One-Shot Unsupervised Cross-Domain Person Re-Identification

被引:7
作者
Han, Guangxing [1 ,2 ]
Zhang, Xuan [1 ,2 ]
Li, Chongrong [1 ,2 ]
机构
[1] Tsinghua Univ, Inst Network Sci & Cyberspace INSC, Beijing 100084, Peoples R China
[2] Zhongguancun Lab, Beijing 100081, Peoples R China
关键词
Training; Adaptation models; Task analysis; Testing; Representation learning; Data models; Training data; Person re-identification; unsupervised domain adaptation; domain generalization; unsupervised image-to-image translation; ATTENTION; NETWORK;
D O I
10.1109/TCSVT.2023.3293130
中图分类号
TM [电工技术]; TN [电子技术、通信技术];
学科分类号
0808 ; 0809 ;
摘要
Cross-domain person re-identification is challenging due to the notorious domain shift problem. Most of the existing unsupervised cross-domain person ReID methods require a large number of unlabeled target-domain samples for adaptation. However, large scale of training data are not always available due to public privacy. Domain generalization methods have inferior adaptation ability without seeing any target domain data. Inspired by the few-shot learning capability of human vision system, we propose a novel setting, one-shot unsupervised cross-domain for person ReID and study the ability of adaptation using the minimum number of image in the target domain during training. Specifically, we first propose a novel Group Normalization (GN) based domain generalizable ReID model. We show that the GN based model could strike a better balance between model discrimination and generalization ability, compared with the Batch Normalization (BN) and Instance Normalization (IN) counterparts, and is more suitable for domain generalizable ReID baseline model. Then besides the supervised feature learning task in the source domain, we introduce two self-supervised learning tasks using the one-shot target domain data to further improve the generalization ability of the ReID model. We carefully design model architecture and perform model training to reduce overfitting to the one-shot target domain. Extensive experiments demonstrate the effectiveness of our approach for one-shot unsupervised cross-domain ReID. Our approach can be extended to few-shot setting and increasing the number of shot up to 1,000 images can steadily increase the performance, which provides practical values to the community.
引用
收藏
页码:1339 / 1351
页数:13
相关论文
共 103 条
[61]   Median Stable Clustering and Global Distance Classification for Cross-Domain Person Re-Identification [J].
Pang, Zhiqi ;
Guo, Jifeng ;
Ma, Zhiqiang ;
Sun, Wenbo ;
Xiao, Yanbang .
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, 2022, 32 (05) :3164-3177
[62]   Contrastive Learning for Unpaired Image-to-Image Translation [J].
Park, Taesung ;
Efros, Alexei A. ;
Zhang, Richard ;
Zhu, Jun-Yan .
COMPUTER VISION - ECCV 2020, PT IX, 2020, 12354 :319-345
[63]  
Radford A, 2016, Arxiv, DOI arXiv:1511.06434
[64]   Performance Measures and a Data Set for Multi-target, Multi-camera Tracking [J].
Ristani, Ergys ;
Solera, Francesco ;
Zou, Roger ;
Cucchiara, Rita ;
Tomasi, Carlo .
COMPUTER VISION - ECCV 2016 WORKSHOPS, PT II, 2016, 9914 :17-35
[65]   U-Net: Convolutional Networks for Biomedical Image Segmentation [J].
Ronneberger, Olaf ;
Fischer, Philipp ;
Brox, Thomas .
MEDICAL IMAGE COMPUTING AND COMPUTER-ASSISTED INTERVENTION, PT III, 2015, 9351 :234-241
[66]  
Schroff F, 2015, PROC CVPR IEEE, P815, DOI 10.1109/CVPR.2015.7298682
[67]   SinGAN: Learning a Generative Model from a Single Natural Image [J].
Shaham, Tamar Rott ;
Dekel, Tali ;
Michaeli, Tomer .
2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019), 2019, :4569-4579
[68]  
Shankar S, 2018, Arxiv, DOI arXiv:1804.10745
[69]   Generalizable Person Re-identification by Domain-Invariant Mapping Network [J].
Song, Jifei ;
Yang, Yongxin ;
Song, Yi-Zhe ;
Xiang, Tao ;
Hospedales, Timothy M. .
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2019), 2019, :719-728
[70]   Unsupervised domain adaptive re-identification: Theory and practice [J].
Song, Liangchen ;
Wang, Cheng ;
Zhang, Lefei ;
Du, Bo ;
Zhang, Qian ;
Huang, Chang ;
Wang, Xinggang .
PATTERN RECOGNITION, 2020, 102