{"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/range-loss-for-deep-face-recognition-with","title":"Range Loss for Deep Face Recognition with Long-tail","arxiv_id":"1611.08976","date":"2016-11-28","proceeding":null,"authors":["Xiao Zhang","Zhiyuan Fang","Yandong Wen","Zhifeng Li","Yu Qiao"],"abstract":"Convolutional neural networks have achieved great improvement on face\nrecognition in recent years because of its extraordinary ability in learning\ndiscriminative features of people with different identities. To train such a\nwell-designed deep network, tremendous amounts of data is indispensable. Long\ntail distribution specifically refers to the fact that a small number of\ngeneric entities appear frequently while other objects far less existing.\nConsidering the existence of long tail distribution of the real world data,\nlarge but uniform distributed data are usually hard to retrieve. Empirical\nexperiences and analysis show that classes with more samples will pose greater\nimpact on the feature learning process and inversely cripple the whole models\nfeature extracting ability on tail part data. Contrary to most of the existing\nworks that alleviate this problem by simply cutting the tailed data for uniform\ndistributions across the classes, this paper proposes a new loss function\ncalled range loss to effectively utilize the whole long tailed data in training\nprocess. More specifically, range loss is designed to reduce overall\nintra-personal variations while enlarging inter-personal differences within one\nmini-batch simultaneously when facing even extremely unbalanced data. The\noptimization objective of range loss is the $k$ greatest range's harmonic mean\nvalues in one class and the shortest inter-class distance within one batch.\nExtensive experiments on two famous and challenging face recognition benchmarks\n(Labeled Faces in the Wild (LFW) and YouTube Faces (YTF) not only demonstrate\nthe effectiveness of the proposed approach in overcoming the long tail effect\nbut also show the good generalization ability of the proposed approach.","url_abs":"http://arxiv.org/abs/1611.08976v1","url_pdf":"http://arxiv.org/pdf/1611.08976v1.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":"range-loss-for-deep-face-recognition-with","repo_url":"https://github.com/DongDem/RangeLoss_MarginalLoss_Tensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"range-loss-for-deep-face-recognition-with","repo_url":"https://github.com/DongDem/RangeLoss_Marginal_Loss_Tensorflow","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"face-recognition","task_name":"Face Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1611.08976","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}