profile

Jiaxi Jiang 蒋佳熹

PhD Student in Computer Science
ETH Zurich, Switzerland
CV | Google Scholar | Github | LinkedIn | X (Twitter) | Email

About Me

I am a PhD student at the Sensing, Interaction & Perception Lab, ETH Zurich with Prof. Christian Holz. I am now also doing my internship at Meta Reality Labs Research with Dr. Federica Bogo. My PhD research focuses on egocentric body motion modeling, tracking, and generation from wearable sensors. My goal is to build learning-based, physics-aware models for modeling, reconstructing, synthesizing, and understanding embodied agents—both virtual (e.g., digital characters, simulated humanoids) and real (e.g., humans, robots)—and their interactions with the environment from sparse and noisy observations. Specifically, I am interested in:

During my research journey, I was fortunate to collaborate with Larissa Laich and Dr. Patrick Snape from Meta Reality Labs, Zurich. Before starting my PhD, I got my Master’s degree in Computational Engineering Science (Specialized in Computer Vision and Robotics) from RWTH Aachen University, Germany. I wrote my Master’s thesis at the Computer Vision Lab, ETH Zurich, with Prof. Kai Zhang and Prof. Radu Timofte, supervised by Prof. Luc Van Gool and Prof. Bastian Leibe. Before that, I did a visiting research with Dr. Shuang Li and Prof. Jianwei Zhang at the University of Hamburg. Besides, I also worked with Dr. Jörg Stückler at Max Planck Institute for Intelligent Systems, and with Dr. Dengxin Dai at Max Planck Institute for Informatics.

News

Publications (* supervised projects)

  • Group Inertial Poser: Multi-Person Pose and Global Translation from Sparse Inertial Sensors and Ultra-Wideband Ranging
    Ying Xue, Jiaxi Jiang*, Rayan Armani, Dominik Hollidt, Yi-Chi Liao, Christian Holz
    International Conference on Computer Vision (ICCV), 2025.
    Project Page Paper Poster Code We tackle multi-person motion capture from only sparse body-worn IMUs — long thought infeasible because IMUs cannot resolve inter-person geometry. By adding low-cost ultra-wideband ranging between users, we recover full-body poses and global trajectories for everyone in the group, anywhere — no cameras, no external infrastructure. 

    @inproceedings{jiang2024egoposer,
title={EgoPoser: Robust real-time egocentric pose estimation from sparse and intermittent observations everywhere},
author={Jiang, Jiaxi and Streli, Paul and Meier, Manuel and Holz, Christian},
booktitle={European Conference on Computer Vision},
year={2024},
organization={Springer}
}

  • Garment Inertial Poser: Human Motion Capture from Loose and Sparse Inertial Sensors with Garment-aware Diffusion Models
    Andela Ilic, Jiaxi Jiang*, Paul Streli, Xintong Liu, Christian Holz
    International Joint Conferences on Artificial Intelligence (IJCAI), 2025.
    Project Page Paper We push sparse-IMU motion capture out of the lab and into real clothing — no more skin-tight straps. Our garment-aware diffusion model explicitly captures how IMUs slide, swing, and rotate with loose fabric, recovering accurate full-body motion in everyday wear for the first time. 

    @article{ilic2025human,
  title={Human Motion Capture from Loose and Sparse Inertial Sensors with Garment-aware Diffusion Models},
  author={Ilic, Andela and Jiang, Jiaxi and Streli, Paul and Liu, Xintong and Holz, Christian},
  journal={arXiv preprint arXiv:2506.15290},
  year={2025}
}

  • MANIKIN: Biomechanically Accurate Neural Inverse Kinematics for Human Motion Estimation
    Jiaxi Jiang, Paul Streli, Xuejing Luo, Christoph Gebhardt, Christian Holz
    European Conference on Computer Vision (ECCV), 2024.
    Project Page Paper Poster We embed human biomechanics directly into a neural inverse kinematics solver, refining the SMPL skeleton and combining it with analytic forward kinematics to enforce realistic joint limits and bone lengths. The result: anatomically plausible full-body pose from sparse sensors, with none of the bent-elbow contortions purely data-driven IK produces. 

    @inproceedings{jiang2024manikin,
  title={Manikin: Biomechanically accurate neural inverse kinematics for human motion estimation},
  author={Jiang, Jiaxi and Streli, Paul and Luo, Xuejing and Gebhardt, Christoph and Holz, Christian},
  booktitle={European Conference on Computer Vision},
  year={2024},
  organization={Springer}
}

  • EgoPoser: Robust Real-Time Egocentric Pose Estimation from Sparse and Intermittent Observations Everywhere
    Jiaxi Jiang, Paul Streli, Manuel Meier, Christian Holz
    European Conference on Computer Vision (ECCV), 2024.
    Project Page Paper Poster Code Real-time full-body pose from just a VR/MR headset and its two hand inputs — robust even when the hands leave the headset's field of view, and generalizing across diverse body shapes. No more fixed play areas: EgoPoser works anywhere the user goes. 

    @inproceedings{jiang2024egoposer,
title={EgoPoser: Robust real-time egocentric pose estimation from sparse and intermittent observations everywhere},
author={Jiang, Jiaxi and Streli, Paul and Meier, Manuel and Holz, Christian},
booktitle={European Conference on Computer Vision},
year={2024},
organization={Springer}
}

  • EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity
    Dominik Hollidt, Paul Streli, Jiaxi Jiang*, Yasaman Haghighi, Changlin Qian, Xintong Liu, Christian Holz
    Neural Information Processing Systems (NeurIPS), 2024
    Project Page Paper Video Dataset Code An egocentric multi-view simulator for body-worn cameras during real human motion, shipped with a paired real-world dataset. Together they unlock training and benchmarking of egocentric vision under the motion blur, occlusion, and viewpoint shifts that body-worn cameras actually experience — conditions no prior dataset captured. 

            @inproceedings{hollidtegosim,
          title={EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity},
          author={Hollidt, Dominik and Streli, Paul and Jiang, Jiaxi and Haghighi, Yasaman and Qian, Changlin and Liu, Xintong and Holz, Christian},
          booktitle={The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track}
        }

  • Ultra Inertial Poser: Scalable Motion Capture and Tracking from Sparse Inertial Sensors and Ultra-Wideband Ranging
    Rayan Armani, Changlin Qian, Jiaxi Jiang*, Christian Holz
    ACM SIGGRAPH 2024
    Project Page Paper Video Code We tackle the global drift that has long plagued sparse-IMU motion capture. By fusing body-worn IMUs with ultra-wideband ranging between sensors, we anchor absolute body position and limb geometry — camera-free, marker-free, and scalable to everyday environments. 

    @article{armani2024ultra,
  title={Ultra Inertial Poser: Scalable Motion Capture and Tracking from Sparse Inertial Sensors and Ultra-Wideband Ranging},
  author={Armani, Rayan and Qian, Changlin and Jiang, Jiaxi and Holz, Christian},
  journal={arXiv preprint arXiv:2404.19541},
  year={2024}
}

  • Structured Light Speckle: Joint Ego-Centric Depth Estimation and Low-Latency Contact Detection via Remote Vibrometry
    Paul Streli, Jiaxi Jiang, Juliete Rossie, Christian Holz
    ACM Symposium on User Interface Software and Technology (UIST), 2023.
    Project Page Paper Video Talk One egocentric camera delivers both dense scene depth and low-latency touch detection on any surface — by decoding tiny remote vibrations in projected laser speckle. No instrumented surface, no extra sensor needed. 

    @inproceedings{uist2023-structured_light_speckle,
  author = {Streli, Paul and Jiang, Jiaxi and Rossie, Juliete and Holz, Christian},
  title = {Structured Light Speckle: Joint Ego-Centric Depth Estimation and Low-Latency Contact Detection via Remote Vibrometry},
  year = {2023},
  isbn = {9798400701320},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  doi = {https://doi.org/10.1145/3586183.3606749},
  booktitle = {Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology},
  articleno = {26},
  numpages = {12}
                }

  • AvatarPoser: Articulated Full-Body Pose Tracking from Sparse Motion Sensing
    Jiaxi Jiang, Paul Streli, Huajian Qiu, Andreas Fender, Larissa Laich, Patrick Snape, Christian Holz
    European Conference on Computer Vision (ECCV), 2022.
    Project Page Paper Poster Video Code The first method to drive a full-body avatar — legs and all — from just the head- and hand-tracking signals a commercial VR/MR headset already provides. Our Transformer infers plausible lower-body motion from inputs that contain zero leg observations, unlocking realistic full-body avatars with no extra sensors to wear. 

    @inproceedings{jiang2022avatarposer,
  title={Avatarposer: Articulated full-body pose tracking from sparse motion sensing},
  author={Jiang, Jiaxi and Streli, Paul and Qiu, Huajian and Fender, Andreas and Laich, Larissa and Snape, Patrick and Holz, Christian},
  booktitle={European Conference on Computer Vision},
  pages={443--460},
  year={2022},
  organization={Springer}
}

  • TapType: Ten-Finger Text Entry on Everyday Surfaces via Bayesian Inference
    Paul Streli, Jiaxi Jiang, Manuel Meier, Hugo Romat, Andreas Fender, Christian Holz
    ACM CHI Conference on Human Factors in Computing Systems (CHI), 2022.
    Project Page Paper Video Demo Type with all ten fingers on any flat surface — a table, a desk, a wall — wearing nothing but two wristbands. Our Bayesian model fuses IMU-based tap detection with a language prior to decode keystrokes accurately, with no keyboard, no markers, no instrumented surface. 

    @inproceedings{chi2022-taptype,
author = {Streli, Paul and Jiang, Jiaxi and Fender, Andreas and Meier, Manuel and Romat, Hugo and Holz, Christian},
title = {TapType: Ten-finger text entry on everyday surfaces via Bayesian inference},
year = {2022},
isbn = {9781450391573},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
doi = {https://doi.org/10.1145/3491102.3501878},
booktitle = {Proceedings of the 2022 CHI Conference on Human Factors in Computing Systems},
numpages = {16}
}

  • Towards Flexible Blind JPEG Artifacts Removal
    Jiaxi Jiang, Kai Zhang, Radu Timofte
    International Conference on Computer Vision (ICCV), 2021.
    Paper Poster Code Master Thesis A controllable blind JPEG artifact removal network: users dial the restoration strength at inference time, with no retraining and no per-quality model. By explicitly predicting and decoupling the quality factor, we are also the first to deeply tackle double JPEG compression and real-world images where the quality is unknown or even varies across the image. 

    @inproceedings{jiang2021towards,
title={Towards Flexible Blind JPEG Artifacts Removal},
author={Jiang, Jiaxi and Zhang, Kai and Timofte, Radu},
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
pages={4997--5006},
year={2021}
}

  • A Mobile Robot Hand-Arm Teleoperation System by Vision and IMU
    Shuang Li, Jiaxi Jiang, Philipp Ruppel, Hongzhuo Liang, Xiaojian Ma, Norman Hendrich, Fuchun Sun, Jianwei Zhang
    International Conference on Intelligent Robots and Systems (IROS), 2020.
    Project Page Paper Video Code Teleoperate a dexterous mobile manipulator with just a webcam (for hand pose) and a single IMU (for arm pose) — no motion-capture suit, no expensive trackers. We show that off-the-shelf consumer sensing is enough to drive a full hand-arm robot through dexterous tasks in real time. 

    @inproceedings{li2020mobile,
title={A mobile robot hand-arm teleoperation system by vision and imu},
author={Li, Shuang and Jiang, Jiaxi and Ruppel, Philipp and Liang, Hongzhuo and Ma, Xiaojian and Hendrich, Norman and Sun, Fuchun and Zhang, Jianwei},
booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
pages={10900--10906},
year={2020},
organization={IEEE}
}

    • Academic Services

    Reviewers

    Computer Vision: CVPR, ICCV, ECCV, TIP
    Computer Graphics: SIGGRAPH, SIGGRAPH Asia, ToG
    Machine Learning: NeurIPS, ICLR, ICML, AAAI
    Human-Computer Interaction: CHI, UIST, IEEE VR

    Teaching

    Egocentric Computer Vision: Perception and Understanding from the First-Person View, ETH Zurich, Spring 2026
    Artificial Intelligence for Digital Characters, ETH Zurich, Spring 2025, Spring 2026
    Egocentric Perception for Human-World Understanding, ETH Zurich, Spring 2025
    Computer Science I, ETH Zurich, Autumn 2024
    Stochastics and Machine Learning, ETH Zurich, Spring 2024
    Mixed Reality, ETH Zurich, Autumn 2023
    Big Data For Engineers, ETH Zurich, Spring 2023

    Supervised Students

    Qi Ma, Master Thesis, MSc in CS, ETH Zurich, 11.2025 - Present
    Liuxin Qing, Semester Thesis, Master Thesis, MSc in EE, ETH Zurich, 10.2024 - 02.2026
    Nicola Witzig, Bachelor Thesis, BSc in CS, ETH Zurich, 03.2025 - 09.2025
    Shuyue Wang, Master Thesis, MSc in CS, ETH Zurich, 04.2024 - 03.2025
    Chang-Jin Lee, Bachelor Thesis, BSc in CS, ETH Zurich, 04.2024 - 10.2024
    Andela Ilic, Master Thesis, MSc in EE, ETH Zurich, 04.2024 - 10.2024, now PhD Student at ETH Zurich
    Ying Xue, Master Thesis, Research Intern, MSc in CS, ETH Zurich, 01.2024 - 05.2025, now PhD Student at ETH Zurich
    Changlin Qian, Research Intern, MSc in Robotics, ETH Zurich, 09.2023 - 07.2024, now Software Engineer at Meta
    Rayan Armani, Research Assistant, MSc in Robotics, ETH Zurich, 2023.09 - 03.2024, now Software Engineer at Apple
    Yasaman Haghighi, Research Intern, MSc in EE, EPFL, 2023.07 - 2024.01, now PhD Student at EPFL
    Xuejing Luo, Master Thesis, MSc in Robotics, ETH Zurich, 05.2023 - 02.2024, now PhD Student at Aalto University
    Xintong Liu, Master Thesis, MSc in Robotics, ETH Zurich, 03.2023 - 03.2024
    Filip Stanisic, Semester Thesis, MSc in EE, ETH Zurich, 03.2023 - 07.2023
    Huajian Qiu, Master Thesis, MSc in CSE, EPFL, 2021.10-2022.01, now PhD Student at Unviersity of Stuttgart

               
    Back to Top