GeoRouter: Dynamic Paradigm Routing for Worldwide Image Geolocalization

TL;DR

GeoRouter is a novel adaptive framework that dynamically routes image geolocalization queries between retrieval and generation paradigms, leveraging a large-scale dataset and a distance-aware training objective to improve accuracy.

Contribution

It introduces a dynamic routing approach for geolocalization that combines retrieval and generation paradigms, along with a new dataset and a distance-aware training method.

Findings

GeoRouter outperforms existing methods on IM2GPS3k and YFCC4k datasets.

The distance-aware preference objective improves paradigm selection accuracy.

Adaptive routing enhances geolocalization precision across diverse images.

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.