{"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/deep-inertial-poser-learning-to-reconstruct","title":"Deep Inertial Poser: Learning to Reconstruct Human Pose from Sparse Inertial Measurements in Real Time","arxiv_id":"1810.04703","date":"2018-10-10","proceeding":null,"authors":["Yinghao Huang","Manuel Kaufmann","Emre Aksan","Michael J. Black","Otmar Hilliges","Gerard Pons-Moll"],"abstract":"We demonstrate a novel deep neural network capable of reconstructing human\nfull body pose in real-time from 6 Inertial Measurement Units (IMUs) worn on\nthe user's body. In doing so, we address several difficult challenges. First,\nthe problem is severely under-constrained as multiple pose parameters produce\nthe same IMU orientations. Second, capturing IMU data in conjunction with\nground-truth poses is expensive and difficult to do in many target application\nscenarios (e.g., outdoors). Third, modeling temporal dependencies through\nnon-linear optimization has proven effective in prior work but makes real-time\nprediction infeasible. To address this important limitation, we learn the\ntemporal pose priors using deep learning. To learn from sufficient data, we\nsynthesize IMU data from motion capture datasets. A bi-directional RNN\narchitecture leverages past and future information that is available at\ntraining time. At test time, we deploy the network in a sliding window fashion,\nretaining real time capabilities. To evaluate our method, we recorded DIP-IMU,\na dataset consisting of $10$ subjects wearing 17 IMUs for validation in $64$\nsequences with $330\\,000$ time instants; this constitutes the largest IMU\ndataset publicly available. We quantitatively evaluate our approach on multiple\ndatasets and show results from a real-time implementation. DIP-IMU and the code\nare available for research purposes.","url_abs":"http://arxiv.org/abs/1810.04703v1","url_pdf":"http://arxiv.org/pdf/1810.04703v1.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":"deep-inertial-poser-learning-to-reconstruct","repo_url":"https://github.com/eth-ait/dip18","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[{"slug":"dip-imu","name":"DIP-IMU","full_name":""}],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}