This publication proposes a method of using neural networks to generate liquid simulations on the fly. Space-time datasets are used to learn deformation networks which can reconstruct a new space-time dataset that the network is fed. We demonstrate the effectiveness of the method, by doing this on a smartphone in real-time.
Recommended citation: Prantl, Lukas; Bonev, Boris; Thuerey Nils. (2019). “Generating Liquid Simulations with Deformation-aware Neural Networks.” International Conference on Learning Representations 2019.