{"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/ghostvlad-for-set-based-face-recognition","title":"GhostVLAD for set-based face recognition","arxiv_id":"1810.09951","date":"2018-10-23","proceeding":null,"authors":["Yujie Zhong","Relja Arandjelović","Andrew Zisserman"],"abstract":"The objective of this paper is to learn a compact representation of image\nsets for template-based face recognition. We make the following contributions:\nfirst, we propose a network architecture which aggregates and embeds the face\ndescriptors produced by deep convolutional neural networks into a compact\nfixed-length representation. This compact representation requires minimal\nmemory storage and enables efficient similarity computation. Second, we propose\na novel GhostVLAD layer that includes {\\em ghost clusters}, that do not\ncontribute to the aggregation. We show that a quality weighting on the input\nfaces emerges automatically such that informative images contribute more than\nthose with low quality, and that the ghost clusters enhance the network's\nability to deal with poor quality images. Third, we explore how input feature\ndimension, number of clusters and different training techniques affect the\nrecognition performance. Given this analysis, we train a network that far\nexceeds the state-of-the-art on the IJB-B face recognition dataset. This is\ncurrently one of the most challenging public benchmarks, and we surpass the\nstate-of-the-art on both the identification and verification protocols.","url_abs":"http://arxiv.org/abs/1810.09951v1","url_pdf":"http://arxiv.org/pdf/1810.09951v1.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":"ghostvlad-for-set-based-face-recognition","repo_url":"https://github.com/keetsky/Net_ghostVLAD-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"ghostvlad-for-set-based-face-recognition","repo_url":"https://github.com/taylorlu/ghostvlad-speaker","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"face-recognition","task_name":"Face Recognition"},{"task_slug":"face-verification","task_name":"Face Verification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/face-verification-on-ijb-a","task":"Face Verification","dataset":"IJB-A","model":"SE-GV-4-g1","rank_in_archive_order":3,"of":17,"metrics":{"TAR @ FAR=0.01":"97.2%"},"uses_additional_data":false},{"leaderboard":"/sota/face-verification-on-ijb-b","task":"Face Verification","dataset":"IJB-B","model":"SE-GV-3-g2","rank_in_archive_order":6,"of":12,"metrics":{"TAR @ FAR=0.01":"96.4%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.09951","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}