Research map / 3D vision / Human pose estimation
Solving the Blind Perspective-n-Point Problem End-To-End With Robust Differentiable Geometric Optimization
Dylan Campbell, Liu Liu, Stephen Gould
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 pose loss to correspondence loss (ablation) | BPnPNet (network output) · ModelNet40 test | median rotation error (deg): 11.3 -> 9.66 | Improved |
| add pose loss to correspondence loss (ablation) | BPnPNet (network output) · ModelNet40 test | median translation error: 0.52 -> 0.08 | Improved |
| add pose loss to correspondence loss (ablation) | BPnPNet (network output) · MegaDepth test | median rotation error (deg): 4.47 -> 3.31 | Improved |
| add pose loss to correspondence loss (ablation) | BPnPNet (network output) · MegaDepth test | median translation error: 1.05 -> 0.46 | Improved |
| use in-network RANSAC pose estimate instead of network output (ablation) | BPnPNet with pose loss · ModelNet40 test | median rotation error (deg): 9.66 -> 8.09 | Improved |
| use in-network RANSAC pose estimate instead of network output (ablation) | BPnPNet with pose loss · MegaDepth test | median rotation error (deg): 3.31 -> 1.00 | Improved |
| add pose loss when RANSAC estimate is used (ablation) | BPnPNet RANSAC output · ModelNet40 test | median rotation error (deg): 11.7 -> 8.09 | Improved |
| add pose loss when RANSAC estimate is used (ablation) | BPnPNet RANSAC output · MegaDepth test | median rotation error (deg): 1.55 -> 1.00 | Improved |
| add pose loss to correspondence loss (ablation) | BPnPNet (network output) · ModelNet40 test | recall at 15 deg rotation and 0.5 translation: 72%, an improvement of 25% | Improved |
| add pose loss to correspondence loss (ablation) | BPnPNet (network output) · MegaDepth test | recall at 10 deg rotation and 1 translation: 73%, 25% better | Improved |
| use in-network RANSAC pose estimate instead of network output (ablation) | BPnPNet with pose loss · MegaDepth test | recall at 10 deg rotation and 1 translation: 73% -> 82% | Improved |
| use end-to-end trained BPnPNet instead of SoftPOSIT | SoftPOSIT · MegaDepth test | median rotation error (deg): 21.4 -> 1.00 | Improved |
Closest papers by meaning
- End-to-End Learnable Geometric Vision by Backpropagating PnP Optimization (CVPR 2020 · similarity 0.90)
- Pose Estimation from Line Correspondences: A Complete Analysis and a Series of Solutions (TPAMI 2016 · similarity 0.89)
- Efficient Solution of Point-Line Absolute Pose (CVPR 2024 · similarity 0.84)
- Unifying Correspondence Pose and NeRF for Generalized Pose-Free Novel View Synthesis (CVPR 2024 · similarity 0.84)
- Sparse-view Pose Estimation and Reconstruction via Analysis by Generative Synthesis (NeurIPS 2024 · similarity 0.84)
- From Correspondences to Pose: Non-minimal Certifiably Optimal Relative Pose without Disambiguation (CVPR 2024 · 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.
