The SimTAI project, proposed by OKTAL-SE within the Franco-British CWITP programme, has just been notified by DGA, DSTL and MBDA.
OKTAL-SE is the team leader, and will carry out this study in partnership with the startup Numalis, the University of Newcastle and MBDA.
The aim of the project is to design an approach which should allow system designers and AI engineers to incrementally build the correct synthetic EO/RF images datasets and to ensure that: the AI system can be effectively used both in simulated and real environments, for the expected reasons, and for the chosen domain of use.
This study will be based on innovative methods of explainability and robustness analysis of CNNs, as well as on the analysis of correlations between the final performance of the CNN and the input parameters of the simulation tools.
Thanks to this project, OKTAL-SE aims to take a significant step forward in the field of simulation exploitation for AI, and to strengthen its network of industrial and academic partners in the AI domain.