Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization

Pengyue Jia, Derong Xu, Yingyi Zhang, Xiaopeng Li, Wenlin Zhang, Yi Wen, Yuanshao Zhu, Xiangyu Zhao

Abstract

Worldwide image geolocalization aims to predict precise GPS coordinates for images captured anywhere on Earth, which is challenging due to the large visual and geographic diversity. Recent methods mainly follow two paradigms: retrieval-based approaches that match queries against a reference database, and generation-based approaches that directly predict coordinates using Large Vision-Language Models (LVLMs). However, we observe distinct error profiles between them: retrieval excels at fine-grained instance matching, while generation offers robust semantic reasoning. This complementary heterogeneity suggests that no single paradigm is universally superior. To harness this potential, we propose GeoRouter, a dynamic routing framework that adaptively assigns each query to the optimal paradigm. GeoRouter leverages an LVLM backbone to analyze visual content and provide routing decisions. To optimize GeoRouter, we introduce a distance-aware preference objective that converts the distance gap between paradigms into a continuous supervision signal, explicitly reflecting relative performance differences. Furthermore, we construct GeoRouting, the first large-scale dataset tailored for training routing policies with independent paradigm predictions. Extensive experiments on IM2GPS3k and YFCC4k demonstrate that GeoRouter significantly outperforms state-of-the-art baselines.

GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization

Abstract

Worldwide image geolocalization aims to predict precise GPS coordinates for images captured anywhere on Earth, which is challenging due to the large visual and geographic diversity. Recent methods mainly follow two paradigms: retrieval-based approaches that match queries against a reference database, and generation-based approaches that directly predict coordinates using Large Vision-Language Models (LVLMs). However, we observe distinct error profiles between them: retrieval excels at fine-grained instance matching, while generation offers robust semantic reasoning. This complementary heterogeneity suggests that no single paradigm is universally superior. To harness this potential, we propose GeoRouter, a dynamic routing framework that adaptively assigns each query to the optimal paradigm. GeoRouter leverages an LVLM backbone to analyze visual content and provide routing decisions. To optimize GeoRouter, we introduce a distance-aware preference objective that converts the distance gap between paradigms into a continuous supervision signal, explicitly reflecting relative performance differences. Furthermore, we construct GeoRouting, the first large-scale dataset tailored for training routing policies with independent paradigm predictions. Extensive experiments on IM2GPS3k and YFCC4k demonstrate that GeoRouter significantly outperforms state-of-the-art baselines.

Paper Structure

This paper contains 29 sections, 2 equations, 10 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Image Geolocalization
  4. Routing and Dynamic Model Selection
  5. Methodology
  6. Dataset Construction
  7. GeoRouter
  8. Optimization
  9. Inference
  10. Experiments
  11. Experimental Setup
  12. Main Results
  13. Ablation Study
  14. Effect of the Scaling Hyperparameter $\alpha$
  15. Effectiveness of Dynamic Routing
  16. ...and 14 more sections

Figures (10)

  • Figure 1: Paradigm Complementarity in Worldwide Image Geolocalization.
  • Figure 2: Overview of the dynamic routing framework -- GeoRouter.
  • Figure 3: Effect of the hyperparameter $\alpha$ on geolocalization accuracy. The subfigures present the results evaluated on the IM2GPS3K dataset across distance thresholds ranging from 1km to 2500km.
  • Figure 4: Effect of backbone model scale on geolocalization performance. The evaluation compares Mean Accuracy across different parameter sizes of the Qwen2-VL and Qwen3-VL families.
  • Figure 5: Data efficiency analysis. The figures illustrate the Mean Accuracy achieved when training GeoRouter with varying proportions of the available dataset.
  • ...and 5 more figures