Training data for the GitHub repository "buildingsFromSentinel"

Training and testing data for machine learning models predicting the building height and footprint from satellite data in urban areas.

Sentinel-1 and -2 data are retrieved from https://scihub.copernicus.eu/ and the GHS built-up grid (here GHSBuilt10) from https://ghsl.jrc.ec.europa.eu/download.php?ds=buS2. GHSBuilt10 is derived from Sentinel-2 global image composite for the reference year 2018 using Convolutional Neural Networks (GHS-S2Net).

The dataset contains the following folders:

footprint: PNG images over urban areas with either three or four features:

XXX_labels.png: true-colour images (TCI) retrieved from Sentinel-2 data

XXX_labels4.png: TCIs with the band 8 (i.e., near-infrared = NIR) as the fourth dimension in the image.

height: data for different cities

building_height.tif: real building height (only for the training data)

sentinel_cropped: satellite images for the same area. Contains Sentinel-1 and -2 data as well as the GHS-Built data with a 10-m resolution.

README.txt: information of the origin of the building height data