CloudSEN12 is a LARGE dataset (~1 TB) for cloud semantic understanding that consists of 49,400 image patches (IP) that are evenly spread throughout all continents except Antarctica. Each IP has an average size of 5090 x 5090 meters and contains data from Sentinel-2 optical levels 1C and 2A, Sentinel-1 Synthetic Aperture Radar (SAR), digital elevation model, surface water occurrence, land cover classes, and cloud mask results from eight cutting-edge cloud detection algorithms.
See a interactive version map here!. More information in https://cloudsen12.github.io/