Research map / 3D vision / Depth and stereo
Large-Scale Localization Datasets in Crowded Indoor Spaces
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 structure-based localization with global retrieval and local feature matching instead of coordinate or absolute pose regression | Visual localization on NAVER LABS indoor datasets · NAVER LABS datasets, Galaxy test images | localized images within 0.1m,1° (%) | Helped |
| Use R2D2 local features rather than SIFT | Structure-based visual localization · NAVER LABS datasets | localized images (%) | Helped |
| Increase ESAC expert count from 1 to 50 (ablation) | ESAC · Dept. 1F test, Galaxy images, 0.1m,1° | localized images (%): 0.0 -> 43.3 | Helped |
| Increase ESAC expert count from 20 to 50 (ablation) | ESAC · Dept. 4F test, Galaxy images, 0.1m,1° | localized images (%): 30.5 -> 45.2 | Helped |
| Increase ESAC expert count from 20 to 50 (ablation) | ESAC · Metro St. B2 test, Galaxy images, 0.1m,1° | localized images (%): 9.3 -> 11.2 | Helped |
| Increase ESAC expert count from 20 to 50 (ablation) | ESAC · Dept. B1 test, Galaxy images, 1m,5° | localized images (%): 7.8 -> 12.6 | Hurt |
| Increase ESAC expert count (ablation) | ESAC · NAVER LABS datasets | localized images (%) | Mixed |
| Use Basler mapping/test images from the same camera family rather than Galaxy test images | Structure-based localization methods · NAVER LABS test sets | localized images (%) | Helped |
| Evaluate only low-frequency images rather than other images | DELG+R2D2 · Dept. 4F Galaxy test, within 0.25m,2° | localized images (%): 95.2 -> 67.1 | Hurt |
| Evaluate only crowded images rather than other images | DELG+R2D2 · Metro St. B1 Galaxy test, within 0.25m,2° | localized images (%): 69.1 -> 50.1 | Hurt |
| Use a spline representation of asynchronous LiDAR SLAM trajectories as pose priors in bundle adjustment | Ground-truth SFM pipeline · NAVER LABS datasets | mean reprojection error: 50 and 200 pixels -> under 1.5 pixels | Helped |
Closest papers by meaning
- A Dataset for Benchmarking Image-Based Localization (CVPR 2017 · similarity 0.92)
- 360Loc: A Dataset and Benchmark for Omnidirectional Visual Localization with Cross-device Queries (CVPR 2024 · similarity 0.86)
- R-SCoRe: Revisiting Scene Coordinate Regression for Robust Large-Scale Visual Localization (CVPR 2025 · similarity 0.85)
- CroCoDL: Cross-device Collaborative Dataset for Localization (CVPR 2025 · similarity 0.85)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.85)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · similarity 0.85)
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.