The FIGARO project (Fission gas release Analysis and Radiation effects in Cr-doped oxide fuel) proposes to study the evolution of an ATF fuel (Cr-doped UO2) microstructure and the fission products behaviour under irradiation, by coupling experimental and simulation data. We wish to develop an innovative TEM image processing based on machine learning, capable to create a robust and scalable method for automated analysis of irradiation defects in nuclear materials.