Research map / 3D vision / Depth and stereo
Learning Icosahedral Spherical Probability Map Based on Bingham Mixture Model for Vanishing Point Estimation
Haoang Li, Kai Chen, Pyojin Kim, Kuk-Jin Yoon, Zhe Liu, Kyungdon Joo, Yun-Hui Liu
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 icosahedral spherical representation instead of equi-angular discretization (ablation) | DL-nL-auto · All datasets | F1-score: 79.41% -> 90.75% | Improved |
| add icosahedral spherical representation instead of equi-angular discretization (ablation) | DL-nL-auto · All datasets | consistency error: 2.984 pix. -> 1.644 pix. | Improved |
| express probability map on the full sphere instead of hemisphere (ablation) | DL-nL-auto · All datasets | F1-score: 88.04% -> 90.75% | Improved |
| express probability map on the full sphere instead of hemisphere (ablation) | DL-nL-auto · All datasets | consistency error: 1.813 pix. -> 1.644 pix. | Improved |
| add AS loss to MSE loss (ablation) | DL-nL-auto · All datasets | F1-score: 88.64% -> 90.02% | Improved |
| add L0 loss to MSE loss (ablation) | DL-nL-auto · All datasets | F1-score: 88.64% -> 89.37% | Improved |
| add L0 loss to MSE and AS losses (ablation) | DL-nL-auto · All datasets | F1-score: 90.02% -> 90.75% | Improved |
| add AS loss to MSE and L0 losses (ablation) | DL-nL-auto · All datasets | consistency error: 1.782 pix. -> 1.644 pix. | Improved |
| use Bingham mixture model fitting instead of non-maximum suppression for DD detection (ablation) | DL-nL-auto · 156 images with close DDs | DD detection precision/recall: 75.00% -> 100% | Improved |
| use DL-nL-auto instead of DL-L-auto | vanishing point estimation · YUD+ | F1-score: 87.34% -> 89.43% | Improved |
| use DL-nL-auto instead of traditional TR-L-auto | vanishing point estimation · All datasets | consistency error: 0.682 pix. -> 1.589 pix. | Hurt |
Closest papers by meaning
- Convex Relaxation for Robust Vanishing Point Estimation in Manhattan World (CVPR 2025 · similarity 0.86)
- Quasi-Globally Optimal and Efficient Vanishing Point Estimation in Manhattan World (ICCV 2019 · similarity 0.86)
- Quasi-Globally Optimal and Near/True Real-Time Vanishing Point Estimation in Manhattan World (TPAMI 2020 · similarity 0.85)
- Detecting Vanishing Points Using Global Image Context in a Non-Manhattan World (CVPR 2016 · similarity 0.85)
- SurGe: Improved Surface Geometry in Point Maps (NeurIPS 2026 · similarity 0.84)
- 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.
