Research map / 3D vision / Depth and stereo
Learning To Detect Scene Landmarks for Camera 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 |
|---|---|---|---|
| Use sparse landmark detections with NBE bearing predictions | camera localization · INDOOR-6 scenes 1–6 | recall (5cm, 5°): 18.7 -> 38.4; 12.3 -> 37.0; 19.7 -> 53.0; 44.9 -> 62.5; 10.6 -> 40.0; 44.3 -> 50.5 | Helped |
| Combine SLD correspondences with HLoc matches | HLoc · INDOOR-6 scenes 1–6 | recall (5cm, 5°): 64.8 -> 68.7; 60.6 -> 62.7; 81.0 -> 81.0; 70.6 -> 73.9; 42.7 -> 45.3; 79.9 -> 82.0 | Helped |
| Use NBE+SLD instead of DSAC* | camera localization · 7-SCENES | average recall (5cm, 5°): 97.8 -> 95.7 | Hurt |
| Train SLD on patches rather than whole images (ablation) | SLD · INDOOR-6 | average recall (5cm, 5°): 18.2 -> 28.2 | Helped |
| Use half-resolution rather than quarter-resolution heatmaps (ablation) | SLD · INDOOR-6 | average recall (5cm, 5°): 28.2 -> 36.9 | Helped |
| Add data augmentation (ablation) | SLD · INDOOR-6 | average recall (5cm, 5°): 36.9 -> 38.4 | Helped |
| Increase the landmark count from 200 to 300 (ablation) | SLD · INDOOR-6 | average recall (5cm, 5°): 38.4 -> 42.7 | Helped |
| Increase the landmark count from 300 to 400 (ablation) | SLD · INDOOR-6 | average recall (5cm, 5°): 42.7 -> 42.4 | Hurt |
| Add visibility-weighted angular loss (ablation) | NBE · INDOOR-6 | average recall (10cm, 10°): 7.11 -> 7.67 | Helped |
| Use 100 rather than 50 selected landmarks (ablation) | NBE · INDOOR-6 | average recall (10cm, 10°): 7.67 -> 8.36 | Helped |
| Choose landmarks randomly rather than selecting them by saliency (ablation) | NBE without visibility weighting · INDOOR-6 | average recall (10cm, 10°): 7.11 -> 7.32 | No clear effect |
Closest papers by meaning
- Learning Less Is More - 6D Camera Localization via 3D Surface Regression (CVPR 2018 · similarity 0.91)
- Back to the Feature: Learning Robust Camera Localization From Pixels To Pose (CVPR 2021 · similarity 0.91)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.89)
- HyperPose: Hypernetwork-Infused Camera Pose Localization and an Extended Cambridge Landmarks Dataset (CVPR 2025 · similarity 0.87)
- LOCATE 3D: Real-World Object Localization via Self-Supervised Learning in 3D (ICML 2025 · similarity 0.86)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · 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.