Multisensor Geospatial Foundation Model: Bridging Remote Sensors for Enhanced Geospatial Analysis

A novel multisensor geospatial pretraining model, msGFM, effectively unifies data from four key sensor modalities (RGB, Sentinel-2, SAR, DSM) to enable joint representation learning and enhance performance across a range of downstream geospatial tasks.