Key-features
The STORE project is dedicated to the development of AI-based image recognition systems for threat assessment and to the creation of a shared European database of annotated defence images from optronic systems. To meet these high level requirements, the project focuses on the following critical key features.
STORE will develop a new type of distributed database architecture based on a set of local database nodes with dedicated synchronization and coordination services, secured by most advanced technologies.
The resulting architecture will enable shared governance among member states by enabling application of exchange restriction rules founded on policy and releasability. Besides, the architecture will allow for strongly decoupling data acquisition, cleaning and consolidation processes from database exploitation for scientific or engineering purposes.
Shared European Database
Leading edge AI factory dedicated to image recognition for detection
STORE will enable to extend the operational use of image recognition systems by making the systems more robust to the variety of observation conditions and by addressing new forms of threats such as small drones and drone swarms, loitering munitions and hypersonic threats. Additionally, STORE potential to mature AI methods will allow an effective way to assess potential novel threats arising with the support of AI based systems.
STORE will address newest technologies for the sharing of data and models across borders, companies and institutions from the defence domain in a controlled and trustworthy manner while maintaining privacy and security.
A software architecture framework will mature decentralized learning techniques to decouple model training from a central database and therefore allowing to train AI models without exposing critical data.
Future technologies for decentralized learning techniques exploiting shared database
Evaluation of AI recognition systems integrated on demonstrators
STORE will ensure an objective evaluation of performances of AI recognition methods through a common annotation platform of acquired images, the use of shared metrics and benchmark tools in order to measure the performance of partners algorithms on the addressed use cases. The benchmark will be independent, as conducted by a dedicated team.
The evaluation will be carried out on integrated demonstrators comprised of optronics sensors used for data collection. AI technologies selected through the benchmark activities will be integrated in customized hardware near sensors and will be tested during field demonstrations implementing the STORE use cases at the end of the project.