{"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/fastfcn-rethinking-dilated-convolution-in-the","title":"FastFCN: Rethinking Dilated Convolution in the Backbone for Semantic Segmentation","arxiv_id":"1903.11816","date":"2019-03-28","proceeding":null,"authors":["Huikai Wu","Junge Zhang","Kaiqi Huang","Kongming Liang","Yizhou Yu"],"abstract":"Modern approaches for semantic segmentation usually employ dilated\nconvolutions in the backbone to extract high-resolution feature maps, which\nbrings heavy computation complexity and memory footprint. To replace the time\nand memory consuming dilated convolutions, we propose a novel joint upsampling\nmodule named Joint Pyramid Upsampling (JPU) by formulating the task of\nextracting high-resolution feature maps into a joint upsampling problem. With\nthe proposed JPU, our method reduces the computation complexity by more than\nthree times without performance loss. Experiments show that JPU is superior to\nother upsampling modules, which can be plugged into many existing approaches to\nreduce computation complexity and improve performance. By replacing dilated\nconvolutions with the proposed JPU module, our method achieves the\nstate-of-the-art performance in Pascal Context dataset (mIoU of 53.13%) and\nADE20K dataset (final score of 0.5584) while running 3 times faster.","url_abs":"http://arxiv.org/abs/1903.11816v1","url_pdf":"http://arxiv.org/pdf/1903.11816v1.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":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/wuhuikai/FastFCN","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/2anchao/VovJpu","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/DongjuSin/computer-vision-proj","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/yiskw713/fastfcn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/yougoforward/Fast_psaa","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/yougoforward/GSNet","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/yougoforward/can","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/yougoforward/cfpn_gsf","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/yougoforward/localUp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/PaddlePaddle/PaddleSeg","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/justld/FastFCN_paddle","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null},{"paper_slug":"fastfcn-rethinking-dilated-convolution-in-the","repo_url":"https://github.com/srihari-humbarwadi/FastFCN_TF2.0","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semantic-segmentation-on-ade20k","task":"Semantic Segmentation","dataset":"ADE20K","model":"EncNet + JPU","rank_in_archive_order":202,"of":235,"metrics":{"Test Score":"55.84","Validation mIoU":"44.34"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-pascal-context","task":"Semantic Segmentation","dataset":"PASCAL Context","model":"Joint Pyramid Upsampling + EncNet","rank_in_archive_order":43,"of":66,"metrics":{"mIoU":"53.1"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1903.11816","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}