In a recent publication, we showed how we can employ GANs for 3D flow reconstructions, overcoming certain limitations of planar estimators. In this work, the 3D-GAN we introduced is challenged to perform the same task with less information available. This problem tries to assess the effects of challenges that might affect the practical implementation of this problem. Different wall-sensor arrangement configurations employing fewer sensors are tested. The effect of employing only one type of sensor is also addressed, in combination with employing less sensors. It is shown that the quality of estimations, that were strongly influenced by the footprint of structures, now has a new influence due how the different sensor arrangements are capable of sensing these footprint patterns. The effect of noise in experimental measurements is also addressed.
Read the full open access publication here. The data sets and codes are also openly accessible.
