| 351 | 2 | 59 |
| 下载次数 | 被引频次 | 阅读次数 |
无监督跨域行人重识别技术，通过将源域的有标签信息迁移到目标域以应对无标签情况，采用聚类方法达到无监督域适应效果，实现跨域行人再识别。然而，仅依赖全局特征的聚类易受域间差异影响产生噪声，且单网络结构训练易导致放大误差影响模型性能。针对此类问题，提出一种双分支注意力特征融合算法，分别抽取并融合域不变特征和域特定特征，以增强目标域上的泛化能力和减少聚类噪声。同时，引入对称网络架构进行同步协同训练，形成互为监督的学习机制，有效抑制过拟合问题。实验表明，在Market-1501和DukeMTMC-ReID公开数据集上，该算法显著提升了无监督跨域行人重识别的mAP和Rank准确率。
Abstract：With unsupervised cross domain pedestrian re-identification technology, labeled information could be transfered from the source domain to the target domain to cope with unlabeled situations, clustering methods were to achieve unsupervised domain adaptation, so that to achieve cross domain pedestrian re-identification. However, clustering based on solely global features was susceptible to noise generated by inter domain differences, and single network structure training could lead to error amplification and affect model performance. A dual-branch attention-based fusion algorithm was proposed to extract & fuse invariant and specific features to enhance target domain generalization and to reduce clustering noise. At the same time, a symmetric network architecture was introduced for synchronous collaborative training, form a mutually supervised learning mechanism to effectively suppress overfitting problems. Experiments showed that on the Market-1501 and DukeMTMC-ReID datasets, the algorithm significantly improved the mAP and Rank accuracy of unsupervised cross domain pedestrian re-identification.
[1] ZHANG M,YU Z,HAN Y,et al.A review of pedestrian re-identification for complex scenes [J].Computer science,2022,49 (10):138-150.
[2] 任向阳，王杰，马天磊，等.红外弱小目标检测技术综述[J].郑州大学学报(理学版),2020,52(2):1-21.REN X Y,WANG J,MA T L,et al.Review on infrared dim and small target detection technology[J].Journal of Zhengzhou university (natural science edition),2020,52(2):1-21.
[3] ZHANG H,WANG X Y,LIU J X,et al.Chinese named entity recognition method for the finance domain based on enhanced features and pretrained language models[J].Information sciences,2023,625:385-400.
[4] LIN X T,REN P Z,YEH C,et al.Unsupervised person re-identification:a systematic survey of challenges and solutions[EB/OL].(2021-10-02)[2023-11-06].https://arxiv.org/abs/2109.06057.
[5] VERMA A,SUBRAMANYAM A V,WANG Z,et al.Unsupervised domain adaptation for person re-identification via individual-preserving and environmental-switching cyclic generation[J].IEEE transactions on multimedia,2023,25:364-377.
[6] QIN H N,XIE W Y,LI Y S,et al.PTGAN:a proposal-weighted two-stage GAN with attention for hyperspectral target detection[C]//IEEE International Geoscience and Remote Sensing Symposium.Piscataway:IEEE Press,2021:4428-4431.
[7] DENG W J,ZHENG L,YE Q X,et al.Image-image domain adaptation with preserved self-similarity and domain-dissimilarity for person re-identification[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2018:994-1003.
[8] CHEN Y B,ZHU X T,GONG S G.Instance-guided context rendering for cross-domain person re-identification[C]//IEEE/CVF International Conference on Computer Vision.Piscataway:IEEE Press,2019:232-242.
[9] YU H X,ZHENG W S,WU A C,et al.Unsupervised person re-identification by soft multilabel learning[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2019:2143-2152.
[10] WU J L,LIAO S C,LEI Z,et al.Clustering and dynamic sampling based unsupervised domain adaptation for person re-identification[C]//IEEE International Conference on Multimedia and Expo.Piscataway:IEEE Press,2019:886-891.
[11] XUAN S Y,ZHANG S L.Intra-inter camera similarity for unsupervised person re-identification[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2021:11921-11930.
[12] JI K K,HUANG P,ZHU S H.Attention-calibration based double-branch cross-domain person re-identification[J].Knowledge-based systems,2022,258:110019.
[13] GE Y X,CHEN D P,LI H S.Mutual mean-teaching:pseudo label refinery for unsupervised domain adaptation on person re-identification[EB/OL].(2020-01-06)[2023-11-06].http://arxiv.org/abs/2001.01526.
[14] HE K M,ZHANG X Y,REN S Q,et al.Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2016:770-778.
[15] MA J,ZHANG T,YANG C,et al.Review of wafer surface defect detection methods[J].Electronics,2023,12:1787-1804.
[16] MA J,MA X,YANG C,et al.An air pollutant forecast correction model based on ensemble learning algorithm[J].Electronics,2023,12:1463-1473.
[17] GAO S H,HAN Q,LI D,et al.Representative batch normalization with feature calibration[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2021:8665-8675.
[18] LIU Z W,LUO P,QIU S,et al.DeepFashion:powering robust clothes recognition and retrieval with rich annotations[C]//IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2016:1096-1104.
[19] 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.Piscataway:IEEE Press,2017:3774-3782.
[20] ZHENG L,SHEN L Y,TIAN L,et al.Scalable person re-identification:a benchmark[C]//IEEE International Conference on Computer Vision.Piscataway:IEEE Press,2015:1116-1124.
[21] FU Y,WEI Y C,WANG G S,et al.Self-similarity grouping:a simple unsupervised cross domain adaptation approach for person re-identification[C]//IEEE/CVF International Conference on Computer Vision.Piscataway:IEEE Press,2019:6111-6120.
[22] CHO Y,KIM W J,HONG S,et al.Part-based pseudo label refinement for unsupervised person re-identification[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2022:7298-7308.
基本信息:
DOI：10.13705/j.issn.1671-6841.2024031
中图分类号:TP391.41;TP18
引用信息:
[1]马建红,靳岩,王亚辉,等.基于双分支注意力特征融合的跨域行人重识别[J].郑州大学学报(理学版),2025,57(06):51-57.DOI:10.13705/j.issn.1671-6841.2024031.
基金信息:
国家重点研发计划项目(2020YFB171240); 郑州市协同创新重大专项(20XTZX06013)