{"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/light-weight-refinenet-for-real-time-semantic","title":"Light-Weight RefineNet for Real-Time Semantic Segmentation","arxiv_id":"1810.03272","date":"2018-10-08","proceeding":null,"authors":["Vladimir Nekrasov","Chunhua Shen","Ian Reid"],"abstract":"We consider an important task of effective and efficient semantic image\nsegmentation. In particular, we adapt a powerful semantic segmentation\narchitecture, called RefineNet, into the more compact one, suitable even for\ntasks requiring real-time performance on high-resolution inputs. To this end,\nwe identify computationally expensive blocks in the original setup, and propose\ntwo modifications aimed to decrease the number of parameters and floating point\noperations. By doing that, we achieve more than twofold model reduction, while\nkeeping the performance levels almost intact. Our fastest model undergoes a\nsignificant speed-up boost from 20 FPS to 55 FPS on a generic GPU card on\n512x512 inputs with solid 81.1% mean iou performance on the test set of PASCAL\nVOC, while our slowest model with 32 FPS (from original 17 FPS) shows 82.7%\nmean iou on the same dataset. Alternatively, we showcase that our approach is\neasily mixable with light-weight classification networks: we attain 79.2% mean\niou on PASCAL VOC using a model that contains only 3.3M parameters and performs\nonly 9.3B floating point operations.","url_abs":"http://arxiv.org/abs/1810.03272v1","url_pdf":"http://arxiv.org/pdf/1810.03272v1.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":"light-weight-refinenet-for-real-time-semantic","repo_url":"https://github.com/AleksBanbur/EE8204---Real-Time-Multi-Task-Learning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"light-weight-refinenet-for-real-time-semantic","repo_url":"https://github.com/DrSleep/light-weight-refinenet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":null,"task_name":"GPU"},{"task_slug":"image-segmentation","task_name":"Image Segmentation"},{"task_slug":"real-time-semantic-segmentation","task_name":"Real-Time Semantic Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/real-time-semantic-segmentation-on-nyu-depth-1","task":"Real-Time Semantic Segmentation","dataset":"NYU Depth v2","model":"Light-Weight-RefineNet-152","rank_in_archive_order":2,"of":10,"metrics":{"Speed(ms/f)":"36","mIoU":"44.4"},"uses_additional_data":false},{"leaderboard":"/sota/real-time-semantic-segmentation-on-nyu-depth-1","task":"Real-Time Semantic Segmentation","dataset":"NYU Depth v2","model":"Light-Weight-RefineNet-101","rank_in_archive_order":3,"of":10,"metrics":{"Speed(ms/f)":"27","mIoU":"43.6"},"uses_additional_data":false},{"leaderboard":"/sota/real-time-semantic-segmentation-on-nyu-depth-1","task":"Real-Time Semantic Segmentation","dataset":"NYU Depth v2","model":"Light-Weight-RefineNet-50","rank_in_archive_order":8,"of":10,"metrics":{"Speed(ms/f)":"20","mIoU":"41.7"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-nyu-depth-v2","task":"Semantic Segmentation","dataset":"NYU Depth v2","model":"Light-Weight-RefineNet-152","rank_in_archive_order":97,"of":121,"metrics":{"Mean IoU":"44.4%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-nyu-depth-v2","task":"Semantic Segmentation","dataset":"NYU Depth v2","model":"Light-Weight-RefineNet-101","rank_in_archive_order":99,"of":121,"metrics":{"Mean IoU":"43.6%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-nyu-depth-v2","task":"Semantic Segmentation","dataset":"NYU Depth v2","model":"Light-Weight-RefineNet-50","rank_in_archive_order":105,"of":121,"metrics":{"Mean IoU":"41.7%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-pascal-voc-2012","task":"Semantic Segmentation","dataset":"PASCAL VOC 2012 test","model":"Light-Weight-RefineNet-152","rank_in_archive_order":24,"of":51,"metrics":{"Mean IoU":"82.7%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-pascal-voc-2012","task":"Semantic Segmentation","dataset":"PASCAL VOC 2012 test","model":"Light-Weight-RefineNet-101","rank_in_archive_order":27,"of":51,"metrics":{"Mean IoU":"82.0%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-pascal-voc-2012","task":"Semantic Segmentation","dataset":"PASCAL VOC 2012 test","model":"Light-Weight-RefineNet-50","rank_in_archive_order":28,"of":51,"metrics":{"Mean IoU":"81.1%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-pascal-voc-2012","task":"Semantic Segmentation","dataset":"PASCAL VOC 2012 test","model":"Light-Weight-RefineNet-MobileNet-v2","rank_in_archive_order":33,"of":51,"metrics":{"Mean IoU":"79.2%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.03272","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}