Research map / 3D vision / Human pose estimation
RIDI: Robust IMU Double Integration
Hang Yan, Qi Shan, Yasutaka Furukawa
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 |
|---|---|---|---|
| add placement classification before velocity regression (cascaded SVM+SVR) (ablation) | RIDI velocity regressor · Velocity regression test set | MSE X axis: 0.32 -> 0.016 | Improved |
| add placement classification before velocity regression (cascaded SVM+SVR) (ablation) | RIDI velocity regressor · Velocity regression test set | MSE Z axis: 0.54 -> 0.015 | Improved |
| correct accelerations with regressed velocities instead of naive double integration, lambda 10,000 -> 0.1 (ablation) | RIDI · Average MPE test sequences | MPE ratio: 33.98% -> 1.45% | Improved |
| add raw IMU acceleration term (lambda 0.0001 -> 0.1) versus integrating regressed velocities only (ablation) | RIDI · Average MPE test sequences | MPE ratio: 11.62% -> 1.45% | Improved |
| vary lambda between 0.001 and 1.0 (ablation) | RIDI · Average MPE test sequences | MPE ratio: 1.49% / 1.45% / 1.47% | No clear effect |
| use RIDI instead of RAW double integration | naive IMU double integration · T5 hand | mean positional error (m): 180.03 -> 0.91 | Improved |
| use regressed direction and magnitude (RIDI) instead of system-rotation direction (RIDI-MAG) | RIDI-MAG · T7 body, backward motions | mean positional error (m): 10.87 -> 1.77 | Improved |
| use regressed speed magnitude (RIDI) instead of constant average speed (RIDI-ORI) | RIDI-ORI · T5 hand, changing speed | mean positional error (m): 8.37 -> 0.91 | Improved |
| use RIDI instead of step counting | STEP · T7 body | mean positional error (m): 16.28 -> 1.77 | Improved |
| use RIDI instead of step counting on leg placement T2 | STEP · T2 leg | mean positional error (m): 0.76 -> 1.23 | Hurt |
| train on four subjects instead of one (ablation) | RIDI (Single) · Unseen subject TS1 (S5, leg) | MPE (m): 5.98 -> 3.51 | Improved |
| train on four subjects instead of one (ablation) | RIDI (Single) · Unseen subject TS6 (S6, bag) | MPE (m): 11.79 -> 8.36 | Improved |
Closest papers by meaning
- iMoT: Inertial Motion Transformer for Inertial Navigation (AAAI 2025 · similarity 0.87)
- EqNIO: Subequivariant Neural Inertial Odometry (ICLR 2025 · similarity 0.87)
- MODA: Motion-Drift Augmentation for Inertial Human Motion Analysis (CVPR 2025 · similarity 0.86)
- Tartan IMU: A Light Foundation Model for Inertial Positioning in Robotics (CVPR 2025 · similarity 0.85)
- Inertia-1: An Open Exploration of Wearable Motion Foundation Models (NeurIPS 2026 · similarity 0.84)
- ImDy: Human Inverse Dynamics from Imitated Observations (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.
