{"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/learning-deep-features-via-congenerous-cosine","title":"Learning Deep Features via Congenerous Cosine Loss for Person Recognition","arxiv_id":"1702.06890","date":"2017-02-22","proceeding":null,"authors":["Yu Liu","Hongyang Li","Xiaogang Wang"],"abstract":"Person recognition aims at recognizing the same identity across time and\nspace with complicated scenes and similar appearance. In this paper, we propose\na novel method to address this task by training a network to obtain robust and\nrepresentative features. The intuition is that we directly compare and optimize\nthe cosine distance between two features - enlarging inter-class distinction as\nwell as alleviating inner-class variance. We propose a congenerous cosine loss\nby minimizing the cosine distance between samples and their cluster centroid in\na cooperative way. Such a design reduces the complexity and could be\nimplemented via softmax with normalized inputs. Our method also differs from\nprevious work in person recognition that we do not conduct a second training on\nthe test subset. The identity of a person is determined by measuring the\nsimilarity from several body regions in the reference set. Experimental results\nshow that the proposed approach achieves better classification accuracy against\nprevious state-of-the-arts.","url_abs":"http://arxiv.org/abs/1702.06890v2","url_pdf":"http://arxiv.org/pdf/1702.06890v2.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":"learning-deep-features-via-congenerous-cosine","repo_url":"https://github.com/sciencefans/coco_loss","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"person-recognition","task_name":"Person Recognition"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1702.06890","atlas_url":"https://app.syntology.ai/?focus=1702.06890","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}