Research map / Point clouds / Point cloud classification
A-Contrario Horizon-First Vanishing Point Detection Using Second-Order Grouping Laws
Gilles Simon, Antoine Fond, Marie-Odile Berger
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 GMM-based candidate horizon line sampling (ablation) | horizon-first VP detector · York Urban (YU) | horizon AUC: 95.24 -> 95.35 | No clear effect |
| add GMM-based candidate horizon line sampling (ablation) | horizon-first VP detector · Eurasian Cities (EC) | horizon AUC: 88.42 -> 91.10 | Improved |
| add GMM-based candidate horizon line sampling (ablation) | horizon-first VP detector · Horizon Lines in the Wild (HLW) | horizon AUC: 45.59 -> 54.43 | Improved |
| use S=300 samples instead of a single sample (ablation) | horizon-first VP detector · Horizon Lines in the Wild (HLW) | horizon AUC: 45.88 -> 54.43 | Improved |
| increase number of line samples S from 150 to 600 (ablation) | horizon-first VP detector · York Urban and Eurasian Cities (first 20 images) | horizon AUC: 93.7% to 94.3% | Improved |
| vary consistency threshold theta_con (ablation) | horizon-first VP detector · York Urban and Eurasian Cities (first 20 images) | horizon AUC: -28.7% | Hurt |
| vary number of histogram bins Lh (ablation) | horizon-first VP detector · York Urban and Eurasian Cities (first 20 images) | horizon AUC: -12.4% | Hurt |
| use consistency measure only for refinement, not VP detection (a-contrario VP detection) | horizon-first VP detector vs Zhai et al. · York Urban and Eurasian Cities | spurious VPs per good VP: one for 2 good -> one for 23 | Improved |
| a-contrario VP detection along horizon line instead of random-subset VP initialization | horizon-first VP detector vs Zhai et al. · York Urban and Eurasian Cities | good VPs per image: 2.11 -> 2.14 | No clear effect |
| use our scoring and PDF instead of Zhai et al. on a large CNN-trained dataset | horizon-first VP detector vs Zhai et al. · Horizon Lines in the Wild (HLW) | horizon AUC: 58.24 -> 54.43 | Hurt |
| use S=1 single sample instead of 300 vs Zhai et al. (ablation) | horizon-first VP detector · Horizon Lines in the Wild (HLW) | horizon AUC: 45.41 -> 45.88 | No clear effect |
| use Ours instead of Zhai et al. with S=1 | horizon-first VP detector · York Urban (YU) | computation time: 0.58 -> 0.29 | Improved |
Closest papers by meaning
- Detecting Vanishing Points Using Global Image Context in a Non-Manhattan World (CVPR 2016 · similarity 0.89)
- Transformer Based Line Segment Classifier With Image Context for Real-Time Vanishing Point Detection in Manhattan World (CVPR 2022 · similarity 0.85)
- VaPiD: A Rapid Vanishing Point Detector via Learned Optimizers (ICCV 2021 · similarity 0.85)
- Convex Relaxation for Robust Vanishing Point Estimation in Manhattan World (CVPR 2025 · similarity 0.85)
- Deep Vanishing Point Detection: Geometric Priors Make Dataset Variations Vanish (CVPR 2022 · similarity 0.85)
- VPDETR: End-to-End Vanishing Point DEtection TRansformers (AAAI 2024 · similarity 0.84)
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.
