{"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/boxsup-exploiting-bounding-boxes-to-supervise","title":"BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation","arxiv_id":"1503.01640","date":"2015-03-05","proceeding":"ICCV 2015 12","authors":["Jifeng Dai","Kaiming He","Jian Sun"],"abstract":"Recent leading approaches to semantic segmentation rely on deep convolutional\nnetworks trained with human-annotated, pixel-level segmentation masks. Such\npixel-accurate supervision demands expensive labeling effort and limits the\nperformance of deep networks that usually benefit from more training data. In\nthis paper, we propose a method that achieves competitive accuracy but only\nrequires easily obtained bounding box annotations. The basic idea is to iterate\nbetween automatically generating region proposals and training convolutional\nnetworks. These two steps gradually recover segmentation masks for improving\nthe networks, and vise versa. Our method, called BoxSup, produces competitive\nresults supervised by boxes only, on par with strong baselines fully supervised\nby masks under the same setting. By leveraging a large amount of bounding\nboxes, BoxSup further unleashes the power of deep convolutional networks and\nyields state-of-the-art results on PASCAL VOC 2012 and PASCAL-CONTEXT.","url_abs":"http://arxiv.org/abs/1503.01640v2","url_pdf":"http://arxiv.org/pdf/1503.01640v2.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":[],"tasks":[{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semantic-segmentation-on-pascal-context","task":"Semantic Segmentation","dataset":"PASCAL Context","model":"BoxSup","rank_in_archive_order":60,"of":66,"metrics":{"mIoU":"40.5"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-pascal-voc-2012","task":"Semantic Segmentation","dataset":"PASCAL VOC 2012 test","model":"BoxSup","rank_in_archive_order":46,"of":51,"metrics":{"Mean IoU":"64.6%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1503.01640","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}