{"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/binarized-convolutional-landmark-localizers","title":"Binarized Convolutional Landmark Localizers for Human Pose Estimation and Face Alignment with Limited Resources","arxiv_id":"1703.00862","date":"2017-03-02","proceeding":"ICCV 2017 10","authors":["Adrian Bulat","Georgios Tzimiropoulos"],"abstract":"Our goal is to design architectures that retain the groundbreaking\nperformance of CNNs for landmark localization and at the same time are\nlightweight, compact and suitable for applications with limited computational\nresources. To this end, we make the following contributions: (a) we are the\nfirst to study the effect of neural network binarization on localization tasks,\nnamely human pose estimation and face alignment. We exhaustively evaluate\nvarious design choices, identify performance bottlenecks, and more importantly\npropose multiple orthogonal ways to boost performance. (b) Based on our\nanalysis, we propose a novel hierarchical, parallel and multi-scale residual\narchitecture that yields large performance improvement over the standard\nbottleneck block while having the same number of parameters, thus bridging the\ngap between the original network and its binarized counterpart. (c) We perform\na large number of ablation studies that shed light on the properties and the\nperformance of the proposed block. (d) We present results for experiments on\nthe most challenging datasets for human pose estimation and face alignment,\nreporting in many cases state-of-the-art performance. Code can be downloaded\nfrom https://www.adrianbulat.com/binary-cnn-landmarks","url_abs":"http://arxiv.org/abs/1703.00862v2","url_pdf":"http://arxiv.org/pdf/1703.00862v2.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":"binarized-convolutional-landmark-localizers","repo_url":"https://github.com/1adrianb/binary-networks-pytorch","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"binarized-convolutional-landmark-localizers","repo_url":"https://github.com/1adrianb/binary-face-alignment","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"torch","reach":null},{"paper_slug":"binarized-convolutional-landmark-localizers","repo_url":"https://github.com/1adrianb/binary-human-pose-estimation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"torch","reach":null}],"tasks":[{"task_slug":"binarization","task_name":"Binarization"},{"task_slug":"face-alignment","task_name":"Face Alignment"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/face-alignment-on-aflw-full-1","task":"Face Alignment","dataset":"AFLW-Full","model":"Binary Face Alignment","rank_in_archive_order":1,"of":2,"metrics":{"Mean NME ":"2.85"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1703.00862","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}