A Unified Benchmark and Modality-Adaptive Network for Day-and-Night Drone-View Geo-Localization

Researchers introduced IRCHN, a new real-world benchmark for drone-view geo-localization across day and night.

The dataset includes matched visible drone, infrared drone, and satellite images from 8,820 locations.

They also proposed MASTR-Net, a modality-adaptive model designed to handle illumination and cross-modality differences.

The method aims to improve localization performance across visible, infrared, and satellite views in a unified setting.