{"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/instance-aware-semantic-segmentation-via","title":"Instance-aware Semantic Segmentation via Multi-task Network Cascades","arxiv_id":"1512.04412","date":"2015-12-14","proceeding":"CVPR 2016 6","authors":["Jifeng Dai","Kaiming He","Jian Sun"],"abstract":"Semantic segmentation research has recently witnessed rapid progress, but\nmany leading methods are unable to identify object instances. In this paper, we\npresent Multi-task Network Cascades for instance-aware semantic segmentation.\nOur model consists of three networks, respectively differentiating instances,\nestimating masks, and categorizing objects. These networks form a cascaded\nstructure, and are designed to share their convolutional features. We develop\nan algorithm for the nontrivial end-to-end training of this causal, cascaded\nstructure. Our solution is a clean, single-step training framework and can be\ngeneralized to cascades that have more stages. We demonstrate state-of-the-art\ninstance-aware semantic segmentation accuracy on PASCAL VOC. Meanwhile, our\nmethod takes only 360ms testing an image using VGG-16, which is two orders of\nmagnitude faster than previous systems for this challenging problem. As a by\nproduct, our method also achieves compelling object detection results which\nsurpass the competitive Fast/Faster R-CNN systems.\n  The method described in this paper is the foundation of our submissions to\nthe MS COCO 2015 segmentation competition, where we won the 1st place.","url_abs":"http://arxiv.org/abs/1512.04412v1","url_pdf":"http://arxiv.org/pdf/1512.04412v1.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":"instance-aware-semantic-segmentation-via","repo_url":"https://github.com/daijifeng001/MNC","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"caffe2","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"instance-aware-semantic-segmentation-via","repo_url":"https://github.com/jfc4050/detect-to-track","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"multi-human-parsing","task_name":"Multi-Human Parsing"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"roiwarp","method_name":"RoIWarp"}],"datasets_introduced":[],"methods_introduced":[{"slug":"roiwarp","name":"RoIWarp","full_name":"RoIWarp"}],"results":[{"leaderboard":"/sota/instance-segmentation-on-coco","task":"Instance Segmentation","dataset":"COCO test-dev","model":"MNC","rank_in_archive_order":111,"of":112,"metrics":{"AP50":"44.3%"},"uses_additional_data":false},{"leaderboard":"/sota/multi-human-parsing-on-pascal-person-part","task":"Multi-Human Parsing","dataset":"PASCAL-Part","model":"MNC","rank_in_archive_order":3,"of":3,"metrics":{"AP 0.5":"38.80%"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1512.04412","atlas_url":"https://app.syntology.ai/?focus=1512.04412","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}