SD4EO: AI-based synthetic satellite Sentinel-2 images of cities and building coverture (RGB+NIR bands)

This dataset has been created as part of the deliverables for ESA’s SD4EO project. It consists of synthetic versions of Sentinel-2 images in urban areas. These images were synthetically generated using schematic representations from Open Street Maps as a guide to create AI-based conditioned diffusion model images in the visible and near-infrared spectrum, along with coverage masks for non-residential buildings and the set of residential buildings combined with the former. We have generated synthetic patches of 1700x1700m which cover the urban area of the eleven cities: Paris Toulouse Poitiers Bordeaux Limoges Clermont-Ferrand Troyes Le Mans Angers Madrid Niort The patches overlap slightly (a 1/6th). There are 50 variants of visible (RGB) and near infrared (NIR) for each patch location.We have split the patches in two main folders: medium_sized_urban_structures: it contains residential and non residential urban areas which are smaller than 1 patch other_urban_and_rural_structures it contains super-structures (like airports, hippodromes, golf clubs) which are underrespresented and typically cover several neighbouring patches, and rural areas around these cities The file names within the ZIP archives follow a very simple schema, where the end of the file name indicates the content: preview_synthetic_patches_3.3m: PNG files of the patch variants for the binary masks, RGB and NIR patches at 3.3m per pixel preview_synthetic_patches_10m: PNG files of the patch variants for the floating point masks, RGB and NIR patches at 10m per pixel (native Sentinel-2 resolution) synthetic_patches_10m: netCDF files with labelled xarrays of each patch variant for the floating point masks, RGB and NIR patches at 10m per pixel (native Sentinel-2 resolution downsampling_scripts: python functions to filter and downsampling the original 3.3m patches into the 10m patches (cubic/order-3 resamping) The image and netCDF files have intuitive filenames and they are oganized in folders according the city and nature of the file. We use the following suffixes: `RGB` for images encoding visible spectrum signals `NIR` for images generated for the near-infrared band `allbuildingmask` for the coverage pixel mask of all building types in floating point `nonresidentialmask` for the coverage pixel mask of non-residential buildings in floating point The full dataset (preview and netcdf files) sums more than 247,000 files. The SD4EO Project is funded by the ESA’s FutureEO programme under contract no. 4000142334/23/I-DT and supervised by ESA Φ-lab.