{"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/deep-feature-learning-with-relative-distance","title":"Deep Feature Learning with Relative Distance Comparison for Person Re-identification","arxiv_id":"1512.03622","date":"2015-12-11","proceeding":null,"authors":["Shengyong Ding","Liang Lin","Guangrun Wang","Hongyang Chao"],"abstract":"Identifying the same individual across different scenes is an important yet\ndifficult task in intelligent video surveillance. Its main difficulty lies in\nhow to preserve similarity of the same person against large appearance and\nstructure variation while discriminating different individuals. In this paper,\nwe present a scalable distance driven feature learning framework based on the\ndeep neural network for person re-identification, and demonstrate its\neffectiveness to handle the existing challenges. Specifically, given the\ntraining images with the class labels (person IDs), we first produce a large\nnumber of triplet units, each of which contains three images, i.e. one person\nwith a matched reference and a mismatched reference. Treating the units as the\ninput, we build the convolutional neural network to generate the layered\nrepresentations, and follow with the $L2$ distance metric. By means of\nparameter optimization, our framework tends to maximize the relative distance\nbetween the matched pair and the mismatched pair for each triplet unit.\nMoreover, a nontrivial issue arising with the framework is that the triplet\norganization cubically enlarges the number of training triplets, as one image\ncan be involved into several triplet units. To overcome this problem, we\ndevelop an effective triplet generation scheme and an optimized gradient\ndescent algorithm, making the computational load mainly depends on the number\nof original images instead of the number of triplets. On several challenging\ndatabases, our approach achieves very promising results and outperforms other\nstate-of-the-art approaches.","url_abs":"http://arxiv.org/abs/1512.03622v1","url_pdf":"http://arxiv.org/pdf/1512.03622v1.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":"person-re-identification","task_name":"Person Re-Identification"},{"task_slug":null,"task_name":"Triplet"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/person-re-identification-on-sysu-30k","task":"Person Re-Identification","dataset":"SYSU-30k","model":"DF (generalization)","rank_in_archive_order":9,"of":10,"metrics":{" Rank-1":"10.3"},"uses_additional_data":true}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1512.03622","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}