Coarse-to-Fine Cascaded Networks with Smooth Predicting for Video Facial Expression Recognition | IEEE Conference Publication | IEEE Xplore