{"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/bihmp-gan-bidirectional-3d-human-motion","title":"BiHMP-GAN: Bidirectional 3D Human Motion Prediction GAN","arxiv_id":"1812.02591","date":"2018-12-06","proceeding":null,"authors":["Jogendra Nath Kundu","Maharshi Gor","R. Venkatesh Babu"],"abstract":"Human motion prediction model has applications in various fields of computer\nvision. Without taking into account the inherent stochasticity in the\nprediction of future pose dynamics, such methods often converges to a\ndeterministic undesired mean of multiple probable outcomes. Devoid of this, we\npropose a novel probabilistic generative approach called Bidirectional Human\nmotion prediction GAN, or BiHMP-GAN. To be able to generate multiple probable\nhuman-pose sequences, conditioned on a given starting sequence, we introduce a\nrandom extrinsic factor r, drawn from a predefined prior distribution.\nFurthermore, to enforce a direct content loss on the predicted motion sequence\nand also to avoid mode-collapse, a novel bidirectional framework is\nincorporated by modifying the usual discriminator architecture. The\ndiscriminator is trained also to regress this extrinsic factor r, which is used\nalongside with the intrinsic factor (encoded starting pose sequence) to\ngenerate a particular pose sequence. To further regularize the training, we\nintroduce a novel recursive prediction strategy. In spite of being in a\nprobabilistic framework, the enhanced discriminator architecture allows\npredictions of an intermediate part of pose sequence to be used as a\nconditioning for prediction of the latter part of the sequence. The\nbidirectional setup also provides a new direction to evaluate the prediction\nquality against a given test sequence. For a fair assessment of BiHMP-GAN, we\nreport performance of the generated motion sequence using (i) a critic model\ntrained to discriminate between real and fake motion sequence, and (ii) an\naction classifier trained on real human motion dynamics. Outcomes of both\nqualitative and quantitative evaluations, on the probabilistic generations of\nthe model, demonstrate the superiority of BiHMP-GAN over previously available\nmethods.","url_abs":"http://arxiv.org/abs/1812.02591v1","url_pdf":"http://arxiv.org/pdf/1812.02591v1.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":"bihmp-gan-bidirectional-3d-human-motion","repo_url":"https://github.com/ThomasDupiereux/ADLproject","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"bihmp-gan-bidirectional-3d-human-motion","repo_url":"https://github.com/ThomasDupiereux/EnGAN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"bihmp-gan-bidirectional-3d-human-motion","repo_url":"https://github.com/maharshi95/Pose2vec","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"human-motion-prediction","task_name":"Human motion prediction"},{"task_slug":"prediction","task_name":"Prediction"},{"task_slug":"motion-prediction","task_name":"motion prediction"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1812.02591","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}