This repository contains a collection of Notebooks used in Stanford course project CS230. The project proposes a method for Least-Squares Migration (Seismic Imaging algorithm) based on a Convolutional Neural Network CNN. The collection herein is not extensive nor self-contained as the data used in the training is not included in this repository.
Seismic imaging is used to map Earth geologic structures for energy resources exploration and characterization. Conventional methods yield blurred images with illumination artifacts. We developed a supervised convolutional neural network (CNN) approach to increase the resolution and compensate for illumination imbalance in conventional seismic images. By letting the CNN learn the appropriate mapping between input and output images, it can recover a better representation of the geologic structures.