{"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/pedestrian-alignment-network-for-large-scale","title":"Pedestrian Alignment Network for Large-scale Person Re-identification","arxiv_id":"1707.00408","date":"2017-07-03","proceeding":null,"authors":["Zhedong Zheng","Liang Zheng","Yi Yang"],"abstract":"Person re-identification (person re-ID) is mostly viewed as an image\nretrieval problem. This task aims to search a query person in a large image\npool. In practice, person re-ID usually adopts automatic detectors to obtain\ncropped pedestrian images. However, this process suffers from two types of\ndetector errors: excessive background and part missing. Both errors deteriorate\nthe quality of pedestrian alignment and may compromise pedestrian matching due\nto the position and scale variances. To address the misalignment problem, we\npropose that alignment can be learned from an identification procedure. We\nintroduce the pedestrian alignment network (PAN) which allows discriminative\nembedding learning and pedestrian alignment without extra annotations. Our key\nobservation is that when the convolutional neural network (CNN) learns to\ndiscriminate between different identities, the learned feature maps usually\nexhibit strong activations on the human body rather than the background. The\nproposed network thus takes advantage of this attention mechanism to adaptively\nlocate and align pedestrians within a bounding box. Visual examples show that\npedestrians are better aligned with PAN. Experiments on three large-scale re-ID\ndatasets confirm that PAN improves the discriminative ability of the feature\nembeddings and yields competitive accuracy with the state-of-the-art methods.","url_abs":"http://arxiv.org/abs/1707.00408v1","url_pdf":"http://arxiv.org/pdf/1707.00408v1.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":"pedestrian-alignment-network-for-large-scale","repo_url":"https://github.com/layumi/Pedestrian_Alignment","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"large-scale-person-re-identification","task_name":"Large-Scale Person Re-Identification"},{"task_slug":"person-re-identification","task_name":"Person Re-Identification"},{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/person-re-identification-on-cuhk03-detected-1","task":"Person Re-Identification","dataset":"CUHK03 (detected)","model":"PAN+re-rank","rank_in_archive_order":1,"of":2,"metrics":{"MAP":"43.8","Rank-1":"41.9"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-cuhk03-detected-1","task":"Person Re-Identification","dataset":"CUHK03 (detected)","model":"PAN(Zheng et al., [2017a])","rank_in_archive_order":2,"of":2,"metrics":{"MAP":"34","Rank-1":"36.3"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-cuhk03-labeled","task":"Person Re-Identification","dataset":"CUHK03 labeled","model":"PAN+re-rank","rank_in_archive_order":15,"of":21,"metrics":{"MAP":"45.8","Rank-1":"43.9"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-cuhk03-labeled","task":"Person Re-Identification","dataset":"CUHK03 labeled","model":"PAN(Zheng et al., [2017a])","rank_in_archive_order":17,"of":21,"metrics":{"MAP":"35.0","Rank-1":"36.9"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-dukemtmc-reid","task":"Person Re-Identification","dataset":"DukeMTMC-reID","model":"PAN + re-rank","rank_in_archive_order":69,"of":94,"metrics":{"Rank-1":"75.94","mAP":"66.74"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-dukemtmc-reid","task":"Person Re-Identification","dataset":"DukeMTMC-reID","model":"PAN","rank_in_archive_order":81,"of":94,"metrics":{"Rank-1":"71.59","mAP":"51.51"},"uses_additional_data":false},{"leaderboard":"/sota/person-re-identification-on-market-1501","task":"Person Re-Identification","dataset":"Market-1501","model":"PAN (GAN)+re-rank","rank_in_archive_order":98,"of":135,"metrics":{"Rank-1":"88.57","mAP":"81.53"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1707.00408","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}