{"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/revisiting-oxford-and-paris-large-scale-image","title":"Revisiting Oxford and Paris: Large-Scale Image Retrieval Benchmarking","arxiv_id":"1803.11285","date":"2018-03-29","proceeding":"CVPR 2018 6","authors":["Filip Radenović","Ahmet Iscen","Giorgos Tolias","Yannis Avrithis","Ondřej Chum"],"abstract":"In this paper we address issues with image retrieval benchmarking on standard\nand popular Oxford 5k and Paris 6k datasets. In particular, annotation errors,\nthe size of the dataset, and the level of challenge are addressed: new\nannotation for both datasets is created with an extra attention to the\nreliability of the ground truth. Three new protocols of varying difficulty are\nintroduced. The protocols allow fair comparison between different methods,\nincluding those using a dataset pre-processing stage. For each dataset, 15 new\nchallenging queries are introduced. Finally, a new set of 1M hard,\nsemi-automatically cleaned distractors is selected.\n  An extensive comparison of the state-of-the-art methods is performed on the\nnew benchmark. Different types of methods are evaluated, ranging from\nlocal-feature-based to modern CNN based methods. The best results are achieved\nby taking the best of the two worlds. Most importantly, image retrieval appears\nfar from being solved.","url_abs":"http://arxiv.org/abs/1803.11285v1","url_pdf":"http://arxiv.org/pdf/1803.11285v1.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":"revisiting-oxford-and-paris-large-scale-image","repo_url":"https://github.com/tensorflow/models","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"revisiting-oxford-and-paris-large-scale-image","repo_url":"https://github.com/tensorflow/models/tree/master/research/delf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"benchmarking","task_name":"Benchmarking"},{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[],"datasets_introduced":[{"slug":"roxford","name":"ROxford","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–HQE+SP","rank_in_archive_order":10,"of":23,"metrics":{"mAP":"49.7"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–HQE","rank_in_archive_order":12,"of":23,"metrics":{"mAP":"41.3"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–ASMK*+SP","rank_in_archive_order":14,"of":23,"metrics":{"mAP":"36.7"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–ASMK*","rank_in_archive_order":15,"of":23,"metrics":{"mAP":"36.4 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–SMK*+SP","rank_in_archive_order":16,"of":23,"metrics":{"mAP":"35.8 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–SMK*","rank_in_archive_order":17,"of":23,"metrics":{"mAP":"35.4 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"HesAff–rSIFT–VLAD","rank_in_archive_order":22,"of":23,"metrics":{"mAP":"13.2"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–HQE+SP","rank_in_archive_order":9,"of":23,"metrics":{"mAP":"71.3 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–HQE","rank_in_archive_order":11,"of":23,"metrics":{"mAP":"66.3 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–ASMK*+SP","rank_in_archive_order":15,"of":23,"metrics":{"mAP":"60.6"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–ASMK*","rank_in_archive_order":16,"of":23,"metrics":{"mAP":"60.4 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–SMK*+SP","rank_in_archive_order":17,"of":23,"metrics":{"mAP":"59.8 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–SMK*","rank_in_archive_order":18,"of":23,"metrics":{"mAP":"59.4"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"HesAff–rSIFT–VLAD","rank_in_archive_order":23,"of":23,"metrics":{"mAP":"33.9 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium-without","task":"Image Retrieval","dataset":"ROxford Medium without fine-tuning","model":"HesAff–rSIFT–VLAD","rank_in_archive_order":1,"of":1,"metrics":{"Average mAP":"33.9"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–HQE+SP","rank_in_archive_order":15,"of":23,"metrics":{"mAP":"45.1"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–HQE","rank_in_archive_order":17,"of":23,"metrics":{"mAP":"44.7"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–ASMK*+SP","rank_in_archive_order":19,"of":23,"metrics":{"mAP":"35.0"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–ASMK*","rank_in_archive_order":20,"of":23,"metrics":{"mAP":"34.5"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–SMK*+SP","rank_in_archive_order":21,"of":23,"metrics":{"mAP":"31.3 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–SMK*","rank_in_archive_order":22,"of":23,"metrics":{"mAP":"31.2 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"HesAff–rSIFT–VLAD","rank_in_archive_order":23,"of":23,"metrics":{"mAP":"17.5 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–HQE+SP","rank_in_archive_order":15,"of":23,"metrics":{"mAP":"70.2"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–HQE","rank_in_archive_order":17,"of":23,"metrics":{"mAP":"68.9 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–ASMK*+SP","rank_in_archive_order":19,"of":23,"metrics":{"mAP":"61.4"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–ASMK*","rank_in_archive_order":20,"of":23,"metrics":{"mAP":"61.2 "},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–SMK*+SP","rank_in_archive_order":21,"of":23,"metrics":{"mAP":"59.2"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–SMK*","rank_in_archive_order":22,"of":23,"metrics":{"mAP":"59.0"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"HesAff–rSIFT–VLAD","rank_in_archive_order":23,"of":23,"metrics":{"mAP":"43.6"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1803.11285","atlas_url":"https://app.syntology.ai/?focus=1803.11285","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}