Synthetic dataset for Deep Learning-based Inversion for Velocity Model Building



Five types of velocity models, namely 1500 pairs each for Linear, Fold, SALT, Image1 and Image2, (x-z domain, size: 201x301 ) simulated according to proposed methodology in our reference below and their corresponding shots (x-t domain, size: 2001 x 301) generated in Devito. Files are written in csv format and then separated in train, test and validation folders in order to be used for training and testing deep neural networks for performing Velocity Model Building. In order for the csv file to be transformed to matlab files, use the savemat() function from python library and in order to plot the pair images, use the matplotlib python library.
When you make use of the dataset please cite: A.Parasyris, L.Stankovic, V. Stankovic "Synthetic data generation for Deep Learning-based Inversion for Velocity Model Building", 2023.
Date made available14 Apr 2023
PublisherUniversity of Strathclyde
Date of data production1 Feb 2023 - 1 Mar 2023

Cite this