Pixel-based forest classification of Sentinel-2 images using automatically generated datasets

Contains six training datasets, composed of 800, 1600 and 3200 images. Each training dataset made up of OSM (OpenStreetMap) masks or HRL (Copernicus pan-European High Resolution Layers). Additional 2 evaluation datasets based on OSM and HRL. Composed of 200 evaluation images. For study area lithuania_2018_06.tiff is provided. This contains a fully preprocessed study area (removed clouds, composed mosaic). We provide additionally a merged mosaic of Lithuanian HRL in lithuania_HRL.tiff file. OpenStreetMap database is not provided, it can be found at https://planet.openstreetmap.org.