{"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/part-aligned-bilinear-representations-for","title":"Part-Aligned Bilinear Representations for Person Re-identification","arxiv_id":"1804.07094","date":"2018-04-19","proceeding":"ECCV 2018 9","authors":["Yumin Suh","Jingdong Wang","Siyu Tang","Tao Mei","Kyoung Mu Lee"],"abstract":"We propose a novel network that learns a part-aligned representation for\nperson re-identification. It handles the body part misalignment problem, that\nis, body parts are misaligned across human detections due to pose/viewpoint\nchange and unreliable detection. Our model consists of a two-stream network\n(one stream for appearance map extraction and the other one for body part map\nextraction) and a bilinear-pooling layer that generates and spatially pools a\npart-aligned map. Each local feature of the part-aligned map is obtained by a\nbilinear mapping of the corresponding local appearance and body part\ndescriptors. Our new representation leads to a robust image matching\nsimilarity, which is equivalent to an aggregation of the local similarities of\nthe corresponding body parts combined with the weighted appearance similarity.\nThis part-aligned representation reduces the part misalignment problem\nsignificantly. Our approach is also advantageous over other pose-guided\nrepresentations (e.g., extracting representations over the bounding box of each\nbody part) by learning part descriptors optimal for person re-identification.\nFor training the network, our approach does not require any part annotation on\nthe person re-identification dataset. Instead, we simply initialize the part\nsub-stream using a pre-trained sub-network of an existing pose estimation\nnetwork, and train the whole network to minimize the re-identification loss. We\nvalidate the effectiveness of our approach by demonstrating its superiority\nover the state-of-the-art methods on the standard benchmark datasets, including\nMarket-1501, CUHK03, CUHK01 and DukeMTMC, and standard video dataset MARS.","url_abs":"http://arxiv.org/abs/1804.07094v1","url_pdf":"http://arxiv.org/pdf/1804.07094v1.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":[],"tasks":[{"task_slug":"2d-human-pose-estimation","task_name":"2D Human Pose Estimation"},{"task_slug":"person-re-identification","task_name":"Person Re-Identification"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/person-re-identification-on-uav-human","task":"Person Re-Identification","dataset":"UAV-Human","model":"Part-Aligned","rank_in_archive_order":4,"of":4,"metrics":{" Rank-1":"60.86"," Rank-5":"81.71","mAP":"60.86"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.07094","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}