Research map / 3D vision / Human pose estimation
Equivalent Constraints for Two-View Geometry: Pose Solution/Pure Rotation Identification and 3D Reconstruction
Qi Cai, Yuanxin Wu, Lilian Zhang, Peike Zhang
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 the two-inequality pose identification instead of traditional reconstruction plus positive-depth check | two-view pose solution selection · Simulation (d=30m, 50m, 80m) | translation discrepancy difference: difference almost all above zero | Improved |
| use analytical depth reconstruction instead of DLT triangulation | 3D point reconstruction · Simulation, small parallax and large noise | reconstruction error ratio: ratio much less than 1 | Improved |
| use analytical reconstruction instead of DLT triangulation at large parallax | 3D point reconstruction · Simulation, large parallax | reconstruction error ratio | No clear effect |
| use analytical reconstruction instead of traditional triangulation | 3D point reconstruction · Real barcode tests (21 view pairs) | reconstruction error ratio: better for all pairs except the 3-4 pair | Improved |
| initialize bundle adjustment with analytical instead of traditional reconstruction | bundle adjustment · Real barcode tests, 1-4 view pair | BA convergence: well-behaved BA vs diverging BA | Improved |
| add bundle adjustment after analytical reconstruction | analytical reconstruction · Real barcode tests (21 view pairs) | reconstruction error ratio | No clear effect |
| use PRI (Proposition 8) instead of the other rotation indicator for pure rotation identification (ablation) | pure rotation identification · Simulation, d=30m, 50m, 80m | threshold separability: threshold 0.015 works for all depths | Improved |
| use inequality-based identification (Methods I and II) instead of traditional DLT with positive depth check | pose identification · Computation cost vs number of feature points (20 Monte Carlo runs) | time cost: reduce the time cost by about 7-9 times | Improved |
Closest papers by meaning
- Revisiting the P3P Problem (CVPR 2023 · similarity 0.86)
- Trifocal Relative Pose From Lines at Points (TPAMI 2022 · similarity 0.86)
- A Pose-Only Geometric Constraint for Multi-Camera Pose Adjustment (TPAMI 2026 · similarity 0.85)
- From Correspondences to Pose: Non-minimal Certifiably Optimal Relative Pose without Disambiguation (CVPR 2024 · similarity 0.85)
- Practical Solutions to the Relative Pose of Three Calibrated Cameras (CVPR 2025 · similarity 0.84)
- EquiPose: Exploiting Permutation Equivariance for Relative Camera Pose Estimation (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.
