Research map / 3D vision / Depth and stereo
Adapting Fine-Grained Cross-View Localization to Areas without Fine Ground Truth
Zimin Xia, Yujiao Shi, Hongdong Li, Julian F. P. Kooij
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 |
|---|---|---|---|
| adapt model with knowledge self-distillation on unlabeled target-area images | CCVPE · VIGOR cross-area test (unknown orientation) | mean error (m): 5.35 -> 4.27 | Improved |
| adapt model with knowledge self-distillation on unlabeled target-area images | GGCVT · VIGOR cross-area test (known orientation) | mean error (m): 5.19 -> 4.34 | Improved |
| adapt model with knowledge self-distillation on unlabeled target-area images | CCVPE · KITTI cross-area test | median lateral error (m): 0.98 -> 0.98 | No clear effect |
| adapt model with knowledge self-distillation on unlabeled target-area images | GGCVT · KITTI cross-area test | mean longitudinal error (m): 9.27 -> 8.56 | Improved |
| use naive best-teacher distillation without proposed designs (ablation) | GGCVT · VIGOR (GGCVT ablation) | mean error (m): 5.16 -> 5.34 | Hurt |
| add mode-based pseudo ground truth (ablation) | GGCVT student · VIGOR (GGCVT ablation) | mean error (m): 5.34 -> 4.67 | Improved |
| add outlier filtering (ablation) | GGCVT student · VIGOR (GGCVT ablation) | mean error (m): 4.67 -> 4.28 | Improved |
| use auxiliary student as pseudo GT instead of for outlier filtering (ablation) | GGCVT student · VIGOR (GGCVT ablation) | mean error (m): 4.28 -> 4.54 | Hurt |
| add entropy minimization on target-area images (ablation) | CCVPE · VIGOR validation | mean and median error | Hurt |
| fuse top-down and bottom-up multi-level pseudo ground truth (ablation) | CCVPE student · VIGOR test | mean error (m): 3.85 -> 4.49 | Hurt |
| use entropy-based outlier filtering instead of teacher-student disagreement (ablation) | CCVPE · VIGOR test | mean error (m): 3.85 -> 4.17 | Hurt |
| keep only 70-80% of samples after outlier filtering vs all (ablation) | CCVPE/GGCVT student · VIGOR | localization error | Improved |
Closest papers by meaning
- FG^2: Fine-Grained Cross-View Localization by Fine-Grained Feature Matching (CVPR 2025 · similarity 0.89)
- CVM-Net: Cross-View Matching Network for Image-Based Ground-to-Aerial Geo-Localization (CVPR 2018 · similarity 0.88)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.87)
- Scaling Image Geo-Localization to Continent Level (NeurIPS 2025 · similarity 0.87)
- Semantic Ambiguity Modeling and Propagation for Fine-Grained Visual Cross View Geo-Localization (AAAI 2025 · similarity 0.85)
- BevSplat: Resolving Height Ambiguity via Feature-Based Gaussian Primitives for Weakly-Supervised Cross-View Localization (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.
