Research map / 3D vision / Human pose estimation
EM-POSE: 3D Human Pose Estimation From Sparse Electromagnetic Trackers
Manuel Kaufmann, Yi Zhao, Chengcheng Tang, Lingling Tao, Christopher Twigg, Jie Song, Robert Wang, Otmar Hilliges
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 subject-specific rotational and translational sensor offsets in training (ablation) | Ours 12 · EM-POSE real test set (subjects 1-4) | MPJPE: 167.6 -> 31.8 | Improved |
| add subject-specific translational offsets t in training (ablation) | Ours 12 · EM-POSE real test set (subjects 1-4) | MPJPE: 35.6 -> 31.8 | Improved |
| add position measurements to orientation-only input (ablation) | Ours 12 · EM-POSE real test set (subjects 1-4) | MPJPE: 50.8 -> 31.8 | Improved |
| add orientation measurements to position-only input (ablation) | Ours 12 · EM-POSE real test set (subjects 1-4) | MPJAE: 16.2 -> 13.3 | Improved |
| add RNN providing the initial estimate to LGD (ablation) | Ours 12 · EM-POSE real test set (subjects 1-4) | MPJPE: 36.9 -> 31.8 | Improved |
| replace BiRNN regression with LGD RNN hybrid, 12 sensors | pure learning baseline · EM-POSE real test set (subjects 1-4) | MPJPE: 37.3 -> 31.8 | Improved |
| replace BiRNN regression with LGD RNN hybrid, 6 sensors | pure learning baseline · EM-POSE real test set (subjects 1-4) | MPJPE: 36.3 -> 35.4 | Improved |
| use temporal BiRNN instead of frame-wise ResNet, 12 sensors | learning baseline · EM-POSE real test set (subjects 1-4) | MPJPE: 41.5 -> 37.3 | Improved |
| use 12 sensors instead of 6 | Ours (LGD RNN) · EM-POSE real test set (subjects 1-4) | MPJPE: 35.4 -> 31.8 | Improved |
| add orientations and subject offsets to MoSh++-style optimization | optimization baseline · EM-POSE real test set (subjects 1-4) | MPJPE: 56.9 -> 44.2 | Improved |
| replace pure optimization with learned gradient descent | pos + ori 12 · EM-POSE real test set (subjects 1-4) | MPJPE: 44.2 -> 31.8 | Improved |
| replace BiRNN with LGD RNN on unseen subject, 12 sensors | pure learning baseline · EM-POSE real test set (subject 5, cross-subject) | MPJPE: 40.7 -> 32.1 | Improved |
Closest papers by meaning
- emg2pose: A Large and Diverse Benchmark for Surface Electromyographic Hand Pose Estimation (NeurIPS 2024 · similarity 0.84)
- EMHI: A Multimodal Egocentric Human Motion Dataset with HMD and Body-Worn IMUs (AAAI 2025 · similarity 0.83)
- HMD-Poser: On-Device Real-time Human Motion Tracking from Scalable Sparse Observations (CVPR 2024 · similarity 0.83)
- UMotion: Uncertainty-driven Human Motion Estimation from Inertial and Ultra-wideband Units (CVPR 2025 · similarity 0.83)
- Full-Body Motion Reconstruction with Sparse Sensing from Graph Perspective (AAAI 2024 · similarity 0.82)
- HiPoser: 3D Human Pose Estimation with Hierarchical Shared Learning at Parts-Level Using Inertial Measurement Units (AAAI 2025 · similarity 0.82)
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.
