Research map / Video / Video-language understanding
Trajectory-aware Cross-view Geo-Localization with Sequential Observations
Tianyi Gao, Jiayu Lin, Danielle Beaulieu, Nathan Jacobs
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 TrajMod trajectory-conditioned modulation (ablation) | TrajLoc CLIP-style baseline · SeqGeo-VL text-to-satellite | R@1%: 26.15 -> 45.48 | Improved |
| add TrajMod trajectory-conditioned modulation (ablation) | TrajLoc · SeqGeo-VL video-to-satellite | R@1: 7.27 -> 12.09 | Improved |
| add trajectory info as text prompt instead of TrajMod (ablation) | Qwen3-VL-Embedding · SeqGeo-VL text-to-satellite | R@1%: 20.91 -> 22.25 | Improved |
| add text-video contrastive loss in TrajMod training (ablation) | TrajLoc · SeqGeo-VL video-to-satellite | R@1: 10.23 -> 12.09 | Improved |
| co-train video and text without TrajMod (ablation) | TrajLoc · SeqGeo-VL video-to-satellite | R@1: 6.87 -> 7.27 | No clear effect |
| co-train video and text with TrajMod (ablation) | TrajLoc · SeqGeo-VL video-to-satellite | R@1: 9.69 -> 12.09 | Improved |
| co-train video and text with TrajMod (ablation) | TrajLoc · SeqGeo-VL text-to-satellite | R@1: 1.90 -> 2.52 | Improved |
| co-train video and text with TrajMod (ablation) | TrajLoc · SeqGeo-VL text-to-satellite | R@1%: 47.25 -> 45.48 | Hurt |
| add two-stage curriculum learning (ablation) | TrajLoc co-training · SeqGeo-VL video-to-satellite | R@1: 5.01 -> 7.27 | Improved |
| increase video frames from 1 to 6 (ablation) | TrajLoc · SeqGeo-VL video-to-satellite | R@1 | Improved |
| blend video and text query embeddings (alpha=0.4) | TrajLoc · SeqGeo-VL multimodal query | R@10: +9.49 at 1 frame to +4.75 at 7 frames | Improved |
| use sequential route descriptions instead of single-view caption (ablation) | CrossText2Loc · SeqGeo-VL text retrieval | R@1%: 10.63 -> 22.29 | Improved |
Closest papers by meaning
- Warp-free Cross-view Geo-localization via Feature-space Consensus Mining (ECCV 2026 · similarity 0.89)
- GeoBridge: A Semantic-Anchored Multi-View Foundation Model Bridging Images and Text for Geo-Localization (CVPR 2026 · similarity 0.88)
- GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization (NeurIPS 2026 · similarity 0.87)
- Semantic Ambiguity Modeling and Propagation for Fine-Grained Visual Cross View Geo-Localization (AAAI 2025 · similarity 0.87)
- From Coarse to Fine: A Matching and Alignment Framework for Unsupervised Cross-View Geo-Localization (AAAI 2025 · similarity 0.87)
- GTR: A General, Multi-View, and Dynamic Framework for Trajectory Representation Learning (ICML 2025 · 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.
