{"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/recurrent-3d-pose-sequence-machines","title":"Recurrent 3D Pose Sequence Machines","arxiv_id":"1707.09695","date":"2017-07-31","proceeding":"CVPR 2017 7","authors":["Mude Lin","Liang Lin","Xiaodan Liang","Keze Wang","Hui Cheng"],"abstract":"3D human articulated pose recovery from monocular image sequences is very\nchallenging due to the diverse appearances, viewpoints, occlusions, and also\nthe human 3D pose is inherently ambiguous from the monocular imagery. It is\nthus critical to exploit rich spatial and temporal long-range dependencies\namong body joints for accurate 3D pose sequence prediction. Existing approaches\nusually manually design some elaborate prior terms and human body kinematic\nconstraints for capturing structures, which are often insufficient to exploit\nall intrinsic structures and not scalable for all scenarios. In contrast, this\npaper presents a Recurrent 3D Pose Sequence Machine(RPSM) to automatically\nlearn the image-dependent structural constraint and sequence-dependent temporal\ncontext by using a multi-stage sequential refinement. At each stage, our RPSM\nis composed of three modules to predict the 3D pose sequences based on the\npreviously learned 2D pose representations and 3D poses: (i) a 2D pose module\nextracting the image-dependent pose representations, (ii) a 3D pose recurrent\nmodule regressing 3D poses and (iii) a feature adaption module serving as a\nbridge between module (i) and (ii) to enable the representation transformation\nfrom 2D to 3D domain. These three modules are then assembled into a sequential\nprediction framework to refine the predicted poses with multiple recurrent\nstages. Extensive evaluations on the Human3.6M dataset and HumanEva-I dataset\nshow that our RPSM outperforms all state-of-the-art approaches for 3D pose\nestimation.","url_abs":"http://arxiv.org/abs/1707.09695v1","url_pdf":"http://arxiv.org/pdf/1707.09695v1.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":"3d-human-pose-estimation","task_name":"3D Human Pose Estimation"},{"task_slug":"3d-pose-estimation","task_name":"3D Pose Estimation"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/3d-human-pose-estimation-on-humaneva-i","task":"3D Human Pose Estimation","dataset":"HumanEva-I","model":"Recurrent 3D Pose Sequence Machines","rank_in_archive_order":22,"of":31,"metrics":{"Mean Reconstruction Error (mm)":"30.8"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1707.09695","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}