{"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/facial-landmarks-detection-by-self-iterative","title":"Facial Landmarks Detection by Self-Iterative Regression based Landmarks-Attention Network","arxiv_id":"1803.06598","date":"2018-03-18","proceeding":null,"authors":["Tao Hu","Honggang Qi","Jizheng Xu","Qingming Huang"],"abstract":"Cascaded Regression (CR) based methods have been proposed to solve facial\nlandmarks detection problem, which learn a series of descent directions by\nmultiple cascaded regressors separately trained in coarse and fine stages. They\noutperform the traditional gradient descent based methods in both accuracy and\nrunning speed. However, cascaded regression is not robust enough because each\nregressor's training data comes from the output of previous regressor.\nMoreover, training multiple regressors requires lots of computing resources,\nespecially for deep learning based methods. In this paper, we develop a\nSelf-Iterative Regression (SIR) framework to improve the model efficiency. Only\none self-iterative regressor is trained to learn the descent directions for\nsamples from coarse stages to fine stages, and parameters are iteratively\nupdated by the same regressor. Specifically, we proposed Landmarks-Attention\nNetwork (LAN) as our regressor, which concurrently learns features around each\nlandmark and obtains the holistic location increment. By doing so, not only the\nrest of regressors are removed to simplify the training process, but the number\nof model parameters is significantly decreased. The experiments demonstrate\nthat with only 3.72M model parameters, our proposed method achieves the\nstate-of-the-art performance.","url_abs":"http://arxiv.org/abs/1803.06598v1","url_pdf":"http://arxiv.org/pdf/1803.06598v1.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"},{"task_slug":"regression-1","task_name":"regression"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/face-alignment-on-300w","task":"Face Alignment","dataset":"300W","model":"SIR-LAN","rank_in_archive_order":48,"of":48,"metrics":{"NME_inter-pupil (%, Challenge)":"8.14","NME_inter-pupil (%, Common)":"4.29","NME_inter-pupil (%, Full)":"5.04"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}