GenAI for cross-view geo-localization

Abstract:

Generative vision models have advanced various research domains, especially in cross-view geo-localization. These generative models have improved the geo-localization accuracy and enabled novel approaches like generating one view conditioned by another—such as ground-to-aerial and aerial-to-ground. This talk will delve into these generative models, exploring their integration with other data modalities like GIS maps to enhance cross-view geo-localization, highlighting their transformative role in advancing geospatial analysis and pushing the boundaries of future cross-view geo-localization research.