{"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/higher-order-conditional-random-fields-in","title":"Higher Order Conditional Random Fields in Deep Neural Networks","arxiv_id":"1511.08119","date":"2015-11-25","proceeding":null,"authors":["Anurag Arnab","Sadeep Jayasumana","Shuai Zheng","Philip Torr"],"abstract":"We address the problem of semantic segmentation using deep learning. Most\nsegmentation systems include a Conditional Random Field (CRF) to produce a\nstructured output that is consistent with the image's visual features. Recent\ndeep learning approaches have incorporated CRFs into Convolutional Neural\nNetworks (CNNs), with some even training the CRF end-to-end with the rest of\nthe network. However, these approaches have not employed higher order\npotentials, which have previously been shown to significantly improve\nsegmentation performance. In this paper, we demonstrate that two types of\nhigher order potential, based on object detections and superpixels, can be\nincluded in a CRF embedded within a deep network. We design these higher order\npotentials to allow inference with the differentiable mean field algorithm. As\na result, all the parameters of our richer CRF model can be learned end-to-end\nwith our pixelwise CNN classifier. We achieve state-of-the-art segmentation\nperformance on the PASCAL VOC benchmark with these trainable higher order\npotentials.","url_abs":"http://arxiv.org/abs/1511.08119v4","url_pdf":"http://arxiv.org/pdf/1511.08119v4.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":"higher-order-conditional-random-fields-in","repo_url":"https://github.com/torrvision/caffe-tvg","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"superpixels","task_name":"Superpixels"}],"methods":[{"method_slug":"crf","method_name":"CRF"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/semantic-segmentation-on-pascal-context","task":"Semantic Segmentation","dataset":"PASCAL Context","model":"HO CRF","rank_in_archive_order":59,"of":66,"metrics":{"mIoU":"41.3"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1511.08119","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}