{"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/fast-exact-and-multi-scale-inference-for","title":"Fast, Exact and Multi-Scale Inference for Semantic Image Segmentation with Deep Gaussian CRFs","arxiv_id":"1603.08358","date":"2016-03-28","proceeding":null,"authors":["Siddhartha Chandra","Iasonas Kokkinos"],"abstract":"In this work we propose a structured prediction technique that combines the\nvirtues of Gaussian Conditional Random Fields (G-CRF) with Deep Learning: (a)\nour structured prediction task has a unique global optimum that is obtained\nexactly from the solution of a linear system (b) the gradients of our model\nparameters are analytically computed using closed form expressions, in contrast\nto the memory-demanding contemporary deep structured prediction approaches that\nrely on back-propagation-through-time, (c) our pairwise terms do not have to be\nsimple hand-crafted expressions, as in the line of works building on the\nDenseCRF, but can rather be `discovered' from data through deep architectures,\nand (d) out system can trained in an end-to-end manner. Building on standard\ntools from numerical analysis we develop very efficient algorithms for\ninference and learning, as well as a customized technique adapted to the\nsemantic segmentation task. This efficiency allows us to explore more\nsophisticated architectures for structured prediction in deep learning: we\nintroduce multi-resolution architectures to couple information across scales in\na joint optimization framework, yielding systematic improvements. We\ndemonstrate the utility of our approach on the challenging VOC PASCAL 2012\nimage segmentation benchmark, showing substantial improvements over strong\nbaselines. We make all of our code and experiments available at\n{https://github.com/siddharthachandra/gcrf}","url_abs":"http://arxiv.org/abs/1603.08358v4","url_pdf":"http://arxiv.org/pdf/1603.08358v4.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":"fast-exact-and-multi-scale-inference-for","repo_url":"https://github.com/siddharthachandra/gcrf","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"image-segmentation","task_name":"Image Segmentation"},{"task_slug":"prediction","task_name":"Prediction"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"structured-prediction","task_name":"Structured Prediction"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semantic-segmentation-on-pascal-voc-2012","task":"Semantic Segmentation","dataset":"PASCAL VOC 2012 test","model":"CentraleSupelec Deep G-CRF","rank_in_archive_order":29,"of":51,"metrics":{"Mean IoU":"80.2%"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1603.08358","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}