Expression snippet transformer for robust video-based facial expression recognition

•We propose the expression snippet Transformer (EST) to achieve accurate video-based facial expression recognition (FER).•To the best of our knowledge, our approach is the first effective snippet-based Transformer method for video-based FER.•We propose an attention-augmented snippet feature extractor to enhance the modelling of intra-snippet expression movements.•We propose a shuffled snippet order prediction mechanism to enhance the modelling of inter-snippet expression movements.•Our approach achieves state-of-the-art performance on four challenging video-based FER datasets, i.e., BU-3DFE, MMI, AFEW, and DFEW.