人脸对齐（二十一）--A Recurrent Encoder-Decoder Network for Sequential Face Alignment