Research map / 3D vision / Depth and stereo
Ground-to-Aerial Image Geo-Localization With a Hard Exemplar Reweighting Triplet Loss
Sudong Cai, Yulan Guo, Salman Khan, Jiwei Hu, Gongjian Wen
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 |
|---|---|---|---|
| use HER triplet loss instead of exhaustive soft-margin triplet loss (ablation) | Siam-FCANet18 · CVUSA | Recall@Top 1%: 95.1% -> 98.3% | Improved |
| use HER triplet loss instead of exhaustive soft-margin triplet loss (ablation) | Siam-FCANet34 · CVUSA | Recall@Top 1%: 96.5% -> 98.3% | Improved |
| use HER triplet loss instead of exhaustive soft-margin triplet loss (ablation) | Siam-FCANet18 · VH Seattle | Recall@Top 1%: 66.2% -> 68.1% | Improved |
| use HER triplet loss instead of exhaustive soft-margin triplet loss (ablation) | Vo network · CVUSA | Recall@Top 1%: 83.9% -> 86.7% | Improved |
| use HER triplet loss instead of exhaustive soft-margin triplet loss (ablation) | Vo network · VH Denver | Recall@Top 1%: 62.4% -> 65.9% | Improved |
| add FCAM dual attention to ResNet34 blocks (ablation) | Siam-PNet34 · CVUSA | Recall@Top 1%: 98.0% -> 98.3% | Improved |
| add FCAM dual attention to ResNet34 blocks (ablation) | Siam-PNet34 · VH Denver | Recall@Top 1%: 77.0% -> 78.3% | Improved |
| add FCAM dual attention to ResNet18 blocks (ablation) | Siam-PNet18 · VH Seattle | Recall@Top 1%: 68.6% -> 68.1% | Hurt |
| add FCAM dual attention to ResNet18 blocks (ablation) | Siam-PNet18 · CVUSA | Recall@Top 1%: 97.7% -> 98.3% | Improved |
| use FC layer with orientation regression instead of NetVLAD aggregation (ablation) | Siam-FCANet18 · CVUSA | Recall@Top 1%: 93.9% -> 98.3% | Improved |
| use FC layer with orientation regression instead of NetVLAD aggregation (ablation) | Siam-FCANet34 · VH Denver | Recall@Top 1%: 67.0% -> 78.3% | Improved |
| average features over 16 rotated test samples (ablation) | Siam-FCANet34 · VH Seattle | Recall@Top 1%: 71.1% -> 73.3% | Improved |
Closest papers by meaning
- CVM-Net: Cross-View Matching Network for Image-Based Ground-to-Aerial Geo-Localization (CVPR 2018 · similarity 0.88)
- Game4Loc: A UAV Geo-Localization Benchmark from Game Data (AAAI 2025 · similarity 0.86)
- Scaling Image Geo-Localization to Continent Level (NeurIPS 2025 · similarity 0.86)
- BevSplat: Resolving Height Ambiguity via Feature-Based Gaussian Primitives for Weakly-Supervised Cross-View Localization (NeurIPS 2025 · similarity 0.85)
- View From Above: Orthogonal-View aware Cross-view Localization (CVPR 2024 · similarity 0.85)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 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.
