Research map / 3D vision / Human pose estimation
AddBiomechanics Dataset: Capturing the Physics of Human Motion at Scale
Keenon Werling, Janelle M Kaneda, Tian Tan, Rishi Agarwal, Six Skov, Tom Van Wouwe, Scott Uhlrich, Scott Delp, Karen Liu, Nicholas A Bianco, Carmichael Ong, Antoine Falisse and 6 more
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 |
|---|---|---|---|
| initialize center-of-mass trajectory using frames with unobserved forces instead of fitting each observed segment independently (ablation) | AddBiomechanics pipeline · Hidden-force ablation (50 s trial) | linear residual (N): 3630N ± 3220N -> 18.0 N ± 2.5N | Improved |
| initialize center-of-mass trajectory using frames with unobserved forces instead of fitting each observed segment independently (ablation) | AddBiomechanics pipeline · Hidden-force ablation (50 s trial) | marker RMS (cm): 7.7cm ± 6.2cm -> 2.4cm ± 0.5cm | Improved |
| hide force data on every third right step versus full force access (oracle) | AddBiomechanics pipeline · Hidden-force ablation (50 s trial) | linear residual (N): 12.9 N -> 18.0 N | Hurt |
| use a learned MLP instead of an analytical F=ma baseline for force prediction | force prediction from motion · AddBiomechanics test set | GRF error (N/kg): 2.79 -> 1.29 | Improved |
| use larger GroundLinkNet architecture instead of two-layer MLP | force prediction from motion · AddBiomechanics test set | joint torque error (Nm/kg): 1.33 -> 1.34 | No clear effect |
| use larger GroundLinkNet architecture instead of two-layer MLP | force prediction from motion · AddBiomechanics test set | GRF error (N/kg): 1.29 -> 1.17 | Improved |
| use larger GroundLinkNet architecture instead of two-layer MLP | force prediction from motion · AddBiomechanics test set | GRM error (Nm/kg): 0.03 -> 0.07 | Hurt |
Closest papers by meaning
- Muscles in Time: Learning to Understand Human Motion In-Depth by Simulating Muscle Activations (NeurIPS 2024 · similarity 0.88)
- PhysPT: Physics-aware Pretrained Transformer for Estimating Human Dynamics from Monocular Videos (CVPR 2024 · similarity 0.88)
- Homogeneous Dynamics Space for Heterogeneous Humans (CVPR 2025 · similarity 0.88)
- Dynamic Inertial Poser (DynaIP): Part-Based Motion Dynamics Learning for Enhanced Human Pose Estimation with Sparse Inertial Sensors (CVPR 2024 · similarity 0.87)
- ImDy: Human Inverse Dynamics from Imitated Observations (ICLR 2025 · similarity 0.87)
- Optimal-state Dynamics Estimation for Physics-based Human Motion Capture from Videos (NeurIPS 2024 · similarity 0.86)
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.
