The objective of this activity is to develop the most accurate free and open cloud mask for Sentinel-2 with global coverage and to make it available for global Earth Observation community and promote its usage. For this it is required to separate cloud free areas from cloud- and cloud shadow-corrupted areas as accurately as possible in the automatic processing chains for higher level products derivation.
Using an AI model all the pixels are divided into four classes following the CMIX standard:
- Cloud free;
- Semi-transparent cloud;
- Cloud shadow.
Output cloud masks are with 10 m spatial resolution and Sentinel-2 data is ready to use for higher level products derivation. Cloud and cloud-shadow corrupted areas are reliably separated. The output of this project targets all users who need Sentinel-2 data, including private companies, academia, and governmental users.
Currently global coverage support is added by increasing the training set using existing free and open labelled Sentinel-2 imagery data sets and labelling additional products by KappaZeta team.