{"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/a-fast-rcnn-hard-positive-generation-via","title":"A-Fast-RCNN: Hard Positive Generation via Adversary for Object Detection","arxiv_id":"1704.03414","date":"2017-04-11","proceeding":"CVPR 2017 7","authors":["Xiaolong Wang","Abhinav Shrivastava","Abhinav Gupta"],"abstract":"How do we learn an object detector that is invariant to occlusions and\ndeformations? Our current solution is to use a data-driven strategy -- collect\nlarge-scale datasets which have object instances under different conditions.\nThe hope is that the final classifier can use these examples to learn\ninvariances. But is it really possible to see all the occlusions in a dataset?\nWe argue that like categories, occlusions and object deformations also follow a\nlong-tail. Some occlusions and deformations are so rare that they hardly\nhappen; yet we want to learn a model invariant to such occurrences. In this\npaper, we propose an alternative solution. We propose to learn an adversarial\nnetwork that generates examples with occlusions and deformations. The goal of\nthe adversary is to generate examples that are difficult for the object\ndetector to classify. In our framework both the original detector and adversary\nare learned in a joint manner. Our experimental results indicate a 2.3% mAP\nboost on VOC07 and a 2.6% mAP boost on VOC2012 object detection challenge\ncompared to the Fast-RCNN pipeline. We also release the code for this paper.","url_abs":"http://arxiv.org/abs/1704.03414v1","url_pdf":"http://arxiv.org/pdf/1704.03414v1.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":"a-fast-rcnn-hard-positive-generation-via","repo_url":"https://github.com/xiaolonw/adversarial-frcnn","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"torch","reach":{"status":"unanswered"}},{"paper_slug":"a-fast-rcnn-hard-positive-generation-via","repo_url":"https://github.com/HusterRC/adversarial-frcnn-master","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":{"status":"unanswered"}},{"paper_slug":"a-fast-rcnn-hard-positive-generation-via","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":"a-fast-rcnn-hard-positive-generation-via","repo_url":"https://github.com/xzabg/fast-adversarial","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"torch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/object-detection-on-pascal-voc-2007","task":"Object Detection","dataset":"PASCAL VOC 2007","model":"FRCN","rank_in_archive_order":20,"of":30,"metrics":{"MAP":"74.2%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1704.03414","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}