Research map / Learning theory and algorithms / Combinatorial and search optimization
Fast Globally Optimal Surface Normal Estimation from an Affine Correspondence
Levente Hajder, Lajos Lóczi, Daniel Barath
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 proposed linear closed-form solver instead of the cubic optimal solver | optimal surface normal estimation · Cambridge Landmarks | run time (ns): 2484 -> 596 | Improved |
| use the proposed linear solver instead of the cubic optimal solver | optimal surface normal estimation · Cambridge Landmarks | AUC@10 deg: 27.3 -> 27.3 | No clear effect |
| use the optimal solver instead of the fastest non-optimal FNE | normal estimation · Cambridge Landmarks | average angular error (deg): 21.0 -> 8.8 | Improved |
| use the optimal solver instead of FNESUPER | normal estimation · Cambridge Landmarks | run time (ns): 1721 -> 596 | Improved |
| use the optimal solver on ScanNet | normal estimation · ScanNet | run time (ns): 3146 -> 747 | Improved |
| add surface normals via P1AC solver using OPTLINEAR instead of P3P | visual localization with GC-RANSAC · Cambridge Landmarks | median position error (cm): 13.5 -> 1.8 | Improved |
| use OPTLINEAR normals instead of ground-truth normals in P1AC | visual localization · Cambridge Landmarks | pose AUC@(1 deg, 10cm): 48.0 vs 48.4 | No clear effect |
| use P1AC with OPTLINEAR normals instead of P3P on Shop Facade | visual localization · Cambridge Landmarks Shop Facade | AUC@(1 deg, 10cm): 65.5 -> 64.1 | Hurt |
| use the Raposo homography-based normal estimator | normal estimation · Cambridge Landmarks | AUC@10 deg: 27.3 -> 1.5 | Hurt |
| use optimal solvers when noise is added to affine parameters | normal estimation · synthetic data | angular error | Improved |
| use optimal solver when only point coordinates are noisy | normal estimation · synthetic data | angular error | Hurt |
| run in single precision instead of double (ablation) | OPTLINEAR · synthetic data | angular error | Hurt |
Closest papers by meaning
- Adaptive Surface Normal Constraint for Geometric Estimation From Monocular Images (TPAMI 2024 · similarity 0.88)
- P1AC: Revisiting Absolute Pose From a Single Affine Correspondence (ICCV 2023 · similarity 0.88)
- Accurate Reconstruction of Oriented 3D Points using Affine Correspondences (ECCV 2020 · similarity 0.87)
- Making Affine Correspondences Work in Camera Geometry Computation (ECCV 2020 · similarity 0.87)
- SurGe: Improved Surface Geometry in Point Maps (NeurIPS 2026 · similarity 0.86)
- Relative Pose Estimation through Affine Corrections of Monocular Depth Priors (CVPR 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.
