Abstract
In this deliverable we report on the work done so far within the EO4EU project with an emphasis on the development of generic, task agnostic, machine learning models. We set to develop and deploy within the platform task-agnostic models that will lead to significant reduction of annotation effort required in downstream supervised models and task agnostic models that will allow us to compress to a significant extend EO data.
To address these two objectives, we relied upon self-supervised models and learned compression models respectively. In this deliverable, we report on the model training and the evaluation of models that address these objectives. In addition, we also report on the EO4EU infrastructure that allows for the deployment of such models and the provision of inference as a service that the platform makes possible.