{"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/monocap-monocular-human-motion-capture-using","title":"MonoCap: Monocular Human Motion Capture using a CNN Coupled with a Geometric Prior","arxiv_id":"1701.02354","date":"2017-01-09","proceeding":null,"authors":["Xiaowei Zhou","Menglong Zhu","Georgios Pavlakos","Spyridon Leonardos","Kostantinos G. Derpanis","Kostas Daniilidis"],"abstract":"Recovering 3D full-body human pose is a challenging problem with many\napplications. It has been successfully addressed by motion capture systems with\nbody worn markers and multiple cameras. In this paper, we address the more\nchallenging case of not only using a single camera but also not leveraging\nmarkers: going directly from 2D appearance to 3D geometry. Deep learning\napproaches have shown remarkable abilities to discriminatively learn 2D\nappearance features. The missing piece is how to integrate 2D, 3D and temporal\ninformation to recover 3D geometry and account for the uncertainties arising\nfrom the discriminative model. We introduce a novel approach that treats 2D\njoint locations as latent variables whose uncertainty distributions are given\nby a deep fully convolutional neural network. The unknown 3D poses are modeled\nby a sparse representation and the 3D parameter estimates are realized via an\nExpectation-Maximization algorithm, where it is shown that the 2D joint\nlocation uncertainties can be conveniently marginalized out during inference.\nExtensive evaluation on benchmark datasets shows that the proposed approach\nachieves greater accuracy over state-of-the-art baselines. Notably, the\nproposed approach does not require synchronized 2D-3D data for training and is\napplicable to \"in-the-wild\" images, which is demonstrated with the MPII\ndataset.","url_abs":"http://arxiv.org/abs/1701.02354v2","url_pdf":"http://arxiv.org/pdf/1701.02354v2.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":"monocap-monocular-human-motion-capture-using","repo_url":"https://github.com/daniilidis-group/monocap","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"torch","reach":null}],"tasks":[{"task_slug":"3d-geometry","task_name":"3D geometry"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1701.02354","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}