基于联合网络的域适应行人重识别算法

doi:10.3969/j.issn.1006-2475.2023.06.009

计算机与现代化 ›› 2023, Vol. 0 ›› Issue (06): 48-55.doi: 10.3969/j.issn.1006-2475.2023.06.009

基于联合网络的域适应行人重识别算法

李国鑫¹, 曲寒冰^1,2, 朱成博², 王鑫轩², 胡嘉宝²

1.北京市新技术应用研究所有限公司,北京 100035;
2.河北工业大学人工智能与数据科学学院,天津 300401

收稿日期:2022-04-27 修回日期:2022-05-25 出版日期:2023-06-28 发布日期:2023-06-28
作者简介:李国鑫（1995—）,男,河北衡水人,硕士研究生,研究方向：计算机视觉,行人重识别,E-mail： lgxxye@163.com; 曲寒冰（1977—）,男,黑龙江哈尔滨人,研究员,博士,研究方向：机器学习,计算机视觉,生物识别,图像处理,E-mail： 147454362@qq.com; 朱成博（1998—）,男,山东德州人,硕士研究生,研究方向：计算机视觉,行人重重识别,E-mail： 849689109@qq.com; 王鑫轩（1996—）,男,河北石家庄人,硕士研究生,研究方向：自然语言处理,知识图谱,E-mail： 1176308922@qq.com; 胡嘉宝（1997—）,女,黑龙江绥化人,硕士研究生,研究方向：自然语言处理,知识图谱,E-mail： 1006000815@qq.com。
基金资助:
国家自然科学基金资助项目（91746207）; 新疆自治区重点研发计划项目（2020B03001-2）

Domain Adapted Person Re-identification Algorithm Based on Joint Network

LI Guo-xin¹, QU Han-bing^1,2, ZHU Cheng-bo², WANG Xin-xuan², HU Jia-bao²

1. Beijing Institute of New Technology Application, Beijing 100035, China;
2. School of Artificial Intelligence and Data Science, Hebei University of Technology, Tianjin 300401, China

Received:2022-04-27 Revised:2022-05-25 Online:2023-06-28 Published:2023-06-28

摘要/Abstract

摘要： 针对域适应方法存在对图像细节信息利用不充分和域间数据分布差异导致跨域行人重识别准确率较低的问题,设计一种基于联合网络的域适应行人重识别算法。首先,为了提高模型对细节信息的获取能力,网络嵌入注意力机制模块,并设计非对称多尺度分支,补充全局特征中缺失的细节信息;其次,为了缓解域间差异造成的影响,在CycleGAN网络的基础上引入在线相关一致性损失,训练样本生成器,生成带有目标域风格的源域样本,降低域间数据分布差异。然后,在域适应阶段,设计基于聚类联合网络的方式,将联合网络中教师网络输出的样本特征和分类概率作为监督信息,监督学生网络的训练,避免聚类硬伪标签在训练中造成的误差被放大。最后,为了缓解噪声伪标签的影响,引入动量对比损失（Momentum Contrast Loss, MoCo Loss）,提高模型自适应能力。在Market-1501和DukeMTMC-ReID上进行实验,实验结果相比主流的算法有所提升,验证了该算法具有一定的竞争力。

关键词: 行人重识别, 域适应, 多尺度特征, 注意力机制, 伪标签, 域风格迁移

Abstract: Aiming at the problems of insufficient use of image details in the domain adaptation method and low accuracy of cross-domain person re-recognition due to data distribution differences between domains, a domain-adapted person re-recognition algorithm based on joint network was designed. Firstly, in order to improve the model's ability to obtain details, an attention mechanism module is embedded in the network, and an asymmetric multi-granularity network was designed to supplement the missing details in the global feature. Secondly, in order to alleviate the influence caused by inter-domain differences, online correlation consistency loss is introduced on the basis of CycleGAN network, and sample generator is trained to generate source domain samples with target domain style to reduce inter-domain data distribution differences. Then, in the domain adaptation stage, the method based on clustering joint network is designed, and the classification probability output by student model in the joint network is used as the supervision information to supervise the training of teacher network, so as to avoid the error amplification caused by clustering hard pseudo-label in training. Finally, Momentum Contrast Loss （MoCo Loss） is introduced to alleviate the influence of noise pseudo-labels to improve the adaptive ability of the model. Experiments are carried out on Market-1501 and DukeMTMC-Reid, and the results show that the algorithm has certain competitiveness compared with the mainstream algorithms.

Key words: person re-identification, domain adaptation, multigrain features, attention mechanism, pseudo tags, domain style migration

中图分类号:

TP391

李国鑫, 曲寒冰, 朱成博, 王鑫轩, 胡嘉宝. 基于联合网络的域适应行人重识别算法[J]. 计算机与现代化, 2023, 0(06): 48-55.

LI Guo-xin, QU Han-bing, ZHU Cheng-bo, WANG Xin-xuan, HU Jia-bao. Domain Adapted Person Re-identification Algorithm Based on Joint Network[J]. Computer and Modernization, 2023, 0(06): 48-55.

参考文献 26

[1]	CHENG D S, CRISTANI M.Person re-identification[M]// Person Re-identification by Articulated Appearance Matching. London: Springer, 2014:139-160.
[2]	CHEN Y C, ZHU X, ZHENG W S, et al.Person re-identification by camera correlation aware feature augmentation[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018,40(2):392-408.
[3]	LIN Y, DONG X, ZHENG L, et al.A bottom-up clustering approach to unsupervised person re-identification[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019,33(1):8738-8745.
[4]	JIN X, LAN C, ZENG W, et al.Global distance distributions separation for unsupervised person reidentification[C]// Proceedings of IEEE International Conference on Computer Vision. 2020:5142-5150.
[5]	ISOLA P, ZHU J, ZHOU T, et al.Image-to-image translation with conditional adversarial networks[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017:5967-5976.
[6]	DENG W, ZHENG L, YE Q, et al.Image-image domain adaptation with preserved self-similarity and domain-dissimilarity for person re-identification[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2018:994-1003.
[7]	ZHU J, PARK T, ISOLA P, et al.Unpaired image-to-image translation using cycle-consistent adversarial networks[C]// 2017 IEEE International Conference on Computer Vision (ICCV). 2017:2242-2251.
[8]	WEI L, ZHANG S, GAO W, et al.Person transfer gan to bridge domain gap for person re-identification[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2018:79-88.
[9]	YANG F, LI K, ZHONG Z, et al.Asymmetric co-teaching for unsupervised cross-domain person re-identification[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020,34(7):12597-12604.
[10]	GE Y, CHEN D, LI H.Mutual mean-teaching: Pseudo label refinery for unsupervised domain adaptation on person re-identification[J]. arXiv preprint arXiv:2001.01526, 2020.
[11]	ZHENG L, SHEN L, TIAN L, et al.Scalable person re-identification: A benchmark[C]// 2015 IEEE International Conference on Computer Vision (ICCV). 2015:1116-1124.
[12]	RISTANI E, SOLERA F, ZOU R S, et al. Performance measures and a data set for multi-target, multi-camera tracking[J]. arXiv preprint arXiv:1609.01775, 2016.
[13]	CHEN H, LAGADEC B, BRÉMOND F. Enhancing diversity in teacher-student networks via asymmetric branches for unsupervised person re-identification[C]// 2021 IEEE Winter Conference on Applications of Computer Vision (WACV). 2021:1-10.
[14]	PAN X, LUO P, SHI J, et al.Two at once: Enhancing learning and generalization capacities via IBN-Net[C]// Computer Vision - ECCV 2018. 2018:484-500.
[15]	GE Y, ZHU F, CHEN D, et al. Structured domain adaptation with online relation regularization for unsupervised person re-ID[J]. IEEE Transactions on Neural Networks and Learning Systems, 2022,PP(99):1-14.
[16]	ZHENG Z D, ZHENG L, YANG Y.Unlabeled samples generated by gan improve the person re-identification baseline in vitro[C]// IEEE International Conference on Computer Vision. 2017:660-669.
[17]	KHAN K, REHMAN S U, AZIZ K, et al.DBSCAN: Past, present and future[C]// The 5th International Conference on the Applications of Digital Information and Web Technologies (ICADIWT 2014). 2014:232-238.
[18]	HE K, FAN H, WU Y, et al.Momentum contrast for unsupervised visual representation learning[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 2020:9726-9735.
[19]	HADSELL R, CHOPRA S, LECUN Y.Dimensionality Reduction by learning an invariant mapping[C]// 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06). 2006:1735-1742.
[20]	ZHONG Z, ZHENG L, LUO Z, et al.Learning to adapt invariance in memory for person re-identification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021,43(8):2723-2738.
[21]	FU Y, WEI Y, WANG G, et al.Self-similarity grouping: A simple unsupervised cross domain adaptation approach for person re-identification[C]// 2019 IEEE/CVF International Conference on Computer Vision (ICCV). 2019:6111-6120.
[22]	LI Y, YANG F, LIU Y, et al.Adaptation and re-identification network: An unsupervised deep transfer learning approach to person re-identification[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). 2018:285-2856.
[23]	WANG D, ZHANG S.Unsupervised person re-identification via multi-label classification[C]// Proceedings of the IEEE/ CVF Conference on Computer Vision and Pattern Recognition. 2020:10981-10990.
[24]	ZHAI Y P, YE Q X, LU S J, et al.Multiple expert brainstorming for domain adaptive person re-identification[C]// European Conference on Computer Vision. 2020:594-611.
[25]	SONG L, WANG C, ZHANG L, et al.Unsupervised domain adaptive re-identification: Theory and practice[J]. Pattern Recognition, 2020,102(3). DOI:10.1016/j.patcog.2019.107173.
[26]	WANG W, ZHAO F, LIAO S, et al.Attentive waveblock: Complementarity-enhanced mutual networks for unsupervised domain adaptation in person re-identification and beyond[J]. IEEE Transactions on Image Processing, 2022, 30(5):1092-1108.

基于联合网络的域适应行人重识别算法

Domain Adapted Person Re-identification Algorithm Based on Joint Network

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献 26

相关文章 15

编辑推荐

Metrics

本文评价

[1]	王秋忆, 周浩, 郑婷婷. 改进RetinaNet的电力设备目标检测方法[J]. 计算机与现代化, 2024, 0(01): 47-52.
[2]	林启钊, 彭志平, 郭棉, 崔得龙. 基于双向多步预测的炉管温度场重构方法[J]. 计算机与现代化, 2024, 0(01): 53-58.
[3]	李亚平, 王军防, 余红梅, 窦一民, 肖媛, 田继林. Regformer：基于稀疏注意力的输油管道水力压降预测方法[J]. 计算机与现代化, 2024, 0(01): 59-66.
[4]	邱凯星, 冯广. 基于双重特征注意力的多标签图像分类模型[J]. 计算机与现代化, 2023, 0(12): 41-47.
[5]	张浩洋, 尹梓名, 乐珺怡, 沈达聪, 束翌俊, 杨自逸, 孔祥勇, 龚伟. 3D-SPRNet: 一种基于并行解码器和双注意力机制的胆囊癌分割模型[J]. 计算机与现代化, 2023, 0(12): 59-66.
[6]	张伯泉, 麦海鹏, 陈嘉敏, 逄锦聚. 基于高灰度值注意力机制的脑白质高信号分割[J]. 计算机与现代化, 2023, 0(12): 67-75.
[7]	宁娟, 周庆华, 曾小为. 改进YOLOv7算法在西林瓶轧盖缺陷检测中的应用[J]. 计算机与现代化, 2023, 0(12): 82-86.
[8]	王宇航, 董宝良, 公超, 尚真真, 姚康宁. 基于意图识别的空中群目标动态威胁评估[J]. 计算机与现代化, 2023, 0(12): 100-104.
[9]	罗明杰, 冯开平. 基于沙漏结构与注意力机制的轻量级人脸表情识别方法[J]. 计算机与现代化, 2023, 0(11): 89-94.
[10]	张嘉琪, 徐啟蕾. 基于NAM-YOLO网络的苹果缺陷检测算法[J]. 计算机与现代化, 2023, 0(10): 53-58.
[11]	叶思佳, 魏延, 杜韩宇, 邓金枝. 结合注意力机制的HRNet图像语义分割算法[J]. 计算机与现代化, 2023, 0(10): 65-69.
[12]	陈嘉敏, 张伯泉, 麦海鹏. 基于特征融合的海马体分割[J]. 计算机与现代化, 2023, 0(08): 1-6.
[13]	刘续, 查可可. 一种用于机场特种车辆作业的环境目标识别方法[J]. 计算机与现代化, 2023, 0(08): 18-24.
[14]	王鸿, 葛红. 基于注意力机制和语义相似度的跨模态哈希检索[J]. 计算机与现代化, 2023, 0(08): 44-53.
[15]	欧阳飞, 吴旭, 向东升. 基于改进YOLOX的垃圾分类检测方法[J]. 计算机与现代化, 2023, 0(08): 68-73.