Shared daTabase for Optronics image Recognition and Evaluation
The defence sector consistently faces the responsibility of swift evaluation and response to various scenarios. In addition, there is a constant flow of new threats by weapons that are more and more difficult to discover. Artificial intelligence (AI)-based image recognition system could be an effective tool to support faster detection and decision-making. However, training such a model requires representative images, which are often not publicly available, nor does the data held by different entities have a unified form that could be easily used for training.
STORE project aims to develop AI-based image recognition systems while also constructing a shared database of annotated defence images that can be used to train such systems. STORE project will explore different AI and machine learning models and goals to evaluate their suitability for this setting. STORE also develops methods to leverage learning with classified data.
Joint efforts in the STORE project unite partners with diverse expertise areas, spanning from optronics and databases management to privacy-preserving computation and AI. Cybernetica will deploy its secure computation tools and knowledge, especially Sharemind. Primarily, we will focus on our tools for privacy-preserving federated machine learning. In addition, prior to implementation, we will model and analyse the use-cases of the project using privacy-enhanced BPMN (PE-BPMN) to map the processes and understand the privacy requirements. This work supports the project, allowing decentralised learning techniques where the parties can input their private data and the final models are shared between the participants without exposing critical data.
STORE is a European Defence Fund project, which is funded by European Commission.