Research map / 3D vision / Depth and stereo
Benchmarking the Robustness of Cross-view Geo-localization Models
Qingwang Zhang, Yingying Zhu
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 |
|---|---|---|---|
| train on stylized images (alone or mixed with standard images) instead of only clean images (ablation) | SAFA, L2LTR, TransGeo · CVUSA-C | R@1cor | Improved |
| train on stylized images only (no standard images) (ablation) | SAFA, L2LTR, TransGeo · CVUSA clean images (severity 0) | R@1 | Hurt |
| interleave stylized and standard images with equal probability during training (ablation) | SAFA, L2LTR, TransGeo · CVUSA-C | R@1cor and clean R@1 | Improved |
| train on CLAHE images only (ablation) | SAFA, L2LTR, TransGeo · CVUSA-C | R@1cor | No clear effect |
| mix CLAHE and standard images in equal proportion during training (ablation) | SAFA, L2LTR, TransGeo · CVUSA-C | R@1cor | Improved |
| higher clean accuracy of a model | cross-view geo-localization models · CVUSA-C | R@1cor: L2LTR R@1cor 82.88 vs GeoDTR 75.48 despite lower clean R@1 (94.05 vs 95.43) | Mixed |
| snow, spatter and zoom blur corruptions versus other corruption types | all benchmarked models · CVUSA-C | RCE: RCE exceeding 28% for all models | Hurt |
| train on CVACT (urban) instead of CVUSA (suburban) | benchmarked models · CVACT_val-C vs CVUSA-C | clean-to-corrupted R@1 drop: drop 4.1% less on CVACT_val-C | Improved |
Closest papers by meaning
- MRGeo: Robust Cross-View Geo-Localization of Corrupted Images via Spatial and Channel Feature Enhancement (AAAI 2026 · similarity 0.90)
- SinGeo: Unlock Single Model's Potential for Robust Cross-View Geo-Localization (CVPR 2026 · similarity 0.89)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.87)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.86)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 2026 · similarity 0.86)
- CVGL: Causal Learning and Geometric Topology (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.
