ESA Open Invitation to Tender: 1-10552
Open Date: 21/06/2021 15:24 CEST
Closing Date: 09/08/2021 13:00 CEST

This activity shall aim to bring together AI, data and Earth players to address challenging EO problems with AI. This collaborationshall result in an increased capacity to solve key scientific questions and real-world problems by proposing and developing new approaches of AI4EO for the next decade.Precisely, the overall objectives of this activity are:Harness the full power of AI and EOto investigate non-supervised learning in a broad sense, coming up with solutions able to advance the state of the art on impactful use-cases by leveraging unlabelled EO data. Indeed, most EO data is not backed up by reference datasets, and thus requires methods to leverage unlabelled data. This requires to extract information and statistics from the data only and to learn to represent the world in a non-task-specific way using techniques such as unsupervised learning, semi-supervised learning, self-supervised learning, AutoML and meta-learning.Gather and focus the attention of key-players of the AI and EO academia and industry on non-supervised learning and non-supervised AI4EO, by building and engaging a community on these topics and organising events and research sprints to benchmark non-(fully)-supervised learning approaches.A key outcome of this activity will be to give insight on the generalisation power of EO models trained by machine learning and how to improve their capacity to be applied to new images everywhere.

Estabilishment: ESRIN
ECOS Required: No
Classified: No
Price Range: 200-500 KEURO
Contracts Officer: Dora Tamas
Initiating Service: EOP-8L
IP Measure: N/A
Prog. Reference: E/A105-01 – EarthScience Block 4
Tender Type: Open Competition
Open To Tenderers From: AT+BE+CA+CH+CZ+DE+DK+EL+ES+FI+FR+GB+HU+IE+IT+LU+LV+NL+NO+PL+PT+RO+SE+SI
Technology Keywords: 26-OTHERS
Products Keywords: 2-P-Other