Machine learning in geoscience: using deep learning to solve the TGS Salt Identification challenge

Marcelo Guarido, Junxiao Li, Raul Cova

Deep learning, or neural networks, contain a widely range of applicability, that goes from regression of business analyses to treats identification on medical images. In this paper, we successfully applied an U-net based image semantic segmentation to identify salt bodies using only seismic images from the TGS Salt Identification Challenge. The process is simple applied with moderate computer requirement for a small set of images, but it can grow exponentially as more images are included. In the end, we could train a model that gives a 0.8 score on the IoU metric.