{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/unsupervised-cross-dataset-person-re","title":"Unsupervised Cross-dataset Person Re-identification by Transfer Learning of Spatial-Temporal Patterns","arxiv_id":"1803.07293","date":"2018-03-20","proceeding":"CVPR 2018 6","authors":["Jianming Lv","Weihang Chen","Qing Li","Can Yang"],"abstract":"Most of the proposed person re-identification algorithms conduct supervised\ntraining and testing on single labeled datasets with small size, so directly\ndeploying these trained models to a large-scale real-world camera network may\nlead to poor performance due to underfitting. It is challenging to\nincrementally optimize the models by using the abundant unlabeled data\ncollected from the target domain. To address this challenge, we propose an\nunsupervised incremental learning algorithm, TFusion, which is aided by the\ntransfer learning of the pedestrians' spatio-temporal patterns in the target\ndomain. Specifically, the algorithm firstly transfers the visual classifier\ntrained from small labeled source dataset to the unlabeled target dataset so as\nto learn the pedestrians' spatial-temporal patterns. Secondly, a Bayesian\nfusion model is proposed to combine the learned spatio-temporal patterns with\nvisual features to achieve a significantly improved classifier. Finally, we\npropose a learning-to-rank based mutual promotion procedure to incrementally\noptimize the classifiers based on the unlabeled data in the target domain.\nComprehensive experiments based on multiple real surveillance datasets are\nconducted, and the results show that our algorithm gains significant\nimprovement compared with the state-of-art cross-dataset unsupervised person\nre-identification algorithms.","url_abs":"http://arxiv.org/abs/1803.07293v1","url_pdf":"http://arxiv.org/pdf/1803.07293v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"unsupervised-cross-dataset-person-re","repo_url":"https://github.com/ahangchen/TFusion","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"incremental-learning","task_name":"Incremental Learning"},{"task_slug":"learning-to-rank","task_name":"Learning-To-Rank"},{"task_slug":"person-re-identification","task_name":"Person Re-Identification"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"unsupervised-person-re-identification","task_name":"Unsupervised Person Re-Identification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1803.07293","atlas_url":"https://app.syntology.ai/?focus=1803.07293","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}