{"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/aggregating-deep-convolutional-features-for","title":"Aggregating Deep Convolutional Features for Image Retrieval","arxiv_id":"1510.07493","date":"2015-10-26","proceeding":null,"authors":["Artem Babenko","Victor Lempitsky"],"abstract":"Several recent works have shown that image descriptors produced by deep\nconvolutional neural networks provide state-of-the-art performance for image\nclassification and retrieval problems. It has also been shown that the\nactivations from the convolutional layers can be interpreted as local features\ndescribing particular image regions. These local features can be aggregated\nusing aggregation approaches developed for local features (e.g. Fisher\nvectors), thus providing new powerful global descriptors.\n  In this paper we investigate possible ways to aggregate local deep features\nto produce compact global descriptors for image retrieval. First, we show that\ndeep features and traditional hand-engineered features have quite different\ndistributions of pairwise similarities, hence existing aggregation methods have\nto be carefully re-evaluated. Such re-evaluation reveals that in contrast to\nshallow features, the simple aggregation method based on sum pooling provides\narguably the best performance for deep convolutional features. This method is\nefficient, has few parameters, and bears little risk of overfitting when e.g.\nlearning the PCA matrix. Overall, the new compact global descriptor improves\nthe state-of-the-art on four common benchmarks considerably.","url_abs":"http://arxiv.org/abs/1510.07493v1","url_pdf":"http://arxiv.org/pdf/1510.07493v1.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":"aggregating-deep-convolutional-features-for","repo_url":"https://github.com/rui-yan/CS229-final-project","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"aggregating-deep-convolutional-features-for","repo_url":"https://github.com/talal579/Deep-image-matching","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"pca","method_name":"PCA"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"R – [O] –SPoC","rank_in_archive_order":23,"of":23,"metrics":{"mAP":"12.4"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"R – [O] –SPoC","rank_in_archive_order":22,"of":23,"metrics":{"mAP":"39.8"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"R – [O] –SPoC","rank_in_archive_order":16,"of":23,"metrics":{"mAP":"69.2"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1510.07493","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}