[1]叶继华,郭 凤,黎 欣,等.DTFBNet:一种面向智能终端的轻量级人脸识别新方法[J].江西师范大学学报(自然科学版),2022,(02):126-133.[doi:10.16357/j.cnki.issn1000-5862.2022.02.03]
 YE Jihua,GUO Feng,LI Xin,et al.DTFBNet:the New Lightweight Face Recognition Method for Smart Terminals[J].Journal of Jiangxi Normal University:Natural Science Edition,2022,(02):126-133.[doi:10.16357/j.cnki.issn1000-5862.2022.02.03]
点击复制

DTFBNet:一种面向智能终端的轻量级人脸识别新方法()
分享到:

《江西师范大学学报》(自然科学版)[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2022年02期
页码:
126-133
栏目:
信息科学与技术
出版日期:
2022-03-25

文章信息/Info

Title:
DTFBNet:the New Lightweight Face Recognition Method for Smart Terminals
文章编号:
1000-5862(2022)02-0126-08
作者:
叶继华郭 凤黎 欣江 蕗江爱文
江西师范大学计算机信息工程学院,江西 南昌 330022
Author(s):
YE JihuaGUO FengLI XinJIANG LuJIANG Aiwen
School of Computer and Information Engineering,Jiangxi Normal University,Nanchang Jiangxi 330022,China
关键词:
DTFBNet DTFBlock 融合损失 轻量级 人脸识别
Keywords:
DTFBNet DTFBlock fusion loss lightweight face recognition
分类号:
TP 391.4
DOI:
10.16357/j.cnki.issn1000-5862.2022.02.03
文献标志码:
A
摘要:
对于智能终端资源不足的问题,目前有许多解决方法,但普遍存在依赖样本数据和参数量的问题.为此,该文先构造了一个深度卷积和传统卷积融合的模块DTFBlock(depthwise convolution and traditional convolution fusion Block); 然后在该基础上提出了一种基于MobileFaceNet的改进方法DTFBNet,DTFBNet参数量较小,在网络的识别效果较好; 最后,在面部识别数据集CASIA-Webface和LFW上进行实验,结果表明:该算法的最高准确率达到了99.40%,达到在同等参数量上具有竞争力的分类准确率.
Abstract:
There are many solutions to the problem of insufficient intelligent terminal resources,which are dependent on sample data and the number of parameters.The depthwise convolution and traditional convolution fusion block(DTFBlock)is proposed.Hence,an improved method DTFBNet based on MobileFaceNet is proposed.The DTFBNet proposed in the paper has a smaller number of parameters and better network results.Experiments on face recognition datasets CASIA-Webface and LFW show that the highest accuracy rate of the algorithm proposed in this paper reaches 99.40%,which is already a competitive classification accuracy for the same parameter amount.

参考文献/References:

[1] DENG Jiankang,GUO Jia,ZHANG Debing,et al.Lightweight face recognition challenge[EB/OL].[2021-11-17].https://ieeexplore.ieee.org/document/9022288.
[2] MÉNDEZ-VÁZQUEZ H,MARTÍNEZ-DÍAZ Y,NICOLÁS-DÍAZ M,et al.Bench-marking lightweight face architectures on specific face recognition scenarios[J].Artificial Intelligence Review,2021,54(8):6201-6244.
[3] SANDLER M,HOWARD A,ZHU Menglong,et al.MobileNet V2:inverted residuals and linear bottlenecks[EB/OL].[2021-11-17].https://arxiv.org/pdf/1801.04381v3.pdf.
[4] MA Ningning,ZHANG Xiangyu,ZHENG Haitao,et al.ShuffleNet V2:practical guidelines for efficient CNN architecture design[EB/OL].[2021-11-17].https://arxiv.org/pdf/1807.11164.pdf.
[5] ZHANG Qian,LI Jianjun,YAO Meng,et al.VarGNet:variable group convolutional neural network for efficient embedded computing[EB/OL].[2021-11-17].https://arxiv.org/abs/1907.05653.
[6] CHEN Sheng,LIU Yang,GAO Xiang,et al.MobileFaceNets:efficient CNNs for accurate real-time face verification on mobile devices[EB/OL].[2021-11-17].https://arxiv.org/ftp/arxiv/papers/1804/1804.07573.pdf.
[7] HOWARD A G,ZHU Menglong,CHEN Bo,et al.MobileNets:efficient convolutional neural networks for mobile vision applications[EB/OL].[2021-11-17].https://arxiv.org/pdf/1704.04861.pdf.
[8] LI Xianyang,WANG Feng,HU Qinghao,et al.AirFace:lightweight and efficient model for face recognition[EB/OL].[2021-11-17].https://www.researchgate.net/publication/339768456_AirFaceLightweight_and_Efficient_Model_for_Face_Recognition.
[9] MARTÍNEZ-DÍAZ Y,LUEVANO L S,MÉNDEZ-VÁZQUEZ H,et al.ShuffleFaceNet:a lightweight face architecture for efficient and highly-accurate face recognition[EB/OL].[2021-11-17].https://ieeexplore.ieee.org/document/902220.
[10] YAN Mengjia,ZHAO Mengao,XU Zining,et al.VargFaceNet:an efficient variable group convolutional neural network for lightweight face recognition[EB/OL].[2021-11-17].https://ieeexplore.ieee.org/document/9022149/.
[11] LI Yuancheng,WANG Yimeng,LI Daoxing.Privacy-preserving lightweight face recognition[J].Neurocomputing,2019,363:212-222.
[12] HUANG Gao,LIU Shichen,VAL DER MAATEN L,et al.CondenseNet:an efficient DenseNet using learned group convolutions[EB/OL].[2021-11-17].https://arxiv.org/pdf/1711.09224.pdf.
[13] SUN Ke,LI Mingjie,LIU Dong,et al.IGCV3:interleaved low-rank group convolutions for efficient deep neural networks[EB/OL].[2021-11-17].https://arxiv.org/pdf/1806.00178.pdf.
[14] WU Bichen,ALVIN W,YUE Xiangyu,et al.Shift:a zero FLOP,zero parameter alternative to spatial convolutions[EB/OL].[2021-11-17].https://arxiv.org/pdf/1711.08141.pdf.
[15] HUANG G B,RAMESH M,BERG T,et al.Labeled faces in the wild:a database for studying face recognition in uncon-strained environments[EB/OL].[2021-11-17].http://cs.brown.edu/courses/cs143/2011/proj4/papers/lfw.pdf.
[16] CHOLLET F.Xception:deep learning with depthwise separable convolutions[EB/OL].[2021-11-17].https://arxiv.org/pdf/1610.02357.pdf.
[17] ZHANG Kaipeng,ZHANG Zhanpeng,LI Zhifeng,et al.Joint face detection and alignment using multitask cascaded convolutional networks[J].IEEE Signal Processing Letters,2016,23(10):1499-1503.
[18] VO D M,LEE S W.Robust face recognition via hierarchical collaborative representation[J].Information Sciences,2018,432:332-346.
[19] 杜星悦,董洪伟,杨振.基于深度网络的人脸区域分割方法[J].计算机工程与应用,2019,55(8):171-174.
[20] 王小玉,韩昌林,胡鑫豪.加权特征融合的密集连接网络人脸识别算法[J].计算机科学与探索,2019,13(7):1195-1205.
[21] DENG Jiankang,GUO Jia,XUE Niannan,et al.ArcFace:additive angular margin loss for deep face recognition[EB/OL].[2021-11-17].https://arxiv.org/pdf/1801.07698.pdf.
[22] DUONG C N,QUACH K G,JALATA I,et al.MobiFace:a lightweight deep learning face recognition on mobile devices[EB/OL].[2021-11-17].https://arxiv.org/pdf/1811.11080.pdf.
[23] WANG Xiaobo,FU Tianyuu,LIAO Shengcai,et al.Exclusivity-consistency regularized knowledge distillation for face recognition[EB/OL].[2021-11-17].https://link.springer.com/chapter/10.1007/978-3-030-58586-0_20.

备注/Memo

备注/Memo:
收稿日期:2022-01-03
基金项目:国家自然科学基金(62167005,61966018)和江西省教育厅重点科研课题(GJJ200302)资助项目.
作者简介:叶继华(1966—),男,江西上饶人,教授,博士生导师,主要从事普适计算、物联网技术、数据融合、图像处理等研究.E-mail:yjhwcl@163.com
更新日期/Last Update: 2022-03-25