Research map / 3D vision / Human pose estimation
Trifocal Relative Pose From Lines at Points
Ricardo Fabbri, Timothy Duff, Hongyi Fan, Margaret H. Regan, David da Costa de Pinho, Elias Tsigaridas, Charles W. Wampler, Jonathan D. Hauenstein, Peter Giblin, Benjamin B. Kimia, Anton Leykin, Tomáš Pajdla
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 |
|---|---|---|---|
| raise maximum number of corrector steps from 4 to 5-7 (ablation) | MINUS · Synthetic solves | runtime: 4 -> 5-7 steps cuts runtime to 280 ms | Improved |
| loosen corrector tolerance by 10^4x (ablation) | MINUS · Synthetic solves | runtime vs error rate: runtime under 200 ms, error rate as high as 50% | Mixed |
| cap predictor iterations per root at N_max=500 (ablation) | MINUS · Synthetic solves | success rate / runtime: success 97%, average 401ms | Mixed |
| rerun twice with N_max=200 instead of once with 500 (ablation) | MINUS · Synthetic solves | success rate: similar success rate | No clear effect |
| use positive depth and third tangent verification to prune solutions (ablation) | MINUS · Synthetic solves | number of solutions: 35 real solutions -> 7 -> about 3-4 | Improved |
| stop at first root meeting positive depth and third tangent when all points have tangents (ablation) | MINUS · Synthetic solves | runtime / success rate: run time on the order of 100 ms with modest decrease in success rate | Mixed |
| use MINUS specialized HC instead of general-purpose HC | trifocal solver · Synthetic solves | runtime: over 1 minute -> 440 ms | Improved |
| use trifocal solver with SIFT orientation instead of COLMAP bifocal pipeline | mugs featureless dataset · Mugs featureless dataset (20 triplets) | triplets with recovered pose: 5 -> 10 out of 20 | Improved |
| compare MINUS with other trifocal methods on rotation error | MINUS · Fountain P-11 / Herz-Jesu-P8 (Chicago problem) | rotation error (deg): 0.257 -> 0.137 | Improved |
| compare MINUS with other trifocal methods on translation error | MINUS · Fountain P-11 / Herz-Jesu-P8 (Chicago problem) | translation error (deg): 0.534 -> 0.673 | No clear effect |
Closest papers by meaning
- TRPLP - Trifocal Relative Pose From Lines at Points (CVPR 2020 · similarity 0.97)
- Minimal Solutions to Generalized Three-View Relative Pose Problem (ICCV 2023 · similarity 0.91)
- Practical Solutions to the Relative Pose of Three Calibrated Cameras (CVPR 2025 · similarity 0.89)
- General Planar Motion from a Pair of 3D Correspondences (ICCV 2023 · similarity 0.88)
- From Correspondences to Pose: Non-minimal Certifiably Optimal Relative Pose without Disambiguation (CVPR 2024 · similarity 0.87)
- Efficient Solution of Point-Line Absolute Pose (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.
