Research map / Learning theory and algorithms / Matrix and tensor methods
Linearly Solving Robust Rotation Estimation
Yinlong Liu, Tianyu Huang, Zhi-Xin Yang
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 rotation voting instead of ARCS decomposition-based methods when same-axis structured outliers are numerous | robust rotation estimation · Synthetic, 10% inlier ratio, same-axis ratio 40% | success rate: 86% -> 100% | Improved |
| use rotation voting instead of ARCS under 20% inlier ratio and 40% same-axis ratio | robust rotation estimation · Synthetic, 20% inlier ratio | success rate: 96% -> 100% | Improved |
| use rotation voting instead of RANSAC at 5% inlier ratio | robust rotation estimation · Synthetic, 5% inlier ratio, same-axis ratio 5% | success rate: 92% -> 100% | Improved |
| use rotation voting instead of RANSAC at 20% inlier ratio | robust rotation estimation · Synthetic, 20% inlier ratio, same-axis ratio 40% | running time (s): 1.66 -> 0.04 | Improved |
| increase same-axis outlier ratio from 5% to 40% for ARCS+O at 5% inlier ratio | ARCS+O · Synthetic, 5% inlier ratio | success rate: 100% -> 84% | Hurt |
| increase same-axis outlier ratio for rotation voting (5% to 40%) | rotation voting · Synthetic, inlier ratios 5-20% | success rate: 100% -> 100% | No clear effect |
| use rotation voting instead of GORE/TEASER++ on 1e5 observations | robust rotation estimation · Synthetic, N=1e5 | feasibility: GORE >=1000 seconds, TEASER++ out of memory -> 100%|0.04 | Improved |
| use rotation voting instead of other robust solvers on PASSTA panorama pairs | robust rotation estimation · PASSTA panorama matching | average time (s): 0.04 -> 0.01 | Improved |
| use rotation voting instead of ARCS+O, RANSAC on PASSTA | robust rotation estimation · PASSTA panorama matching | success rate: 100% -> 100% | No clear effect |
| replace SVD-based closed form with quaternion-circle linear closed form | outlier-free rotation estimation · Synthetic, N=1e3, noise 0.00-0.10 | rotation error | No clear effect |
| replace SVD with quaternion-circle linear closed form | outlier-free rotation estimation · Synthetic, N=1e3 | running time: more than x5 times slower | Hurt |
| increase outlier ratio in pairwise 6-DoF pipeline (inlier check removes outliers) | rotation voting 6DoF · Synthetic 6DoF, N=5000 | running time | Improved |
Closest papers by meaning
- Robust Relative Rotation Averaging (TPAMI 2017 · similarity 0.86)
- A Quaternion-Based Certifiably Optimal Solution to the Wahba Problem With Outliers (ICCV 2019 · similarity 0.86)
- Certifiably Optimal Anisotropic Rotation Averaging (ICCV 2025 · similarity 0.86)
- Certifiably Optimal Outlier-Robust Geometric Perception: Semidefinite Relaxations and Scalable Global Optimization (TPAMI 2022 · similarity 0.85)
- Hybrid Rotation Averaging: A Fast and Robust Rotation Averaging Approach (CVPR 2021 · similarity 0.85)
- Symmetry-Robust 3D Orientation Estimation (ICML 2025 · similarity 0.85)
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.
