Research map / 3D vision / Human pose estimation
Towards Viewpoint Invariant 3D Human Pose Estimation
Albert Haque, Boya Peng, Zelun Luo, Alexandre Alahi, Serena Yeung, Li Fei-Fei
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 LSTM recurrent connections to iterative error feedback (ablation) | Iterative feedback · ITOP top-view, Upper Body | mAP at 10 cm: 51.7 -> 91.4 | Improved |
| add LSTM recurrent connections to iterative error feedback (ablation) | Iterative feedback · ITOP front-view, Upper Body | mAP at 10 cm: 61.0 -> 84.0 | Improved |
| add LSTM recurrent connections to iterative error feedback (ablation) | Iterative feedback · ITOP front-view, Full Body | mAP at 10 cm: 71.0 -> 77.4 | Improved |
| add LSTM recurrent connections to iterative error feedback (ablation) | Iterative feedback · ITOP top-view, Full Body | mAP at 10 cm: 51.2 -> 75.5 | Improved |
| add iterative error feedback instead of direct single-pass regression (ablation) | Direct prediction · ITOP front-view, Full Body | mAP at 10 cm: 21.8 -> 71.0 | Improved |
| add iterative error feedback instead of direct single-pass regression (ablation) | Direct prediction · ITOP top-view, Upper Body | mAP at 10 cm: 17.8 -> 51.7 | Improved |
| use glimpse inputs instead of heatmap (indicator mask) inputs (ablation) | Recurrent error feedback · ITOP localization error vs iterations | Localization error | Improved |
| use learned viewpoint-invariant model instead of random tree walk on frontal-to-top transfer | Random tree walk · ITOP viewpoint transfer, Full Body | mAP at 10 cm: 2.0 -> 20.4 | Improved |
| use recurrent error feedback instead of random tree walk on front view full body | Random tree walk · ITOP front-view, Full Body | mAP at 10 cm: 80.5 -> 77.4 | Hurt |
| use recurrent error feedback instead of random tree walk on top view upper body | Random tree walk · ITOP top-view, Upper Body | mAP at 10 cm: 84.8 -> 91.4 | Improved |
| use deep model instead of random tree walk regarding runtime | Random tree walk · Runtime per frame | seconds per frame: 0.1 -> 1.7 | Hurt |
Closest papers by meaning
- A Simple yet Effective Baseline for 3D Human Pose Estimation (ICCV 2017 · similarity 0.90)
- Coarse-To-Fine Volumetric Prediction for Single-Image 3D Human Pose (CVPR 2017 · similarity 0.89)
- Lifting From the Deep: Convolutional 3D Pose Estimation From a Single Image (CVPR 2017 · similarity 0.89)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.89)
- PoseIRM: Enhance 3D Human Pose Estimation on Unseen Camera Settings via Invariant Risk Minimization (CVPR 2024 · similarity 0.88)
- NOPE: Novel Object Pose Estimation from a Single Image (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.
