{"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/local-similarity-aware-deep-feature-embedding","title":"Local Similarity-Aware Deep Feature Embedding","arxiv_id":"1610.08904","date":"2016-10-27","proceeding":"NeurIPS 2016 12","authors":["Chen Huang","Chen Change Loy","Xiaoou Tang"],"abstract":"Existing deep embedding methods in vision tasks are capable of learning a\ncompact Euclidean space from images, where Euclidean distances correspond to a\nsimilarity metric. To make learning more effective and efficient, hard sample\nmining is usually employed, with samples identified through computing the\nEuclidean feature distance. However, the global Euclidean distance cannot\nfaithfully characterize the true feature similarity in a complex visual feature\nspace, where the intraclass distance in a high-density region may be larger\nthan the interclass distance in low-density regions. In this paper, we\nintroduce a Position-Dependent Deep Metric (PDDM) unit, which is capable of\nlearning a similarity metric adaptive to local feature structure. The metric\ncan be used to select genuinely hard samples in a local neighborhood to guide\nthe deep embedding learning in an online and robust manner. The new layer is\nappealing in that it is pluggable to any convolutional networks and is trained\nend-to-end. Our local similarity-aware feature embedding not only demonstrates\nfaster convergence and boosted performance on two complex image retrieval\ndatasets, its large margin nature also leads to superior generalization results\nunder the large and open set scenarios of transfer learning and zero-shot\nlearning on ImageNet 2010 and ImageNet-10K datasets.","url_abs":"http://arxiv.org/abs/1610.08904v1","url_pdf":"http://arxiv.org/pdf/1610.08904v1.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":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"zero-shot-learning","task_name":"Zero-Shot Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/metric-learning-on-cub-200-2011","task":"Metric Learning","dataset":"CUB-200-2011","model":"PDDM Quadruplet","rank_in_archive_order":27,"of":30,"metrics":{"R@1":"58.3"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1610.08904","atlas_url":"https://app.syntology.ai/?focus=1610.08904","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}