Research map / 3D vision / Depth and stereo
Rectified Embedding Flow Learning for Aerial Multi-view Geo-localization
Hao Ruan, Jinliang Lin, Yingxin Lai, Zhiming Luo, Shaozi Li, Yu Zang, Cheng Wang
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 REFL (flow learning + embedding rectification) | Qwen3-VL-Embedding-2B · Aerial MVGL (10 sub-tasks avg) | R@1 (%): 40.77 -> 44.83 | Improved |
| add REFL | Qwen3-VL-Embedding-2B · Aerial MVGL (10 sub-tasks avg) | R@10 (%): 62.29 -> 65.85 | Improved |
| add Stage 1 velocity-prior flow learning (100k steps) (ablation) | REFL · Aerial MVGL | mR@1 (%): 42.95 -> 44.41 | Improved |
| extend Stage 1 training from 100k to 400k steps (ablation) | REFL · Aerial MVGL | mR@1 (%): 44.41 -> 44.83 | Improved |
| use 6-layer instead of 1-layer velocity MLP (ablation) | REFL · Aerial MVGL | mR@1 (%): 43.89 -> 44.83 | Improved |
| use 6-layer instead of 4-layer velocity MLP (ablation) | REFL · Aerial MVGL | mR@1 (%): 43.45 -> 44.83 | Mixed |
| replace static contrastive learning with trajectory-guided rectification (ER-CL) (ablation) | REFL · Aerial MVGL | mR@1 (%): 40.77 -> 42.84 | Improved |
| add Stage 1 prior and joint L_FL to ER-CL (ablation) | REFL · Aerial MVGL | mR@1 (%): 42.84 -> 44.83 | Improved |
| add Stage 1 flow learning without contrastive supervision (ablation) | Backbone (zero-shot) · Aerial MVGL | mR@1 (%): 19.91 -> 16.15 | Hurt |
| add Stage 1 prior before static contrastive learning (ablation) | Static contrastive baseline · Aerial MVGL | mR@1 (%): 40.77 -> 37.31 | Hurt |
| set joint loss weight lambda to 0.5 instead of 0 (ablation) | REFL · Aerial MVGL | mR@1 (%): 43.06 -> 44.83 | Improved |
| set joint loss weight lambda to 0.5 instead of 1.0 (ablation) | REFL · Aerial MVGL | mR@1 (%): 43.82 -> 44.83 | Improved |
Closest papers by meaning
- Spatial-Aware Feature Aggregation for Image based Cross-View Geo-Localization (NeurIPS 2019 · similarity 0.88)
- Semantic Ambiguity Modeling and Propagation for Fine-Grained Visual Cross View Geo-Localization (AAAI 2025 · similarity 0.88)
- Cross-view Geo-localization with Layer-to-Layer Transformer (NeurIPS 2021 · similarity 0.88)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.87)
- Multi-Modal Aerial-Ground Cross-View Place Recognition with Neural ODEs (CVPR 2025 · similarity 0.86)
- Aligning Geometric Spatial Layout in Cross-View Geo-Localization via Feature Recombination (AAAI 2024 · similarity 0.86)
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.
