{"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/discobox-weakly-supervised-instance","title":"DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision","arxiv_id":"2105.06464","date":"2021-05-13","proceeding":"ICCV 2021 10","authors":["Shiyi Lan","Zhiding Yu","Christopher Choy","Subhashree Radhakrishnan","Guilin Liu","Yuke Zhu","Larry S. Davis","Anima Anandkumar"],"abstract":"We introduce DiscoBox, a novel framework that jointly learns instance segmentation and semantic correspondence using bounding box supervision. Specifically, we propose a self-ensembling framework where instance segmentation and semantic correspondence are jointly guided by a structured teacher in addition to the bounding box supervision. The teacher is a structured energy model incorporating a pairwise potential and a cross-image potential to model the pairwise pixel relationships both within and across the boxes. Minimizing the teacher energy simultaneously yields refined object masks and dense correspondences between intra-class objects, which are taken as pseudo-labels to supervise the task network and provide positive/negative correspondence pairs for dense constrastive learning. We show a symbiotic relationship where the two tasks mutually benefit from each other. Our best model achieves 37.9% AP on COCO instance segmentation, surpassing prior weakly supervised methods and is competitive to supervised methods. We also obtain state of the art weakly supervised results on PASCAL VOC12 and PF-PASCAL with real-time inference.","url_abs":"https://arxiv.org/abs/2105.06464v2","url_pdf":"https://arxiv.org/pdf/2105.06464v2.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":"discobox-weakly-supervised-instance","repo_url":"https://github.com/NVlabs/DiscoBox","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"discobox-weakly-supervised-instance","repo_url":"https://github.com/voidrank/DiscoBox","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"discobox-weakly-supervised-instance","repo_url":"https://github.com/voidrank/SCOT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"box-supervised-instance-segmentation","task_name":"Box-supervised Instance Segmentation"},{"task_slug":"instance-segmentation","task_name":"Instance Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-correspondence","task_name":"Semantic correspondence"},{"task_slug":"weakly-supervised-instance-segmentation","task_name":"Weakly-supervised instance segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/box-supervised-instance-segmentation-on-coco","task":"Box-supervised Instance Segmentation","dataset":"COCO test-dev","model":"DiscoBox","rank_in_archive_order":3,"of":7,"metrics":{"mask AP":"37.9"},"uses_additional_data":false},{"leaderboard":"/sota/box-supervised-instance-segmentation-on","task":"Box-supervised Instance Segmentation","dataset":"PASCAL VOC 2012 val","model":"DiscoBox","rank_in_archive_order":5,"of":5,"metrics":{"AP_25":"75.2","AP_50":"63.6","AP_70":"45.5","AP_75":"37.5"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-instance-segmentation-on-1","task":"Weakly-supervised instance segmentation","dataset":"COCO 2017 val","model":"DiscoBox (ResNet-50)","rank_in_archive_order":1,"of":2,"metrics":{"AP":"31.4","AP@50":"52.6","AP@75":"32.2","AP@L":"50.1","AP@M":"33.8","AP@S":"11.5"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-instance-segmentation-on-2","task":"Weakly-supervised instance segmentation","dataset":"COCO test-dev","model":"DiscoBox (ResNeXt-101-DCN-FPN)","rank_in_archive_order":1,"of":7,"metrics":{"AP":"37.9","AP@50":"61.4","AP@75":"40.0","AP@L":"53.9","AP@M":"41.1","AP@S":"18.0"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-instance-segmentation-on-2","task":"Weakly-supervised instance segmentation","dataset":"COCO test-dev","model":"DiscoBox (ResNet-101-DCN-FPN)","rank_in_archive_order":2,"of":7,"metrics":{"AP":"35.8","AP@50":"59.8","AP@75":"36.4","AP@L":"52.1","AP@M":"38.7","AP@S":"16.9"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-instance-segmentation-on-2","task":"Weakly-supervised instance segmentation","dataset":"COCO test-dev","model":"DiscoBox (ResNet-50-FPN)","rank_in_archive_order":7,"of":7,"metrics":{"AP":"32.0","AP@50":"53.6","AP@75":"32.6","AP@L":"48.4","AP@M":"33.7","AP@S":"11.7"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2105.06464","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}