Abstract
Summary Deep learning and deep convolutional neural network (CNN) models have shown promising results and are gaining popularity in the geoscientific community. In contrast to traditional machine learning methodologies based on a suite of carefully selected attributes, deep learning is based on the raw images themselves. Deep CNNs are currently the tools of choice for computer vision tasks such as self-driving cars. Unfortunately, deep learning is encumbered by jargon that is unfamiliar to most geoscientists, providing black box applications resulting in two common reactions: deep learning models are the solution for everything or deep learning models are a modern fad that discards the interpreter's insight or experience with a given problem. In this presentation, we show that CNN models are based on attributes similar to those we use in seismic interpretation and remote sensing. We also show that through a process called transfer learning based on the analysis of 2D colour images, we can exploit much of the previous work developed for image recognition applications to rocks. We illustrate the successful use of transfer learning to microfossil classification, core description, petrographic analysis, and hand specimen identification. We also discuss some of the challenges in CNN analysis of 3D seismic data volumes.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.