In addition, a sketch-photo cross-reconstructed feature alignment module is proposed to obtain cross-modal aligned features for effective recognition by reconstructing and embedding global features of ...