Master Thesis Defense by Ask Hammer

Title: A comparison of deep learning approaches for seismic fault detection

Abstract:
The aim of the thesis is to explorer different approaches to the application of machine learning for fault detection on seismic profiles. The thesis examines the benefits and drawbacks to a variety of methods, but with a particular focus on dataset creation for neural network training and their resulting impact on the finished models. The project attempts to create a variety of synthetic seismic datasets containing fault annotation trough standard modelling and the use of generative adversarial networks. The synthetic data are then used to train machine learning models for fault detection, one such model is a convolutional neural network. The finished machine learning models are compared to each other and to conventional fault detection methods such as manual interpretation.

Supervisor: Klaus Mosegaard