PixBox_AI4QC dataset

The dataset was used in the AI4QC project (Atrificial Intelligence for Quality Control), in the context of cloud detection. It originates from the PixBox-S2-CMIX dataset, which was used as a validation reference within the first Cloud Masking Inter-comparison eXercise (CMIX) in 2019. The PixBox-S2-CMIX pixel collection was conducted already in 2018. The PixBox_AI4QC dataset was derived from this PixBox dataset, where the cloud masks were reclassified and rasterized to TIFF files. The new classes are the following:

0 - clear

1 - thick cloud

2 - thin cloud

3 - cloud shadow

5 - no-data

The dataset contains 29 scenes, which are spatially, temporally (year 2018), and thematically well distributed.

The "raster_masks" folder contains the cloud masks, which have a size of 10980x10980 pixels with 10m resolution. 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: 10m

B05, B06, B07, B8A, B11, B12: 20m

B01, B09, B10: 60m

The original PixBox dataset is available at https://zenodo.org/records/5036991.