This is a supporting repositories of our manuscript 'A Conditional GAN-based Approach to Build 3D Facies Models Sequentially Upwards' submitted to Computers & Geosciences
The dataset is available at https://github.com/GeoDataScienceUQ/GANRiverI
The Fluvial GAN for 2D simulation is available at https://github.com/GeoDataScienceUQ/Fluvial_GAN
To train your own version, please download the dataset (this work uses the 7-facies version as default) and set up your path.
Then run 'train_3d.py' in the fold that you'd like to try.
After training both 2D (Fluvial GAN) and 3D models (FluvialGAN_3DR), use the simulator (FluvialGAN3D) in 'test_all.py' to start random simulation or feed your latent/noise vectors. Feel free to change the num and Nav value in the class Simulation() to have different number of layers/thickness and correlation strength between layers/slices.
This code borrows heavily from https://github.com/NVlabs/SPADE