Numerous steps of seismic data processing rely on a velocity model of the subsurface.
Although full-waveform inversion has shown its capability to provide high-resolution images of the subsurface,
this method suffers from different problems such as cycle-skipping and crosstalk between parameters.
Currently, I'm working on generative adversarial networks (GANs) to build an accurate image of the subsurface while avoiding the issues of FWI.
I will update this post in the future.