{"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/learning-pose-specific-representations-by","title":"Learning Pose Specific Representations by Predicting Different Views","arxiv_id":"1804.03390","date":"2018-04-10","proceeding":"CVPR 2018 6","authors":["Georg Poier","David Schinagl","Horst Bischof"],"abstract":"The labeled data required to learn pose estimation for articulated objects is\ndifficult to provide in the desired quantity, realism, density, and accuracy.\nTo address this issue, we develop a method to learn representations, which are\nvery specific for articulated poses, without the need for labeled training\ndata. We exploit the observation that the object pose of a known object is\npredictive for the appearance in any known view. That is, given only the pose\nand shape parameters of a hand, the hand's appearance from any viewpoint can be\napproximated. To exploit this observation, we train a model that -- given input\nfrom one view -- estimates a latent representation, which is trained to be\npredictive for the appearance of the object when captured from another\nviewpoint. Thus, the only necessary supervision is the second view. The\ntraining process of this model reveals an implicit pose representation in the\nlatent space. Importantly, at test time the pose representation can be inferred\nusing only a single view. In qualitative and quantitative experiments we show\nthat the learned representations capture detailed pose information. Moreover,\nwhen training the proposed method jointly with labeled and unlabeled data, it\nconsistently surpasses the performance of its fully supervised counterpart,\nwhile reducing the amount of needed labeled samples by at least one order of\nmagnitude.","url_abs":"http://arxiv.org/abs/1804.03390v2","url_pdf":"http://arxiv.org/pdf/1804.03390v2.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":"learning-pose-specific-representations-by","repo_url":"https://github.com/poier/PreView","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"learning-pose-specific-representations-by","repo_url":"https://github.com/RenFeiTemp/murauer/tree/master/source/my_net","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"hand-pose-estimation","task_name":"Hand Pose Estimation"},{"task_slug":"object","task_name":"Object"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1804.03390","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}