Research map / Video / Action recognition
MoBind: Motion Binding for Fine-Grained IMU-Video Pose Alignment
Duc Duy Nguyen, Tat-Jun Chin, Minh Hoai
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 Masked Token Prediction auxiliary task (ablation) | MoBind · TotalCapture action recognition (finetune) | accuracy: 0.55 -> 0.72 | Improved |
| Add Masked Token Prediction auxiliary task (ablation) | MoBind · TotalCapture action recognition (1-NN) | accuracy: 0.53 -> 0.71 | Improved |
| Add Masked Token Prediction auxiliary task (ablation) | MoBind · mRi action recognition (1-NN) | accuracy: 0.76 -> 0.86 | Improved |
| Add Masked Token Prediction auxiliary task (ablation) | MoBind · mRi action recognition (finetune) | accuracy: 0.97 -> 0.98 | Improved |
| Add hierarchical contrastive levels (token, local, global) (ablation) | MoBind · mRi | R@1, sync, localization | Improved |
| Align IMU with 2D pose rather than raw video, with local plus global alignment vs global-only baselines | MoBind · TotalCapture retrieval | R@1 IMU->Video: 0.51 -> 0.87 | Improved |
| Hierarchical body-part alignment vs best baseline for synchronization | MoBind · EgoHumans synchronization | MAE (s): 1.01 -> 0.04 | Improved |
| Use longer clips for synchronization | MoBind · mRi synchronization | accuracy: 97% (2-min) -> 100% (3-min) | Improved |
| Offset-classification baseline instead of contrastive embedding | IMUSync · EgoHumans synchronization | accuracy: 0.82 vs 1.00 (MoBind) | No clear effect |
| Add random sensor dropout training | MoBind · TotalCapture retrieval | R@k | Improved |
| Use IMU-to-CLIP style global-only alignment | IMU2CLIP · TotalCapture retrieval | R@1 IMU->Video: 0.87 (MoBind) vs 0.06 | Hurt |
Closest papers by meaning
- MotionBind: Multi-Modal Human Motion Alignment for Retrieval, Recognition, and Generation (NeurIPS 2025 · similarity 0.87)
- Spatial-Related Sensors Matters: 3D Human Motion Reconstruction Assisted with Textual Semantics (AAAI 2024 · similarity 0.86)
- MODA: Motion-Drift Augmentation for Inertial Human Motion Analysis (CVPR 2025 · similarity 0.85)
- HiPoser: 3D Human Pose Estimation with Hierarchical Shared Learning at Parts-Level Using Inertial Measurement Units (AAAI 2025 · similarity 0.85)
- Dynamic Inertial Poser (DynaIP): Part-Based Motion Dynamics Learning for Enhanced Human Pose Estimation with Sparse Inertial Sensors (CVPR 2024 · similarity 0.84)
- I'M HOI: Inertia-aware Monocular Capture of 3D Human-Object Interactions (CVPR 2024 · 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.
