MTMSN: Multi-Task and Multi-Modal Sequence Network for Facial Action Unit and Expression Recognition | IEEE Conference Publication | IEEE Xplore