Hollstein_AI4QC dataset

This dataset was used for the AI4QC project (Artificial Intelligence for Quality Control), in the context of cloud detection through semantic segmentation. It consists of a database of 58 manually labeled Sentinel-2 spectra of clouds. The scenes are globally distributed and were acquired in 2016. The dataset originates from the Hollstein dataset, where the cloud masks were reclassified and converted into raster TIFF format. The new classes are the following: 

0 - clear

1 - thick cloud

2 - thin cloud

3 - cloud shadow

The "masks" folder contains the cloud masks, which have varying sizes with 60m resolution. The folder "S2_images_true_color" contains RGB images combining the bands B04, B03 and B02 of Sentinel-2. These images have a size and a resolution matching the corresponding cloud masks. The other 13 folders contain the Sentinel-2 scenes for each Sentinel-2 band. Their resolution follows the Sentinel-2 specifications:

B02, B03, B04, B08: 10mB05, B06, B07, B8A, B11, B12: 20mB01, B09, B10: 60m

The original Hollstein dataset is available at https://git.gfz-potsdam.de/EnMAP/sentinel2_manual_classification_clouds.