Research map / 3D vision / Human pose estimation
Egocentric Pose Estimation From Human Vision Span
Hao Jiang, Vamsi Krishna Ithapu
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 Stage 2 3D pose refinement (ablation) | Stage 1 egopose model · Synthetic data (60 test sequences) | 3D keypoint error (cm): 12.53 -> 11.76 | Improved |
| add shape features to motion features (ablation) | MotionOnly · Synthetic data (60 test sequences) | 3D keypoint error (cm): 14.55 -> 11.76 | Improved |
| add motion history features to shape features (ablation) | ShapeOnly · Synthetic data (60 test sequences) | 3D keypoint error (cm): 13.89 -> 11.76 | Improved |
| add camera height to the motion history image (ablation) | NoHeight · Synthetic data (60 test sequences) | 3D keypoint error (cm): 14.52 -> 11.76 | Improved |
| add camera height to the motion history image (ablation) | NoHeight · Synthetic data (60 test sequences) | head orientation vector 1 error (deg): 13.01 -> 11.26 | Improved |
| use body shape input instead of hand keypoint map (ablation) | HandMap · Synthetic data (60 test sequences) | 3D keypoint error (cm): 26.10 -> 11.76 | Improved |
| use non-recurrent network instead of RNN in Stage 1 (ablation) | Stage1 (RNN variant) · Synthetic data (60 test sequences) | 3D keypoint error (cm): 12.52 -> 12.53 | No clear effect |
| use non-recurrent network instead of RNN with refinement (ablation) | Full model vs RNN variant · Synthetic data (60 test sequences) | 3D keypoint error (cm): 11.76 | Improved |
| add Stage 2 refinement (head pose unchanged) (ablation) | Stage 1 model · Synthetic data (60 test sequences) | head orientation vector 1 error (deg): 11.26 -> 11.26 | No clear effect |
| replace xr-egopose with proposed model | xr-egopose · Synthetic data (60 test sequences) | 3D keypoint error (cm): 13.26 -> 11.76 | Improved |
| add synthetic data to real training data | Ours I (real only) · Real video test | 3D keypoint error: 15.70 -> 14.87 | Improved |
| add synthetic data to real training data | xr-egopose I · Real video test | 3D keypoint error: 16.86 -> 17.29 | Hurt |
Closest papers by meaning
- SelfPose: 3D Egocentric Pose Estimation From a Headset Mounted Camera (TPAMI 2020 · similarity 0.93)
- EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity (NeurIPS 2024 · similarity 0.91)
- FRAME: Floor-aligned Representation for Avatar Motion from Egocentric Video (CVPR 2025 · similarity 0.90)
- Estimating Ego-Body Pose from Doubly Sparse Egocentric Video Data (NeurIPS 2024 · similarity 0.89)
- 3D Human Pose Perception from Egocentric Stereo Videos (CVPR 2024 · similarity 0.88)
- Estimating Body and Hand Motion in an Ego-sensed World (CVPR 2025 · similarity 0.88)
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.
