Evaluating Precise Geolocation Inference Capabilities of Vision Language Models

TL;DR

This paper assesses the ability of vision-language models to infer geographic location from images, revealing their potential privacy risks due to accurate geolocation capabilities without specialized training.

Contribution

It introduces a new global geolocation benchmark dataset and evaluates foundation models' geolocation accuracy, highlighting their unexpected proficiency in this task.

Findings

Many models achieve median errors under 300 km

Supplemental tools improve accuracy by up to 30.6%

Modern VLMs can infer location without specific training

Abstract

The prevalence of Vision-Language Models (VLMs) raises important questions about privacy in an era where visual information is increasingly available. While foundation VLMs demonstrate broad knowledge and learned capabilities, we specifically investigate their ability to infer geographic location from previously unseen image data. This paper introduces a benchmark dataset collected from Google Street View that represents its global distribution of coverage. Foundation models are evaluated on single-image geolocation inference, with many achieving median distance errors of <300 km. We further evaluate VLM "agents" with access to supplemental tools, observing up to a 30.6% decrease in distance error. Our findings establish that modern foundation VLMs can act as powerful image geolocation tools, without being specifically trained for this task. When coupled with increasing accessibility of these models, our findings have greater implications for online privacy. We discuss these risks, as well as future work in this area.