Company-Project:
GMV
Description:
The main goal is to demonstrate how Artificial Intelligence Applications on EO can benefit from Federated Learning (FL) while keeping data privacy.

First of all, a very brief Power Point presentation with the following points will be made:
• Introducing some concepts of federated learning.
• Presentation of the federated training platform.
• Brief description of the EO use case to solve.

Secondly, a demo will be run in the laptop to show the use case chosen. This use case, from a business point of view, consist of the identification of buildings in images. From a technical point of view the objective is to demonstrate that it is possible perform a vertical partitioning [1] of the data, this means that each data owner has separate information of the same individual (pixel of the image) and they can be shared in a private and secure way.

To illustrate the use case the demo considers a scenario where data owners can publish in a private a secure way their tags (building / non-building) in the FL platform. These tag providers market their tags, but they do not want other to see them, they only allow to use them, for example to train classification models. On the other hand, data scientists can connect to the FL platform, namely to their working lab, where they can use their images and search for compatible tags within that private catalogue, and then build their classification models.