Research map / 3D vision / Depth and stereo
Hierarchical Scene Coordinate Classification and Regression for Visual Localization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 hierarchical coarse-to-fine classification and conditioning to scene-coordinate regression | Regression-only network · 7-Scenes average | accuracy (<5 cm, 5°): 78.8 -> 84.8 | Helped |
| add hierarchical coarse-to-fine classification and conditioning to scene-coordinate regression | Regression-only network · Cambridge Landmarks average | median translation error: 0.38 -> 0.16 | Helped |
| remove data augmentation (ablation) | Ours · 7-Scenes | accuracy: 75.5% vs. 84.8% | Hurt |
| remove data augmentation (ablation) | Ours · 12-Scenes | accuracy: 99.4% vs. 99.3 | No clear effect |
| remove data augmentation (ablation) | Regression-only network · 7-Scenes | accuracy: 70.9% vs. 78.8% | Hurt |
| use a much larger 409×409 receptive field for direct coordinate regression (ablation) | Regression-only network · 7-Scenes | accuracy: 59.2% vs. 78.8% | Hurt |
| use a much larger 409×409 receptive field for direct coordinate regression (ablation) | Regression-only network · i7-Scenes | accuracy: 39.3% vs. 37.9% | Helped |
| remove all conditioning layers and generators (ablation) | Hierarchical scene coordinate network · Combined scenes | localization accuracy | Hurt |
| use a larger receptive field across all hierarchy levels (ablation) | Hierarchical scene coordinate network · Combined scenes | localization accuracy | Hurt |
| vary the hierarchy depth and branching factor (ablation) | Hierarchical scene coordinate network · 7-Scenes | accuracy: 82.9% to 85.9% | Mixed |
| use a four-level classification-only network instead of classification-regression (ablation) | Aachen localization network · Aachen Day | localization accuracy at 0.25m, 2°: 71.1% vs. 47.8% | Helped |
| use top-10 retrieved images instead of only top-1 (ablation) | Aachen localization network · Aachen Day | localization accuracy at 0.25m, 2°: 58.3% -> 71.1% | Helped |
Closest papers by meaning
- R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization (CVPR 2025 · similarity 0.92)
- Learning Less Is More - 6D Camera Localization via 3D Surface Regression (CVPR 2018 · similarity 0.91)
- GLACE: Global Local Accelerated Coordinate Encoding (CVPR 2024 · similarity 0.88)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.88)
- HyperPose: Hypernetwork-Infused Camera Pose Localization and an Extended Cambridge Landmarks Dataset (CVPR 2025 · similarity 0.87)
- Learning to Produce Semi-dense Correspondences for Visual Localization (CVPR 2024 · similarity 0.86)
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.