{"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-person-re-identification-by-soft","title":"Unsupervised Person Re-identification by Soft Multilabel Learning","arxiv_id":"1903.06325","date":"2019-03-15","proceeding":"CVPR 2019 6","authors":["Hong-Xing Yu","Wei-Shi Zheng","An-Cong Wu","Xiaowei Guo","Shaogang Gong","Jian-Huang Lai"],"abstract":"Although unsupervised person re-identification (RE-ID) has drawn increasing\nresearch attentions due to its potential to address the scalability problem of\nsupervised RE-ID models, it is very challenging to learn discriminative\ninformation in the absence of pairwise labels across disjoint camera views. To\novercome this problem, we propose a deep model for the soft multilabel learning\nfor unsupervised RE-ID. The idea is to learn a soft multilabel (real-valued\nlabel likelihood vector) for each unlabeled person by comparing (and\nrepresenting) the unlabeled person with a set of known reference persons from\nan auxiliary domain. We propose the soft multilabel-guided hard negative mining\nto learn a discriminative embedding for the unlabeled target domain by\nexploring the similarity consistency of the visual features and the soft\nmultilabels of unlabeled target pairs. Since most target pairs are cross-view\npairs, we develop the cross-view consistent soft multilabel learning to achieve\nthe learning goal that the soft multilabels are consistently good across\ndifferent camera views. To enable effecient soft multilabel learning, we\nintroduce the reference agent learning to represent each reference person by a\nreference agent in a joint embedding. We evaluate our unified deep model on\nMarket-1501 and DukeMTMC-reID. Our model outperforms the state-of-the-art\nunsupervised RE-ID methods by clear margins. Code is available at\nhttps://github.com/KovenYu/MAR.","url_abs":"http://arxiv.org/abs/1903.06325v2","url_pdf":"http://arxiv.org/pdf/1903.06325v2.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-person-re-identification-by-soft","repo_url":"https://github.com/KovenYu/MAR","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"person-re-identification","task_name":"Person Re-Identification"},{"task_slug":"unsupervised-person-re-identification","task_name":"Unsupervised Person Re-Identification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/person-re-identification-on-dukemtmc-reid","task":"Person Re-Identification","dataset":"DukeMTMC-reID","model":"MAR","rank_in_archive_order":84,"of":94,"metrics":{"Rank-1":"79.8","mAP":"48"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-market-1501","task":"Person Re-Identification","dataset":"Market-1501","model":"MAR","rank_in_archive_order":118,"of":135,"metrics":{"Rank-1":"67.7","Rank-5":"81.9","mAP":"40"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1903.06325","atlas_url":"https://app.syntology.ai/?focus=1903.06325","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}