Predicting the 3D microstructure of SOFC anodes from 2D SEM images using stochastic microstructure modeling and CNNs

This repository provides the datasets used in this study. It includes 2000 virtual 3D microstructures of SOFC anodes generated using a low-dimensional stochastic model, along with their corresponding ground truth parameters, simulated 2D SEM images and computed geometrical descriptors. The data is structured in two hdf5 files, distinguishing between data used for training and validation, respectively. A README file gives all further information regarding the exct structure of the hdf5 files. The repository also includes a Python script (networks.py) defining the neural network architecture employed in this work.