OpenFWI is a collection of large-scale, multi-structural benchmark datasets for machine learning-driven seismic full waveform inversion (FWI). There are twelve datasets synthesized from different priors, including one 3D dataset. It also provides baseline experimental results with four deep learning methods: InversionNet, VelocityGAN, UPFWI and InversionNet3D.
We have placed one of the Vel Family datasets, FlatVel_A, under the following directory to demonstrate OpenFWI training, testing and inference at Gadi.
/g/data/up99/sandbox/openFWI/FWIOpenData
You could run OpenFWI tutorials and models with module 'openfwi' under /g/data/up99, i.e.
$ module use /g/data/up99/modulefiles $ module load openfwi/23.11
You need to join NCI project up99 to access the OpenFWI dataset and environment.