Research map / Label-efficient and robust learning / Recognition and long-tail
Must All Negatives Be Pushed Away Equally? Uncertainty-Aware Cross-View Geo-Localization via Normal Inverse Gamma Distribution
Songsong Ouyang, Le Wu, 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 |
|---|---|---|---|
| replace uniform negative labels with uncertainty-aware Soft InfoNCE (zero-shot transfer) | UA-Geo · CVUSA -> CVACT test | R@1: 35.52 -> 54.99 (vs CLGT) | Improved |
| add attention-weight statistics (entropy, max, top-2 ratio, variance) to the Uncertainty Head (ablation) | UA-Geo · VIGOR same-area | R@1: 78.92 -> 79.48 | Improved |
| swap ConvNeXt-B for DINOv3-ViT-B encoder in the Sample4Geo framework (ablation) | Sample4Geo · VIGOR same-area | R@1: 77.86 -> 77.68 | No clear effect |
| set Top-K to 5 instead of 3 (ablation) | UA-Geo · VIGOR same-area | R@1: 78.68 -> 79.48 | Improved |
| set Top-K to 8 instead of 5 (ablation) | UA-Geo · VIGOR same-area | R@1: 79.48 -> 79.09 | Hurt |
| set soft label scale w to 0.1 instead of 0.05 (ablation) | UA-Geo · VIGOR same-area | R@1: 78.38 -> 79.48 | Improved |
| set soft label scale w to 0.3 instead of 0.1 (ablation) | UA-Geo · VIGOR same-area | R@1: 79.48 -> 78.69 | Hurt |
| UA-Geo vs Sample4Geo in-domain on CVUSA | UA-Geo · CVUSA | R@1: 98.68 -> 99.30 | Improved |
| UA-Geo vs Sample4Geo on VIGOR cross-area | UA-Geo · VIGOR cross-area | R@1: 61.70 -> 66.70 | Improved |
| UA-Geo vs CLGT on CVACT test in-domain | UA-Geo · CVACT test | R@1: 73.22 -> 73.63 | Improved |
Closest papers by meaning
- Semantic Ambiguity Modeling and Propagation for Fine-Grained Visual Cross View Geo-Localization (AAAI 2025 · similarity 0.89)
- Sample4Geo: Hard Negative Sampling For Cross-View Geo-Localisation (ICCV 2023 · similarity 0.88)
- Unleashing Unlabeled Data: A Paradigm for Cross-View Geo-Localization (CVPR 2024 · similarity 0.87)
- CVGL: Causal Learning and Geometric Topology (NeurIPS 2025 · similarity 0.86)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.86)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · 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.
