Research map / 3D vision / 3D generation
Beyond 2D Matching: A Unified Single-Stage Framework for Geometry-Aware Cross-View Object Geo-Localization
Liyao Wang, Ruipu Wu, Haojun Xu, Lei Shi, Linjiang Huang, Si Liu
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 |
|---|---|---|---|
| add deep supervision loss (ablation) | GAGeo base · CMA-Loc D->S seen | mAcc: 62.86 -> 66.38 | Improved |
| add deep supervision loss (ablation) | GAGeo base · CMA-Loc G->S seen | mAcc: 67.41 -> 70.06 | Improved |
| add contrastive loss (ablation) | GAGeo + deep supervision · CMA-Loc D->S seen | mAcc: 66.38 -> 67.63 | Improved |
| add contrastive loss (ablation) | GAGeo + deep supervision · CMA-Loc G->S seen | mAcc: 70.06 -> 71.33 | Improved |
| add camera pose loss (ablation) | GAGeo + contrastive · CMA-Loc D->S seen | mAcc: 67.63 -> 68.48 | Improved |
| add camera pose loss (ablation) | GAGeo + contrastive · CMA-Loc G->S seen | mAcc: 71.33 -> 71.50 | Improved |
| add camera pose loss (ablation) | GAGeo + contrastive · CMA-Loc D->S seen | mIoU: 79.51 -> 79.39 | Hurt |
| increase learnable tokens from 1 to 2 or 4 (ablation) | GAGeo · CMA-Loc G->S seen | accuracy | Hurt |
| use pi3 backbone instead of VGGT or WorldMirror (ablation) | GAGeo · CMA-Loc zero-shot G->D | accuracy | Improved |
| use mask prompt instead of point prompt | GAGeo · CMA-Loc D->S seen | mAcc: 68.48 -> 76.00 | Improved |
| add SAM Prompt Stage | TROGeo · CMA-Loc D->S seen | mIoU: 46.86 -> 69.09 | Improved |
| use GAGeo instead of OCGNet in zero-shot transfer | OCGNet · CVOGL-Seg D->S zero-shot | mAcc: 12.75 -> 29.43 | Improved |
Closest papers by meaning
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.90)
- Weather-Robust Cross-View Geo-Localization via Prototype-Based Semantic Part Discovery (NeurIPS 2026 · similarity 0.90)
- Seeing Across Skies and Streets: Feedforward 3D Reconstruction from Satellite, Drone, and Ground Images (NeurIPS 2026 · similarity 0.89)
- Game4Loc: A UAV Geo-Localization Benchmark from Game Data (AAAI 2025 · similarity 0.89)
- View From Above: Orthogonal-View aware Cross-view Localization (CVPR 2024 · similarity 0.89)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.89)
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.
