基于YOL0v5与ArcFace的轻量化人脸识别系统研究

打开文本图片集

中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2025）15-0038-05

Research on Lightweight Face Recognition System Based on YOLOv5 and ArcFace

ZHANG Jie,MU Xiya,HUANG Xinyu, CHEN Si (Changchun Universityof Science and Technology,Changchun 13oo22, China)

Abstract: With the advent of the inteligent era,higher and higher requirements are put forward for the accuracy, procesing time andlightweight deployment offace recognition technology.Inviewofthe above problems,this paper designs a lightweightfacerecognitionsystembyusingthecharacteristicsofhighaccuracyandfastresponsetimeofYOLOv5andArcFace networkstructure.ComparedwiththetraditionalConvolutionalNeuralNetwork(CNN),thesystemonlyusesless thanlillion parameters,far less than the 6.8 milion parametersofFaceNet. On the LFWverification set,itsaccuracyreaches 99.43% ， showingextremelyhighaccuracyAtthedatasourcelevel,throughthedataaugmentationstrategy,thesystemalsohasagood performance infaceoclusionandlow-lightscenes,whichprovidesanewideaforthefuture inteligentrapidsecurityfrontier inspection.

Keywords: face recognition; YOLOv5; ArcFace; Deep Learning; lightweight structure; feature extraction; data augmentation

0 引言

近年来，随着人工智能技术的飞速发展，人脸识别技术也逐渐走进千家万户，在通勤、安保和社会治理等领域发挥着越来越重要的作用[]。（剩余10675字）