Seismic inversion with deep learning

Abstract This article investigates bypassing the inversion steps involved in a standard litho-type classification pipeline and performing directly from imaged seismic data. We consider set of deep learning methods that map data into classes, trained on two variants synthetic data: (i) one which we image using local Radon transform to obtain angle gathers, (ii) another start subsurface-offset based correlations over Our results indicate this single-step approach provides faster alternative established while being convincingly accurate. observe adding background model as input network optimization is essential correctly categorizing litho-types. Also, starting gathers obtained by imaging domain more informative than subsurface offset input.