MultiQG-TI: Towards Question Generation from Multi-modal Sources