Would you like to try Deep Learning-based seismic velocity model building (VMB)?

Tomography and FWI solutions for the VMB problem struggle with requiring high computational power, long computational time, no guarantee whatever it would converge to a global solution, sensitivity to the low-frequency content, and requiring human interventions.

The GitHub repository that I'm taking today promises to solve some of the limitations of the conventional methods by using a convolutional neural network for velocity-model building directly from raw seismograms.

Authors: Yang, Fangshu and Ma, Jianwei
Deep-learning inversion: A next-generation seismic velocity model building method

Link to the GitHub repo:
Link to the LinkedIn post