--Overview--
3D-ESRGAN is developed to reconstruct 3D super-resolution channel flow from low-resolution data.
--Dependencies--
1. Python 3.6-3.8
2. tensorflow >=2.2.0<2.4.0 (cuDNN=7.6, CUDA=10.1 for tensorflow-gpu)
3. Numpy < 1.19
--Data preparation--
100 snapshots (channel flow at Reτ = 180) and its low-resolution data are provided for tutorial. Before the input, the data should be normalized using code nor_fluc3d.py to get normalized data.
--Training--
Use 3d-esrgan.py to train the deep learning model. It will save architecture and weights files automatically.
--BibTex citation--
@article{doi:10.1063/5.0129203,
author = {Yu,Linqi and Yousif,Mustafa Z. and Zhang,Meng and Hoyas,Sergio and Vinuesa,Ricardo and Lim,Hee-Chang},
title = {Three-dimensional ESRGAN for super-resolution reconstruction of turbulent flows with tricubic interpolation-based transfer learning},
journal = {Physics of Fluids},
volume = {34},
number = {12},
pages = {125126},
year = {2022},
doi = {10.1063/5.0129203},
URL = { https://doi.org/10.1063/5.0129203},
eprint = { https://doi.org/10.1063/5.0129203}
}
--Note--
There is a typo in eq(4) and (5). The sigmoid function should be applied for the whole right side of the equation.