Figure 5
From: A deep-learning approach for reconstructing 3D turbulent flows from 2D observation data
Architecture of 2D3DGAN and 3DCAE. (a) Schematic of 2D3DGAN showing the generator network (left) and the discriminator network (right). The generator network receives a dataset represented by two planes in the flow, the first one is the observation plane, and the second one is the plane that is matched with the observation plane by applying the procedure explained in Eq. (2), which is indicated by the superscript ‘\(*\)’. The output from the generator network, i.e. the artificial 3D flow data is fed to the discriminator network, and the latter tries to distinguish if the data is artificial or true. In the generator network, \(\gamma\) represents the residual scaling parameter, which is set to 0.2. More information can be found in29,43. (b) Schematic of the feature extractor (3DCAE). The main features of the flow fields are extracted using three layers in the encoder part of the 3DCAE.
