{"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/peak-piloted-deep-network-for-facial","title":"Peak-Piloted Deep Network for Facial Expression Recognition","arxiv_id":"1607.06997","date":"2016-07-24","proceeding":null,"authors":["Xiangyun Zhao","Xiaodan Liang","Luoqi Liu","Teng Li","Yugang Han","Nuno Vasconcelos","Shuicheng Yan"],"abstract":"Objective functions for training of deep networks for face-related\nrecognition tasks, such as facial expression recognition (FER), usually\nconsider each sample independently. In this work, we present a novel\npeak-piloted deep network (PPDN) that uses a sample with peak expression (easy\nsample) to supervise the intermediate feature responses for a sample of\nnon-peak expression (hard sample) of the same type and from the same subject.\nThe expression evolving process from non-peak expression to peak expression can\nthus be implicitly embedded in the network to achieve the invariance to\nexpression intensities. A special purpose back-propagation procedure, peak\ngradient suppression (PGS), is proposed for network training. It drives the\nintermediate-layer feature responses of non-peak expression samples towards\nthose of the corresponding peak expression samples, while avoiding the inverse.\nThis avoids degrading the recognition capability for samples of peak expression\ndue to interference from their non-peak expression counterparts. Extensive\ncomparisons on two popular FER datasets, Oulu-CASIA and CK+, demonstrate the\nsuperiority of the PPDN over state-ofthe-art FER methods, as well as the\nadvantages of both the network structure and the optimization strategy.\nMoreover, it is shown that PPDN is a general architecture, extensible to other\ntasks by proper definition of peak and non-peak samples. This is validated by\nexperiments that show state-of-the-art performance on pose-invariant face\nrecognition, using the Multi-PIE dataset.","url_abs":"http://arxiv.org/abs/1607.06997v2","url_pdf":"http://arxiv.org/pdf/1607.06997v2.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":[],"tasks":[{"task_slug":"face-recognition","task_name":"Face Recognition"},{"task_slug":"facial-expression-recognition-1","task_name":"Facial Expression Recognition"},{"task_slug":"facial-expression-recognition","task_name":"Facial Expression Recognition (FER)"},{"task_slug":"robust-face-recognition","task_name":"Robust Face Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/facial-expression-recognition-on-oulu-casia","task":"Facial Expression Recognition (FER)","dataset":"Oulu-CASIA","model":"PPDN","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy (10-fold)":"84.59"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1607.06997","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}