Proposing a novel approach, FACE-BE-SELF, for classifying adult and child facial expressions through deep domain adaptation and feature fusion.