,有利于学生今后的就业。

Based on deep learning,this paper develops a face recognition system that combines face detection method( MTCNN),OpenCV affine transformation method,face recognition method( FaceNet),similarity finding method( HNSW) and face attribute recognition method( FAN). Model training and testing are carried out on the deep learning development framework( Caffe),and the pre-training model is applied to face registration,face retrieval and face attribute recognition tasks. The example test shows that it has certain robustness in real scenes. The system also has great application potential in classroom attendance,identity authentication,access control systems,login and unlocking,and social media platforms. The research results will be given to undergraduate students to broaden students’ horizons,increase students’ interest in studying and scientific research,improve student’ academic performance,and exercise their hands-on ability,which is conducive to students’ future employment.