Research map / Vision-language and multimodal / Human and embodied modeling
GeoSURGE: Geo-localization using Semantic Fusion with Hierarchy of Geographic Embeddings
Angel Daruna, Nicholas Meegan, Han-Pang Chiu, Supun Samarasekera, Rakesh Kumar
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 |
|---|---|---|---|
| use 7 hierarchy levels instead of 1 (ablation) | GeoSURGE · YFCC26k | GCD accuracy 1km: 8.9 -> 17.8 | Improved |
| use 7 hierarchy levels instead of 1 (ablation) | GeoSURGE · IM2GPS | GCD accuracy 1km: 18.1 -> 27.0 | Improved |
| use 7 hierarchy levels instead of 1 (ablation) | GeoSURGE · YFCC4k | GCD accuracy 1km: 8.4 -> 19.9 | Improved |
| use 7 hierarchy levels instead of 1 (ablation) | GeoSURGE · GWS15k | GCD accuracy 1km: 0.1 -> 1.0 | Improved |
| add 3 semantic fusion blocks vs none (ablation) | GeoSURGE · YFCC26k | GCD accuracy 1km: 13.8 -> 17.8 | Improved |
| add 3 semantic fusion blocks vs none (ablation) | GeoSURGE · IM2GPS | GCD accuracy 1km: 19.8 -> 27.0 | Improved |
| add 3 semantic fusion blocks vs none (ablation) | GeoSURGE · GWS15k | GCD accuracy 200km: 23.0 -> 21.9 | Hurt |
| use geographic embeddings (retrieval) instead of classification, full hierarchy (ablation) | GeoSURGE · YFCC26k | GCD accuracy 1km: 15.1 -> 17.8 | Improved |
| use geographic embeddings (retrieval) instead of classification, full hierarchy (ablation) | GeoSURGE · IM2GPS3k | GCD accuracy 25km: 43.5 -> 42.5 | Hurt |
| use geographic embeddings instead of classification, flat representation (ablation) | GeoSURGE · GWS15k | GCD accuracy 750km: 51.8 -> 53.8 | Improved |
| use geographic embeddings instead of classification, flat representation (ablation) | GeoSURGE · YFCC26k | GCD accuracy 1km: 14.0 -> 14.2 | No clear effect |
| fine-tune last CLIP block instead of freezing all (ablation) | GeoSURGE · YFCC26k | GCD accuracy 1km: 15.7 -> 17.8 | Improved |
Closest papers by meaning
- GeoRanker: Distance-Aware Ranking for Worldwide Image Geolocalization (NeurIPS 2025 · similarity 0.90)
- Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language Models (NeurIPS 2025 · similarity 0.90)
- Vision-Language Reasoning for Geolocalization: A Reinforcement Learning Approach (AAAI 2026 · similarity 0.89)
- GeoViS: Geospatially Rewarded Visual Search for Remote Sensing Visual Grounding (CVPR 2026 · similarity 0.89)
- GRE Suite: Geo-localization Inference via Fine-Tuned Vision-Language Models and Enhanced Reasoning Chains (NeurIPS 2025 · similarity 0.89)
- Around the World in 80 Timesteps: A Generative Approach to Global Visual Geolocation (CVPR 2025 · 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.
