Is it right to transform 3d face recognition data to 2d and perform 2d recognition in 3d face recognition research?