We propose TransFlowNet, a novel physics-constrained deep learning framework that
focuses on the spatio-temporal super-resolution (STSR) of flow simulations. A key insight is
how to combine both statistical and physical properties in the process of an STSR network.
Therefore, we elaborately design stacked convolutional layers and Transformer blocks to
extract shallow and deep features. Besides, we employ an automatic differentiation process
for solving the physical constraints. Unlike existing physics-informed solutions, our method is …