Your address will show here +12 34 56 78

Signals & Pixels

January 2022 - December 2023

Satellite images track the changing human footprint on territories, including specific changing features around major infrastructures like harbours or airports. The amount of images and their resolution is continuously increasing.
Unfortunately, the size of the teams of analysts assessing the features changes in structured reports (e.g. following the STANAG structuration) remains most of the time constant.

The emergence of Convolutional Deep Neural Networks in Artificial Intelligence (AI) is an opportunity to partly solve this issue. Automated annotations in image and automated production of structured reports have been recently proposed in the literature through static AI.

The original approach of the project "Secure Active Learning for Territorial Observations" (SALTO) is to address this issue by designing new active learning algorithms which optimize the global analyst annotation budget through an optimal selection of the areas to be annotated.
SALTO will moreover provide an entrusting mechanism for coalitions of analysts sharing the same active learning model. In practice, SALTO will provide a prototype of secure active learning implementation which will allow a pool of analysts to annotate 4 times more data than without SALTO.