FAIR (findable, accessible, interoperable and re-usable) data principles are at the heart of this specification, which provides a common structure for EO Training Datasets. Innovations for fairifying data include documentation of data provenance, proposed standardised quality indicators, automation of quality indicator checking and the introduction of AIREO Compliance Levels to rapidly assess the maturity and completeness of a dataset.
The AIREO Best Practice Guidelines outline how to generate and document AIREO-compliant datasets following the AIREO specifications. The guidelines consider best practice from both the EO and AI/ML communities, as well as specific recommendations relevant to the AIREO specifications. The innovations introduced in the AIREO specification are described in more detail in the Guidelines from a data providers perspective.
Four pilot datasets are provided for users to demonstrate the AIREO innovations in practical terms. Each dataset is accompanied by a Jupyter Notebook using the AIREO Python Library functionality.
• AI4Arctic Automated Sea Ice Products dataset
• Common Agricultural Practice (CAP) Austria dataset
• Forest Observation System (FOS) dataset
• Spacenet7 Dataset
The AIREO Python library is being developed to support users in creation and application of AIREO-compliant datasets. For the initial version, basic functionality is provided allowing loading and exploring the pilot datasets as well as populating critical metadata and running automated checking.
The Jupyter notebooks allow each of the pilot datasets to be explored from the perspective of both a data creator and a data user. The functionality of the Python library is demonstrated in the notebooks.
To subscribe to the AIREO network or to contact the AIREO Team please email email@example.com