Research map / 3D vision / Human pose estimation
3D Ego-Pose Estimation via Imitation Learning
Ye Yuan, Kris Kitani
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 policy fine-tuning (20 iterations) on the test sequence, starting from estimated initial state (ablation) | Ours (video-conditioned policy) · Virtual test set (20 trajectories) | Pose error: 1.2219 -> 0.9779 | Improved |
| add policy fine-tuning (20 iterations) on the test sequence, starting from estimated initial state (ablation) | Ours (video-conditioned policy) · Virtual test set (20 trajectories) | Velocity error: 7.5337 -> 6.5143 | Improved |
| add policy fine-tuning (20 iterations) on the test sequence, starting from estimated initial state (ablation) | Ours (video-conditioned policy) · Virtual test set (20 trajectories) | Smoothness: 12.2472 -> 11.9876 | Improved |
| use ground-truth initial state instead of estimated initial state (ablation) | Ours (video-conditioned policy) · Virtual test set (20 trajectories) | Pose error: 0.9779 -> 0.6688 | Improved |
| use ground-truth initial state instead of estimated initial state (ablation) | Ours (video-conditioned policy) · Virtual test set (20 trajectories) | Velocity error: 6.5143 -> 6.0761 | Improved |
| use ground-truth initial state instead of estimated initial state (ablation) | Ours (video-conditioned policy) · Virtual test set (20 trajectories) | Smoothness: 11.9876 -> 12.2968 | No clear effect |
| physics-based control policy instead of direct pose regression | Ego-pose estimator · Virtual test set (20 trajectories) | Smoothness: 36.1628 -> 11.9876 | Improved |
| physics-based control policy instead of direct pose regression | Ego-pose estimator · Virtual test set (20 trajectories) | Pose error: 0.8310 -> 0.9779 | Hurt |
| physics-based control policy instead of Path pose | Ego-pose estimator · Virtual test set (20 trajectories) | Pose error: 1.7643 -> 0.9779 | Improved |
| physics-based control policy with fine-tuning instead of pose regression on real videos | Ego-pose estimator · Real-world egocentric videos (12 sequences) | 2D projection error: 0.1621 -> 0.1325 | Improved |
| physics-based control policy with fine-tuning instead of pose regression on real videos | Ego-pose estimator · Real-world egocentric videos (12 sequences) | Smoothness: 44.11 -> 11.54 | Improved |
Closest papers by meaning
- EgoSim: An Egocentric Multi-view Simulator and Real Dataset for Body-worn Cameras during Motion and Activity (NeurIPS 2024 · similarity 0.90)
- Estimating Ego-Body Pose from Doubly Sparse Egocentric Video Data (NeurIPS 2024 · similarity 0.89)
- Estimating Body and Hand Motion in an Ego-sensed World (CVPR 2025 · similarity 0.89)
- Whole-Body Conditioned Egocentric Video Prediction (NeurIPS 2025 · similarity 0.88)
- FRAME: Floor-aligned Representation for Avatar Motion from Egocentric Video (CVPR 2025 · similarity 0.87)
- 3D Human Pose Perception from Egocentric Stereo Videos (CVPR 2024 · similarity 0.86)
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.
