The ExtremeEarth software architecture for Copernicus earth observation data

Hagos, D.H.; Kakantousis, T.; Vlassov, V.; Sheikholeslami, S.; Wang, T.; Dowling, J.; Fleming, A.; Cziferszky, A.; Muerth, M.; Appel, F.; Pantazi, D-A.; Bilidas, D.; Papadakis, G.; Mandilaras, G.; Stamoulis, G.; Koubarakis, M.; Troumpoukis, A.; Konstantopoulos, S.

Hagos, D.H.; Kakantousis, T.; Vlassov, V.; Sheikholeslami, S.; Wang, T.; Dowling, J.; Fleming, A. ORCID: https://orcid.org/0000-0002-0143-4527; Cziferszky, A. ORCID: https://orcid.org/0000-0002-1330-6733; Muerth, M.; Appel, F.; Pantazi, D-A.; Bilidas, D.; Papadakis, G.; Mandilaras, G.; Stamoulis, G.; Koubarakis, M.; Troumpoukis, A.; Konstantopoulos, S.. 2021 The ExtremeEarth software architecture for Copernicus earth observation data. In: Soille, P.; Loekken, S.; Albani, S., (eds.) Proceedings of the 2021 conference on Big Data from Space. Publications Office of the European Union, 4pp.

[A][B][+][-]

Abstract

Current deep learning architectures for remote sensing are trained on small datasets typically using 1 GPU without taking advantage of new innovative approaches such as distributed scale-out deep learning. In this paper, we present the ExtremeEarth software architecture for Copernicus Earth Observation data. We show how we go beyond the state-of-the-art by scaling to the petabytes of data using Hopsworks and demonstrate our big data technologies in two Thematic Exploitation Platforms (TEPs): Food Security and Polar. Furthermore, we present the integration of Hopsworks with the Polar and Food Security use cases and the flow of events for the products offered through the TEPs.

Documents