Evaluating Precise Geolocation Inference Capabilities of Vision Language Models
This work highlights privacy risks for individuals due to VLMs' capability to geolocate images without specific training, though it is incremental in evaluating existing models on a new task.
The paper investigated the ability of Vision-Language Models (VLMs) to infer geographic locations from unseen images, finding that many models achieved median distance errors under 300 km, and using supplemental tools reduced errors by up to 30.6%.
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.