Research map / 3D vision / Human pose estimation
SelfPose: 3D Egocentric Pose Estimation From a Headset Mounted Camera
Denis Tomè, Thiemo Alldieck, Patrick Peluse, Gerard Pons‐Moll, Lourdes Agapito, Hernán Badino, Fernando De la Torre
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 heatmap regression branch to 3D pose branch (ablation) | multi-branch auto-encoder · xR-EgoPose test (synthetic) | MPJPE (mm): 130.4 -> 58.2 | Improved |
| add rotation regression branch to 3D pose branch (ablation) | multi-branch auto-encoder · xR-EgoPose test (synthetic) | MPJPE (mm): 130.4 -> 91.2 | Improved |
| add rotation branch to p3d+hm (ablation) | multi-branch auto-encoder · xR-EgoPose test (synthetic) | MPJPE (mm): 58.2 -> 54.7 | Mixed |
| use multi-branch auto-encoder instead of Martinez et al. lifting (ablation) | multi-branch auto-encoder · xR-EgoPose test (synthetic) | MPJPE (mm): 122.1 -> 54.7 | Improved |
| add ImageNet pre-training to ResNet 50 heatmap network (ablation) | ResNet 50 · xR-EgoPose test (synthetic) | MPJPE (mm): 54.7 -> 51.1 | Improved |
| use pre-trained U-Net instead of pre-trained ResNet 50 for 2D heatmaps (ablation) | multi-branch auto-encoder · xR-EgoPose test (synthetic) | MPJPE (mm): 51.1 -> 41.0 | Improved |
| use embedding size 50 instead of 10 (ablation) | U-Net (p) · xR-EgoPose test, embedding size, U-Net (p) | MPJPE (mm): 45.8 -> 41.0 | Improved |
| increase embedding size from 50 to 500 (ablation) | U-Net (p) · xR-EgoPose test, embedding size, U-Net (p) | MPJPE (mm): 41.0 -> 41.2 | Hurt |
| use full heatmap size 48 instead of 8 (ablation) | U-Net (p) · xR-EgoPose test, heatmap size, U-Net (p) | MPJPE (mm): 41.7 -> 41.0 | Improved |
| add 2D-only annotated images to training (ablation) | multi-branch auto-encoder · xR-EgoPose test | MPJPE (mm) | Improved |
| use U-Net (p) instead of ResNet 50 in Mo2Cap2 indoor (ablation) | multi-branch auto-encoder · Mo2Cap2 indoor | MPJPE total (mm): 48.16 -> 46.61 | Improved |
| use U-Net (p) instead of ResNet 50 in Mo2Cap2 outdoor (ablation) | multi-branch auto-encoder · Mo2Cap2 outdoor | MPJPE total (mm): 60.19 -> 54.61 | Improved |
Closest papers by meaning
- xR-EgoPose: Egocentric 3D Human Pose From an HMD Camera (ICCV 2019 · similarity 0.98)
- Egocentric Pose Estimation From Human Vision Span (ICCV 2021 · similarity 0.93)
- 3D Human Pose Perception from Egocentric Stereo Videos (CVPR 2024 · similarity 0.91)
- EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity (NeurIPS 2024 · similarity 0.90)
- FRAME: Floor-aligned Representation for Avatar Motion from Egocentric Video (CVPR 2025 · similarity 0.90)
- EgoGen: An Egocentric Synthetic Data Generator (CVPR 2024 · 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.
