Research map / 3D vision / Human pose estimation
Towards Real-World Wearable Motion Reconstruction
Andrea Boscolo Camiletto, Rishabh Dabral, Eduardo Alvarado, Thabo Beeler, Marc Habermann, Christian Theobalt
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use flow-matching generative model instead of matched-size transformer regressor | WHIP · Unseen actor (all sensors) | MPJPE: 78.3 -> 56.7 | Improved |
| use flow-matching generative model instead of matched-size MLP regressor | WHIP · Unseen actor (all sensors) | MPJPE: 127.8 -> 56.7 | Improved |
| use generative model instead of transformer regressor | WHIP · Unseen action (all sensors) | MPJPE: 83.3 -> 60.0 | Improved |
| use generative model instead of transformer regressor | WHIP · Unseen actor (insoles only) | N-MPJPE: 147.6 -> 151.7 | Hurt |
| use generative model instead of transformer regressor | WHIP · Unseen actor (HMD only) | MPJPE: 135.9 -> 107.4 | Improved |
| train WHIP on arbitrary sensor subsets and compare with IMUPoser on IMU-only inputs | WHIP · Comparison with prior methods (retrained on our dataset), IMUs only, all phones and watches | N-MPJPE: 75.5 -> 61.7 | Improved |
| compare WHIP with HMDPoser when all sensors and HMD are available | WHIP · Comparison with prior methods (retrained on our dataset), HMD + all sensors | N-MPJPE: 74.6 -> 52.6 | Improved |
| add HMD to the set of sensors | WHIP · Sensor complementarity study | N-MPJPE | Improved |
| add smartwatches to the sensor set | WHIP · Sensor complementarity study | MPJPE | Improved |
| add action labels as conditioning | WHIP · Sensor complementarity study | MPJPE | No clear effect |
| add insoles to a full sensor set | WHIP · Sensor complementarity study | N-MPJPE | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- HMDPoser — HMD-Poser: On-Device Real-time Human Motion Tracking from Scalable Sparse Observations (CVPR 2024) · better on 1 of 1 shared result
- AvatarPoser — AvatarPoser: Articulated Full-Body Pose Tracking from Sparse Motion Sensing (ECCV 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Inertia-1: An Open Exploration of Wearable Motion Foundation Models (NeurIPS 2026 · similarity 0.89)
- Spatial-Related Sensors Matters: 3D Human Motion Reconstruction Assisted with Textual Semantics (AAAI 2024 · similarity 0.87)
- DiffusionPoser: Real-time Human Motion Reconstruction From Arbitrary Sparse Sensors Using Autoregressive Diffusion (CVPR 2024 · similarity 0.86)
- Mocap Everyone Everywhere: Lightweight Motion Capture With Smartwatches and a Head-Mounted Camera (CVPR 2024 · similarity 0.85)
- UMotion: Uncertainty-driven Human Motion Estimation from Inertial and Ultra-wideband Units (CVPR 2025 · similarity 0.84)
- RelCon: Relative Contrastive Learning for a Motion Foundation Model for Wearable Data (ICLR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
