Data Fusion with Transformers for Sentinel-2 and Sentinel-3 Missions using Non-EO and EO Training Data

To optimally exploit past and current space missions there is a growing need for the development of new tools capable of processing and analysing large amounts of heterogeneous data sets. Artificial Intelligence (AI) is a powerful tool that is becoming more common in remote sensing applications and Big Data from space. A main challenge for their exploitation is the combination of data sets from different (spectral-imaging) instruments to produce high-quality products. The focus of this work is to develop data fusion tools for the Sentinel 2 and 3 missions from the EU Copernicus program. We propose two approaches, with different datasets, to tackle this challenge, where we try to solve, without any ground truth available, combining multi-spectral imagery with a spatial resolution factor difference of 30 while reconstructing a result with almost twice the spectral definition.