Research map / 3D vision / Human pose estimation
Orientation Keypoints for 6D Human Pose Estimation
Martin Fisch, Ronald Clark
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 |
|---|---|---|---|
| use 3D detections with SVD instead of 2D detections with PnP (ablation) | orientation keypoints · Human3.6M rotations | MPJPAS rad: 0.265 -> 0.213 | Improved |
| add orientation keypoints to joint keypoints (lifter, ground-truth 2D) (ablation) | lifter · Human3.6M | MPJPE mm: 48.3 -> 33.3 | Improved |
| add orientation keypoints to Crosshairs 2D detections (ablation) | lifter · Human3.6M | MPJPE mm: 50.6 -> 47.1 | Improved |
| add orientation keypoints to regression 2D detections (ablation) | lifter · Human3.6M | MPJPE mm: 57.8 -> 53.0 | Improved |
| use Crosshairs 3D detector instead of Crosshairs 2D (ablation) | lifter with OKPs · Human3.6M | MPJPE mm: 47.1 -> 44.6 | Improved |
| use Crosshairs instead of regression 2D detector (ablation) | lifter with OKPs · Human3.6M | MPJPE mm: 53.0 -> 47.1 | Improved |
| raise resolution from 256x192 to 384x288 (ablation) | Crosshairs 1-layer XY · Human3.6M | MPJPE-17 P1 mm: 52.2 -> 48.1 | Improved |
| use 4-layer instead of 1-layer XY crosshairs (ablation) | Crosshairs 384x288 · Human3.6M | MPJPE-17 P1 mm: 48.1 -> 48.2 | No clear effect |
| add second-stage refiner (ablation) | Crosshairs 384x288 1-layer · Human3.6M | MPJPE-17 P1 mm: 48.1 -> 45.5 | Improved |
| train with feet/hands included (21j, 19r) (ablation) | Crosshairs 384x288 4-layer · Human3.6M | MPJPE-17 P1 mm: 48.2 -> 47.4 | Improved |
| add centroid regularizer loss (ablation) | Crosshairs · Human3.6M | accuracy | No clear effect |
| fine-tune on 3DPW and use orientation keypoints | orientation kps · 3DPW | MPJPE mm: 115.4 -> 70.7 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- RepNet — RepNet: Weakly Supervised Training of an Adversarial Reprojection Network for 3D Human Pose Estimation (CVPR 2019) · better on 1 of 2 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
- FlexPose: Pose Distribution Adaptation with Limited Guidance (AAAI 2025 · similarity 0.86)
- A Comparative Assessment of Accuracy in Video-Based Monocular Human Pose Estimation Frameworks (TPAMI 2026 · similarity 0.86)
- Human Pose Estimation Using Global and Local Normalization (ICCV 2017 · similarity 0.86)
- ProbPose: A Probabilistic Approach to 2D Human Pose Estimation (CVPR 2025 · similarity 0.84)
- Optimizing Human Pose Estimation Through Focused Human and Joint Regions (AAAI 2025 · similarity 0.84)
- Reconstructing Humans with a Biomechanically Accurate Skeleton (CVPR 2025 · similarity 0.84)
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.
