{"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/decafa-deep-convolutional-cascade-for-face","title":"DeCaFA: Deep Convolutional Cascade for Face Alignment In The Wild","arxiv_id":"1904.02549","date":"2019-04-04","proceeding":"ICCV 2019 10","authors":["Arnaud Dapogny","Kévin Bailly","Matthieu Cord"],"abstract":"Face Alignment is an active computer vision domain, that consists in\nlocalizing a number of facial landmarks that vary across datasets.\nState-of-the-art face alignment methods either consist in end-to-end\nregression, or in refining the shape in a cascaded manner, starting from an\ninitial guess. In this paper, we introduce DeCaFA, an end-to-end deep\nconvolutional cascade architecture for face alignment. DeCaFA uses\nfully-convolutional stages to keep full spatial resolution throughout the\ncascade. Between each cascade stage, DeCaFA uses multiple chained transfer\nlayers with spatial softmax to produce landmark-wise attention maps for each of\nseveral landmark alignment tasks. Weighted intermediate supervision, as well as\nefficient feature fusion between the stages allow to learn to progressively\nrefine the attention maps in an end-to-end manner. We show experimentally that\nDeCaFA significantly outperforms existing approaches on 300W, CelebA and WFLW\ndatabases. In addition, we show that DeCaFA can learn fine alignment with\nreasonable accuracy from very few images using coarsely annotated data.","url_abs":"http://arxiv.org/abs/1904.02549v1","url_pdf":"http://arxiv.org/pdf/1904.02549v1.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-alignment","task_name":"Face Alignment"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/face-alignment-on-300w","task":"Face Alignment","dataset":"300W","model":"DeCaFA","rank_in_archive_order":26,"of":48,"metrics":{"NME_inter-ocular (%, Challenge)":"5.26","NME_inter-ocular (%, Common)":"2.93","NME_inter-ocular (%, Full)":"3.39"},"uses_additional_data":false},{"leaderboard":"/sota/face-alignment-on-wflw","task":"Face Alignment","dataset":"WFLW","model":"DeCaFA","rank_in_archive_order":22,"of":36,"metrics":{"AUC@10 (inter-ocular)":"56.3","FR@10 (inter-ocular)":"4.84","NME (inter-ocular)":"4.62"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1904.02549","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}