Research map / Learning theory and algorithms / Combinatorial and search optimization
Accelerating Globally Optimal Consensus Maximization in Geometric Vision
X. Zhang, Liangzu Peng, Wanting Xu, Laurent Kneip
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 |
|---|---|---|---|
| replace plain 1-D branch and bound with interval stabbing (ACM-0) | branch-and-bound consensus maximization · Synthetic 3D-2D registration (IMU-aided, 1-DoF) | runtime: 200x-300x speed-up | Improved |
| replace plain 1-D branch and bound with interval stabbing (ACM-0) | branch-and-bound consensus maximization · Synthetic 3D-2D registration (IMU-aided, 1-DoF) | angular error: similar angular errors | No clear effect |
| branch over one dimension and solve the other by interval stabbing (ACM-1) | 2D-2D planar relative pose BnB · Synthetic 2D-2D registration | runtime: 2x-20x speed-up | Improved |
| use ACM-1 instead of plain BnB | 2D-2D planar relative pose BnB · Synthetic 2D-2D registration | angular error: similarly low angular errors for both angles | No clear effect |
| use ACM-1 instead of plain BnB at extreme outlier rates | 2D-2D planar relative pose BnB · Synthetic 2D-2D registration (extreme outlier rates) | runtime: 20x speed-up | Improved |
| use ACM-1 instead of plain BnB on low-outlier real data | 2D-2D planar relative pose BnB · KITTI (11 sequences, SIFT correspondences) | runtime: about 2x speed-up | Improved |
| use ACM-1 instead of plain BnB | 2D-2D planar relative pose BnB · KITTI (11 sequences, SIFT correspondences) | iterations: 16,234 -> 1,692 on sequence 00 | Improved |
| use ACM-1 instead of plain BnB | 2D-2D planar relative pose BnB · KITTI (11 sequences, SIFT correspondences) | rotation and translation error: 0.1304 -> 0.1293 deg rotation on sequence 00 | No clear effect |
| use ACM-1 instead of plain BnB when outliers are few | 2D-2D planar relative pose BnB · KITTI (11 sequences, SIFT correspondences) | speed-up ratio: 1.5778 to 2.1045 | Mixed |
| use ACM-1 as outlier ratio grows | ACM-1 · Synthetic 2D-2D registration | runtime stability: about 0.03 sec vs 0.0615 -> 0.5833 sec for plain BnB | Improved |
| use interval stabbing (extra log M cost per bound) | ACM-0 · Synthetic 3D-2D registration | iterations: 1,138 -> 1 at 80% outliers; 1,899 -> 1 at 95% | Improved |
Closest papers by meaning
- Efficient Globally Optimal Consensus Maximisation with Tree Search (TPAMI 2016 · similarity 0.91)
- Model-free Consensus Maximization for Non-Rigid Shapes (ECCV 2018 · similarity 0.90)
- An Exact Penalty Method for Locally Convergent Maximum Consensus (CVPR 2017 · similarity 0.90)
- Deterministic Approximate Methods for Maximum Consensus Robust Fitting (TPAMI 2019 · similarity 0.89)
- Deterministic Consensus Maximization with Biconvex Programming (ECCV 2018 · similarity 0.89)
- Convex Relaxations for Consensus and Non-Minimal Problems in 3D Vision (ICCV 2019 · 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.
