Research map / 3D vision / Human pose estimation
EgoPoseFormer: A Simple Baseline for Stereo Egocentric 3D Human Pose Estimation
Chenhongyi Yang, Anastasia Tkach, Shreyas Hampali, Linguang Zhang, Elliot J Crowley, Cem Keskin
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 PRFormer with Deformable Stereo Attention and self-attention on top of pose proposals (ablation) | EgoPoseFormer PPN · UnrealEgo (stereo) | MPJPE: 46.7 -> 33.4 | Improved |
| PRFormer with self-attention only (ablation) | EgoPoseFormer · UnrealEgo (stereo) | MPJPE: 46.9 -> 45.9 | No clear effect |
| add Deformable Stereo Attention to PRFormer (ablation) | EgoPoseFormer · UnrealEgo (stereo) | MPJPE: 45.9 -> 33.4 | Improved |
| add self-attention among joint query tokens to DS-Attn (ablation) | EgoPoseFormer PRFormer · UnrealEgo (stereo) | MPJPE: 38.1 -> 33.4 | Improved |
| use stereo input instead of monocular left view (ablation) | EgoPoseFormer · UnrealEgo | MPJPE: 48.3 -> 33.4 | Improved |
| increase PRFormer layers from 3 to 4 (ablation) | EgoPoseFormer · UnrealEgo (stereo) | MPJPE: 33.4 -> 33.0 | No clear effect |
| use larger backbone (ResNet-18 to ResNet-101) (ablation) | EgoPoseFormer · UnrealEgo | MPJPE: 33.4 -> 27.9 | Improved |
| use larger backbone (ResNet-18 to ResNet-101) (ablation) | UnrealEgo heatmap baseline · UnrealEgo | MPJPE: 79.1 -> 80.1 | No clear effect |
| apply AdamW and EgoPoseFormer training recipe to baseline UnrealEgo (ablation) | UnrealEgo baseline · UnrealEgo | MPJPE: 79.1 -> 70.5 | Improved |
| use AdamW instead of Adam in PPN training (ablation) | Pose Proposal Network · UnrealEgo | MPJPE: 168.9 -> 59.6 | Improved |
| add 2D heatmap pre-training of feature extractor (ablation) | Pose Proposal Network · UnrealEgo | MPJPE: 59.6 -> 43.5 | Improved |
| add 2D heatmap pre-training (ablation) | EgoPoseFormer · SceneEgo (monocular) | MPJPE: 122.9 -> 93.0 | Improved |
Later papers that compared with this method
4 later accepted papers report EgoPoseFormer as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- EgoPoseFormer v2: Accurate Egocentric Human Motion Estimation for AR/VR (CVPR 2026 · better on 6 of 6 shared results)
- E-3DPSM: A State Machine for Event-based Egocentric 3D Human Pose Estimation (CVPR 2026 · better on 4 of 5 shared results)
- Egocentric Visibility-Aware Human Pose Estimation (CVPR 2026 · better on 2 of 2 shared results)
- Bring Your Rear Cameras for Egocentric 3D Human Pose Estimation (ICCV 2025 · better on 8 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- 3D Human Pose Perception from Egocentric Stereo Videos (CVPR 2024 · similarity 0.94)
- SelfPose: 3D Egocentric Pose Estimation From a Headset Mounted Camera (TPAMI 2020 · similarity 0.92)
- EgoPoseFormer v2: Accurate Egocentric Human Motion Estimation for AR/VR (CVPR 2026 · similarity 0.92)
- Scene-Aware Egocentric 3D Human Pose Estimation (CVPR 2023 · similarity 0.91)
- FusionFormer: A Concise Unified Feature Fusion Transformer for 3D Pose Estimation (AAAI 2024 · similarity 0.89)
- Attention-Propagation Network for Egocentric Heatmap to 3D Pose Lifting (CVPR 2024 · similarity 0.89)
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.
