Data related to Space4iCity WP5: Characterization of Vegetation
===============================================================

Instead of simply computing NDVI, a UNet model was trained to identify
different land cover classes at 10m resolution based on Sentinel-2
bands 2, 3, 4 and 8. This has the advantage of being able to
distinguish taller vegetation (trees) from lower vegetation (grass).

Target classes are based on the ESA WorldCover scheme:
https://esa-worldcover.org/en

Details on the training procedure and evaluation results are found in
the attached report.