Momentum contrast transformer for COVID-19 diagnosis with knowledge distillation

•A CT image diagnostic method based on momentum contrast Transformer is proposed.•This method classifies COVID-19 from the perspective of global and local features.•Momentum contrast Transformer enables ViT to be trained well on small datasets.•Knowledge distillation is used to integrate locality of convolution into ViT.•We experiment with multiple metrics on multiple COVID-19 datasets.