Research map / Vision-language and multimodal / Human and embodied modeling
Large-Scale Location Recognition and the Geometric Burstiness Problem
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 |
|---|---|---|---|
| Replace raw inlier-count ranking with inter-place burstiness weighting | DisLoc · San Francisco Landmarks at 95% precision | recall@1: 53.4% -> 70.1% | Helped |
| Replace raw inlier-count ranking with inter-place burstiness weighting | DisLoc · Pittsburgh at 90% precision | recall@1: 7.2% -> 51.1% | Helped |
| Add place-popularity weighting to the inter-place count (ablation) | DisLoc · San Francisco Landmarks at 95% precision | recall@1: 71.2% -> 72.4% | Helped |
| Add place-popularity weighting to the inter-place count (ablation) | DisLoc · Pittsburgh at 90% precision | recall@1: 54.3% -> 59.4% | Helped |
| Use image-level burst counts instead of clustering matching images into places (ablation) | Geometric burstiness weighting · Pittsburgh | recall@1: 25.8% -> 51.1% at 90% precision | Hurt |
| Remove inliers that participate in geometric bursts (ablation) | Geometric burstiness weighting · San Francisco Landmarks and Pittsburgh | recall@1 | Hurt |
| Use linear rather than square-root weighting for the inter-place burst count (ablation) | Inter-place-weighted inlier count · San Francisco Landmarks and Pittsburgh | recall@1 | No clear effect |
| Use a regular grid rather than adaptive clustering to define places (ablation) | Inter-place-weighted inlier count · San Francisco Landmarks at 95% precision | recall@1: 71.2% -> 65.6% | Hurt |
| Define places with carto IDs rather than adaptive clustering (ablation) | Inter-place-weighted inlier count · San Francisco Landmarks | recall@1 | No clear effect |
| Replace adaptive clustering with a regular grid to define places (ablation) | Inter-place-weighted inlier count · Pittsburgh | recall@1 | No clear effect |
| Use the effective inlier count instead of the raw inlier count | DisLoc · San Francisco Landmarks and Pittsburgh | recall@1: 53.4% -> 57.3% at 95% precision; 7.2% -> 18% at 90% precision | Helped |
| Apply geometric burstiness weighting to verified inliers | DisLoc · Oxford105k | mAP: 0.710 -> 0.735; 0.710 -> 0.745 | Mixed |
Closest papers by meaning
- BoQ: A Place is Worth a Bag of Learnable Queries (CVPR 2024 · similarity 0.85)
- On the Estimation of Image-matching Uncertainty in Visual Place Recognition (CVPR 2024 · similarity 0.85)
- Focus on Local: Finding Reliable Discriminative Regions for Visual Place Recognition (AAAI 2025 · similarity 0.84)
- GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization (NeurIPS 2025 · similarity 0.84)
- Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era (NeurIPS 2025 · similarity 0.84)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.83)
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.