DaFab Project Begins

By Jean-Thomas Acquaviva February 7, 2024

Data economy is an ever-increasing market. Thanks to the Copernicus constellation Europe has amassed a considerable amount of Earth Observation data already used by technology-intensive SMEs and large industry in several sectors of the economy. However, the management of data at scale remains challenging, and lowering the entry cost to exploitation of Copernicus data is a key element to foster innovation and fuel this data economy. #DaFab is a new project, launched with the support of the European Space Agency: https://www.euspa.europa.eu/dafab-ai-factory-copernicus-data-scale

During the three years, a consortium, with #CERN, #ECMWF, #FORTH, #GCORE, #JozefStefanInstitute, #LuxProvide, #ThalesAleniaSpace and coordinated by #DDN will work on optimizing Copernicus European Earth Observational Data. The core idea is to ease navigation within the 50 PB of Copernicus data, by injecting a semantic structure. #DaFab aims to create this structure using a fully automated, AI-based approach. With AI algorithms, such as feature extraction, #DaFab will automatically tag and correlate sets of images. Enhancing data with dynamic structure makes superfast and super-efficient data retrieving possible, even in the huge 50 PB of Copernicus archives.

The generation of the semantic structure on Copernicus datasets requires the involvement of the most modern computing techniques, #DaFab will assess CPU, IPU, GPU, and neuromorphic processing architectures to determine the most cost-effective way to enhance earth observation data with structure.

As use cases, #DaFab will showcase the benefit of AI for the common good, with measurable improvement in flood prevention and agricultural prediction.

During these three years, based on European supercomputers, we will keep pushing the boundaries of AI, using unconventional hardware and novel data management techniques to unleash the potential of European Copernicus images. 🛰 📡

Stay tuned for updates! #DaFab #Copernicus #AI #supercomputer

Jean-Thomas Acquaviva
Datadirect Networks (DDN) France