Research map / 3D vision / Human pose estimation
MGDHand: Multi-Granularity Prior-to-Inertial Distillation Framework for Sequential 3D Hand Pose Estimation from Sparse IMUs
Xinyi Wang, Pengfei Ren, Haoyang Zhang, Hanling Zhan, Yingxi Li, Liang Xie, Yue Gao, Erwei Yin
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 MGDistill multi-granularity distillation to the IMU-only student | MGDHand student · VIHand (7 IMUs) | MPJPE (mm): 15.40 -> 9.13 | Improved |
| add MGDistill to the student with 3 IMUs | MGDHand student · VIHand (3 IMUs) | MPJPE (mm): 20.64 -> 15.48 | Improved |
| add MGDistill to the student with 2 IMUs | MGDHand student · VIHand (2 IMUs) | MPJPE (mm): 22.63 -> 17.54 | Improved |
| add dynamic pose distillation (DPD) (ablation) | MGDistill · VIHand (7 IMUs) | MPJPE (mm): 12.39 -> 9.13 | Improved |
| add static shape distillation (SSD) (ablation) | MGDistill · VIHand (7 IMUs) | MPVPE (mm): 12.58 -> 10.46 | Improved |
| add temporal motion distillation (TMD) (ablation) | MGDistill · VIHand (7 IMUs) | MPJPE (mm): 9.65 -> 9.13 | Improved |
| replace SCJD distillation with MGDistill (ablation) | student distillation method · VIHand (7 IMUs) | MPJPE (mm): 11.24 -> 9.13 | Improved |
| replace coupled global feature distillation (VIFNet-S) with decoupled MGDistill | IMU-only student · VIHand (7 IMUs) | MPJPE (mm): 13.69 -> 9.13 | Improved |
| use MANO-IMU teacher instead of RGB-IMU teacher | teacher model · VIHand (7 IMUs) | MPJPE (mm): 7.86 -> 5.84 | Improved |
| use generic KD methods (SimKD, DWPose) instead of decoupled distillation (ablation) | student · VIHand (7 IMUs) | MPJPE (mm): 15.40 -> 13.78 (SimKD), 13.61 (DWPose) | Improved |
Closest papers by meaning
- Prior-aware Dynamic Temporal Modeling Framework for Sequential 3D Hand Pose Estimation (ICCV 2025 · similarity 0.87)
- UniHand: A Unified Model for Diverse Controlled 4D Hand Motion Modeling (ICLR 2026 · similarity 0.85)
- Dense 3D Regression for Hand Pose Estimation (CVPR 2018 · similarity 0.85)
- HandMCM: Multi-modal Point Cloud-based Correspondence State Space Model for 3D Hand Pose Estimation (AAAI 2026 · similarity 0.85)
- HiPoser: 3D Human Pose Estimation with Hierarchical Shared Learning at Parts-Level Using Inertial Measurement Units (AAAI 2025 · similarity 0.85)
- StableHand: Quality-Aware Flow Matching for World-Space Dual-Hand Motion Estimation from Egocentric Video (NeurIPS 2026 · 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.
