{"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/training-region-based-object-detectors-with","title":"Training Region-based Object Detectors with Online Hard Example Mining","arxiv_id":"1604.03540","date":"2016-04-12","proceeding":"CVPR 2016 6","authors":["Abhinav Shrivastava","Abhinav Gupta","Ross Girshick"],"abstract":"The field of object detection has made significant advances riding on the\nwave of region-based ConvNets, but their training procedure still includes many\nheuristics and hyperparameters that are costly to tune. We present a simple yet\nsurprisingly effective online hard example mining (OHEM) algorithm for training\nregion-based ConvNet detectors. Our motivation is the same as it has always\nbeen -- detection datasets contain an overwhelming number of easy examples and\na small number of hard examples. Automatic selection of these hard examples can\nmake training more effective and efficient. OHEM is a simple and intuitive\nalgorithm that eliminates several heuristics and hyperparameters in common use.\nBut more importantly, it yields consistent and significant boosts in detection\nperformance on benchmarks like PASCAL VOC 2007 and 2012. Its effectiveness\nincreases as datasets become larger and more difficult, as demonstrated by the\nresults on the MS COCO dataset. Moreover, combined with complementary advances\nin the field, OHEM leads to state-of-the-art results of 78.9% and 76.3% mAP on\nPASCAL VOC 2007 and 2012 respectively.","url_abs":"http://arxiv.org/abs/1604.03540v1","url_pdf":"http://arxiv.org/pdf/1604.03540v1.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":"training-region-based-object-detectors-with","repo_url":"https://github.com/Bennie-Han/Image-augementation-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"gone","observed_at":"2026-09-17","how":"tree_404+repo_404"}},{"paper_slug":"training-region-based-object-detectors-with","repo_url":"https://github.com/abhi2610/ohem","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"caffe2","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"training-region-based-object-detectors-with","repo_url":"https://github.com/busyboxs/Some-resources-useful-for-me","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"training-region-based-object-detectors-with","repo_url":"https://github.com/hh-xiaohu/Image-augementation-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"gone","observed_at":"2026-09-17","how":"tree_404+repo_404"}},{"paper_slug":"training-region-based-object-detectors-with","repo_url":"https://github.com/tkuanlun350/Kaggle_Ship_Detection_2018","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"ohem","method_name":"OHEM"}],"datasets_introduced":[],"methods_introduced":[{"slug":"ohem","name":"OHEM","full_name":"Online Hard Example Mining"}],"results":[{"leaderboard":"/sota/face-identification-on-trillion-pairs-dataset","task":"Face Identification","dataset":"Trillion Pairs Dataset","model":"HM-Softmax","rank_in_archive_order":6,"of":6,"metrics":{"Accuracy":"36.75"},"uses_additional_data":false},{"leaderboard":"/sota/face-verification-on-trillion-pairs-dataset","task":"Face Verification","dataset":"Trillion Pairs Dataset","model":"HM-Softmax","rank_in_archive_order":6,"of":6,"metrics":{"Accuracy":"34.46"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-pascal-voc-2007","task":"Object Detection","dataset":"PASCAL VOC 2007","model":"OHEM","rank_in_archive_order":14,"of":30,"metrics":{"MAP":"78.9%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1604.03540","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}