Introducing DEFAIR: Destination Earth Framework for Preparing AI-Ready Data

The Python library transforms EO and meteorological satellite data into usable, AI-ready data.

Implemented by EUMETSAT and targeted at the Destination Earth Data Lake (DEDL) data portfolio, the DEFAIR (Destination Earth Framework for Preparing AI-Ready Data) service bridges the gap between supported Earth observation satellite datasets and their transformation into an AI/ML-ready form.

Built on xarray and Dask, DEFAIR provides a unified API for reading data from multiple satellite instruments, pre-processing it and reprojecting it into standard coordinate reference systems and grids, including geographic coordinates and HEALPix. Users can export the results in widely used formats such as NetCDF, GeoTIFF and Zarr.

By handling the complexity of different data formats and instrument-specific projections, DEFAIR helps researchers and data scientists focus on AI model development rather than data engineering.

Users can install DEFAIR locally or use it within the DEDL JupyterHub environment. Please follow the instructions in the service documentation for installation, tutorials and access requirements.

 

Explore DEFAIR.

Last updates