{"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/self-supervised-learning-of-3d-human-pose","title":"Self-Supervised Learning of 3D Human Pose using Multi-view Geometry","arxiv_id":"1903.02330","date":"2019-03-06","proceeding":"CVPR 2019 6","authors":["Muhammed Kocabas","Salih Karagoz","Emre Akbas"],"abstract":"Training accurate 3D human pose estimators requires large amount of 3D\nground-truth data which is costly to collect. Various weakly or self supervised\npose estimation methods have been proposed due to lack of 3D data.\nNevertheless, these methods, in addition to 2D ground-truth poses, require\neither additional supervision in various forms (e.g. unpaired 3D ground truth\ndata, a small subset of labels) or the camera parameters in multiview settings.\nTo address these problems, we present EpipolarPose, a self-supervised learning\nmethod for 3D human pose estimation, which does not need any 3D ground-truth\ndata or camera extrinsics. During training, EpipolarPose estimates 2D poses\nfrom multi-view images, and then, utilizes epipolar geometry to obtain a 3D\npose and camera geometry which are subsequently used to train a 3D pose\nestimator. We demonstrate the effectiveness of our approach on standard\nbenchmark datasets i.e. Human3.6M and MPI-INF-3DHP where we set the new\nstate-of-the-art among weakly/self-supervised methods. Furthermore, we propose\na new performance measure Pose Structure Score (PSS) which is a scale\ninvariant, structure aware measure to evaluate the structural plausibility of a\npose with respect to its ground truth. Code and pretrained models are available\nat https://github.com/mkocabas/EpipolarPose","url_abs":"http://arxiv.org/abs/1903.02330v2","url_pdf":"http://arxiv.org/pdf/1903.02330v2.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":"self-supervised-learning-of-3d-human-pose","repo_url":"https://github.com/mkocabas/EpipolarPose","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"3d-human-pose-estimation","task_name":"3D Human Pose Estimation"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"},{"task_slug":"self-supervised-learning","task_name":"Self-Supervised Learning"},{"task_slug":"weakly-supervised-3d-human-pose-estimation","task_name":"Weakly-supervised 3D Human Pose Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/3d-human-pose-estimation-on-mpi-inf-3dhp","task":"3D Human Pose Estimation","dataset":"MPI-INF-3DHP","model":"EpipolarPose (fully-supervised)","rank_in_archive_order":74,"of":108,"metrics":{"MPJPE":"108.99","PCK":"77.5"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-3d-human-pose-estimation-on","task":"Weakly-supervised 3D Human Pose Estimation","dataset":"Human3.6M","model":"EpipolarPose (SS + RU)","rank_in_archive_order":11,"of":33,"metrics":{"Average MPJPE (mm)":"60.56"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-3d-human-pose-estimation-on","task":"Weakly-supervised 3D Human Pose Estimation","dataset":"Human3.6M","model":"EpipolarPose (S1)","rank_in_archive_order":17,"of":33,"metrics":{"Average MPJPE (mm)":"65.35"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-3d-human-pose-estimation-on","task":"Weakly-supervised 3D Human Pose Estimation","dataset":"Human3.6M","model":"EpipolarPose (self-supervised)","rank_in_archive_order":20,"of":33,"metrics":{"Average MPJPE (mm)":"76.6"},"uses_additional_data":false},{"leaderboard":"/sota/weakly-supervised-3d-human-pose-estimation-on","task":"Weakly-supervised 3D Human Pose Estimation","dataset":"Human3.6M","model":"Kocabas et al.","rank_in_archive_order":30,"of":33,"metrics":{"3D Annotations":"S1","Number of Frames Per View":"1","Number of Views":"2"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1903.02330","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}