Research map / Learning theory and algorithms / Combinatorial and search optimization
Certifiably Optimal Anisotropic Rotation Averaging
Carl Olsson, Yaroslava Lochman, Johan Malmport, Christopher Zach
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 conv(SO(3)) constraints to the anisotropic SDP (SDP-O(3)-aniso -> SDP-cSO(3)) (ablation) | SDP-O(3)-aniso · LU Sphinx (real SfM) | rotation error: 18.6037 -> 0.0740 | Improved |
| add conv(SO(3)) constraints to the anisotropic SDP (ablation) | SDP-O(3)-aniso · Temple Singapore (real SfM) | rotation error: 26.0824 -> 0.1696 | Improved |
| add conv(SO(3)) constraints to the anisotropic SDP: rank of solution on 1000 synthetic instances (ablation) | SDP-O(3)-aniso · synthetic anisotropic (1000 instances) | rank-3 recovery: no rank-3 solutions -> rank 3 in all cases | Improved |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints) | SDP-O(3)-iso · Tsar Nikolai I (real SfM) | rotation error: 0.1170 -> 0.0534 | Improved |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints) | SDP-O(3)-iso · UWO (real SfM) | rotation error: 0.3142 -> 0.2274 | Improved |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints) | SDP-O(3)-iso · King's College (real SfM) | rotation error: 0.1656 -> 0.0796 | Improved |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints) | SDP-O(3)-iso · Kronan (real SfM) | rotation error: 0.2149 -> 0.3892 | Hurt |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints) | SDP-O(3)-iso · Mus. Barcelona (real SfM) | rotation error: 0.2255 -> 0.1310 | Improved |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints), Mahalanobis error | SDP-O(3)-iso · UWO (real SfM) | Mahalanobis error: 1.481 -> 0.727 | Improved |
| replace isotropic chordal cost with anisotropic cost (with cSO(3) constraints), Mahalanobis error | SDP-O(3)-iso · Tsar Nikolai I (real SfM) | Mahalanobis error: 0.687 -> 0.188 | Improved |
| use proposed anisotropic cost and conv(SO(3)) relaxation versus isotropic baselines on synthetic data | SDP-cSO(3) · synthetic (2-100 cameras) | rotation error | Improved |
| add conv(SO(3)) constraints (also to isotropic cost) for solver speed (ablation) | SDP-cSO(3)-iso · synthetic (2-100 cameras) | runtime | Improved |
Closest papers by meaning
- Hybrid Rotation Averaging: A Fast and Robust Rotation Averaging Approach (CVPR 2021 · similarity 0.90)
- Robust Relative Rotation Averaging (TPAMI 2017 · similarity 0.90)
- Shonan Rotation Averaging: Global Optimality by Surfing SO(p)(n) (ECCV 2020 · similarity 0.89)
- Rotation Averaging and Strong Duality (CVPR 2018 · similarity 0.89)
- Rotation Averaging in a Split Second: A Primal-Dual Method and a Closed-Form for Cycle Graphs (ICCV 2021 · similarity 0.89)
- On the Distribution of Minima in Intrinsic-Metric Rotation Averaging (CVPR 2020 · similarity 0.88)
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.
