Research map / 3D vision / Depth and stereo
Vanishing Point Estimation in Uncalibrated Images with Prior Gravity Direction
Rémi Pautrat, Shaohui Liu, Petr Hruby, Marc Pollefeys, 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 |
|---|---|---|---|
| add proposed non-minimal-solver local optimization (LO) to RANSAC (ablation) | Hybrid RANSAC with prior gravity · YorkUrban DB | median rotation error (deg): 4.59 -> 1.13 | Improved |
| use proposed LO instead of Iter LO (ablation) | Hybrid RANSAC with prior gravity · YorkUrban DB | median rotation error (deg): 1.51 -> 1.13 | Improved |
| add proposed LO to 2-0-0g with prior gravity (ablation) | 2-0-0g solver · YorkUrban DB | median rotation error (deg): 25.09 -> 1.25 | Improved |
| add proposed LO (longer runtime) (ablation) | RANSAC solvers · YorkUrban DB | runtime (ms): 43 -> 100 | Hurt |
| give solvers accurate IMU gravity | new 2-line solvers · YorkUrban DB | median rotation error (deg): 1.51 -> 1.10 | Improved |
| use only a rough upright prior on gravity instead of no gravity (hybrid) | Hybrid RANSAC · ScanNet | rotation AUC at 5 deg: 7.0 -> 7.9 | Improved |
| use noisy prior gravity (24.8 deg error) with single-config gravity solvers instead of no-gravity solvers | 1-1-0g, 0-1-1g · ScanNet | median rotation error (deg): 42.07 -> 43.22 | No clear effect |
| add gravity (IMU) to relative-rotation estimation from vanishing points | VP solvers on LaMAR · LaMAR | median rotation error (deg): 10.06 -> 2.89 | Improved |
| use 1-1-0g solver instead of 2-0-0g solver | gravity-based solvers · YorkUrban DB prior gravity | median rotation error (deg): 1.25 -> 1.20 | Improved |
| use proposed LO instead of Iter for focal length estimation (ablation) | 1-1-0g on ScanNet · ScanNet | relative focal length error: 0.123 -> 0.129 | No clear effect |
Closest papers by meaning
- Quasi-Globally Optimal and Efficient Vanishing Point Estimation in Manhattan World (ICCV 2019 · similarity 0.90)
- Quasi-Globally Optimal and Near/True Real-Time Vanishing Point Estimation in Manhattan World (TPAMI 2020 · similarity 0.89)
- Convex Relaxation for Robust Vanishing Point Estimation in Manhattan World (CVPR 2025 · similarity 0.87)
- Analytical Modeling of Vanishing Points and Curves in Catadioptric Cameras (CVPR 2018 · similarity 0.87)
- City-Scale Localization for Cameras with Known Vertical Direction (TPAMI 2016 · similarity 0.87)
- Globally Optimal Relative Pose Estimation With Gravity Prior (CVPR 2021 · similarity 0.87)
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.
