Research map / Architectures and efficiency / Attention and transformers
Cross-view Geo-localization with Layer-to-Layer Transformer
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 self-cross attention (attend across adjacent layers) in place of self-attention (ablation) | Polar-L2LTR · CVUSA test | r@1 (%): 93.26 -> 94.05 | Helped |
| add self-cross attention in place of self-attention (ablation) | Polar-L2LTR · CVUSA test | r@1% (%): 99.68 -> 99.67 | No clear effect |
| add self-cross attention under 20% training data (ablation) | Polar-L2LTR · CVUSA few-shot 20% | r@1 (%): 75.37 -> 76.01 | Helped |
| add self-cross attention under 60% training data (ablation) | Polar-L2LTR · CVUSA few-shot 60% | r@1 (%): 88.74 -> 90.30 | Helped |
| add learnable positional embeddings (ablation) | L2LTR · CVUSA test | r@1 (%): 89.04 -> 91.99 | Helped |
| add learnable positional embeddings (ablation) | Polar-L2LTR · CVUSA test | r@1 (%): 90.90 -> 94.05 | Helped |
| add polar transform to aerial images (ablation) | L2LTR · CVUSA test | r@1 (%): 91.99 -> 94.05 | Helped |
| Transformer-based L2LTR vs CNN-based SAFA without polar transform | L2LTR · CVACT_val | r@1 (%): 78.28 -> 83.14 | Helped |
| L2LTR vs Shi et al. under limited field of view | L2LTR · CVUSA 70 deg FoV | r@1 (%): 8.78 -> 13.95 | Helped |
| add polar transform to aerial images on fine-grained test (text and table disagree on L2LTR value) | L2LTR · CVACT_test | r@1 (%): 58.33 -> 60.72 | Helped |
Closest papers by meaning
- Cross-View Geo-Localization via Learning Disentangled Geometric Layout Correspondence (AAAI 2023 · similarity 0.92)
- Optimal Feature Transport for Cross-View Image Geo-Localization (AAAI 2020 · similarity 0.91)
- Lending Orientation to Neural Networks for Cross-View Geo-Localization (CVPR 2019 · similarity 0.91)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.91)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.90)
- Semantic Ambiguity Modeling and Propagation for Fine-Grained Visual Cross View Geo-Localization (AAAI 2025 · similarity 0.87)
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.