Petrographic analysis with deep convolutional neural networks

Pires de Lima, Rafael

Petrographic analysis with deep convolutional neural networks

dc.contributor.advisor	Nicholson, Charles
dc.contributor.author	Pires de Lima, Rafael
dc.contributor.committeeMember	Shehab, Randa
dc.contributor.committeeMember	Slatt, Roger
dc.date.accessioned	2019-12-06T21:37:35Z
dc.date.available	2019-12-06T21:37:35Z
dc.date.issued	2019-12
dc.date.manuscript	2019-11-22
dc.description.abstract	Petrographic analysis is based on the microscopic description and classification of rocks and is a crucial technique for sedimentary and diagenetic studies. When compared to hand specimens, thin sections of rocks provide better and more accurate means for analysis of mineral distribution and percentage, pore space analysis, and cement composition. Because of the rich information they contain, thin section data are commonly used not only by the mining and petroleum industry, but by the academic community as well. Most petrographic analysis relies on visual inspection of rock thin sections under a microscope, a task that is laborious even for experienced geologists. Large projects with a tight time frame requiring the analysis of a large amount of thin sections may require multiple petrographers, thereby risking the introduction of inconsistency in the analysis. To address this challenge, we explore the use of deep convolutional neural networks (CNN) as a tool that can allow the petrographer to analyze and classify more samples in a consistent manner. Unlike previous studies using deep learning models trained on large volumes of thin section data, we make use of transfer learning based on robust and reliable CNN models trained with a large amount of non-geological images. With a much smaller number of labeled thin sections used in training followed by “fine-tuning” we are able to construct convolutional neural networks that achieve low error levels (<5% when images of same quality are used for training and testing) in thin section classification. While becoming widely accepted as a useful tool in the biological and manufacturing disciplines, CNN is currently underutilized in the geoscience community; we foresee an increase of use of such techniques to help accelerate and quantify a wide variety of geological tasks.	en_US
dc.identifier.uri	https://hdl.handle.net/11244/322836
dc.language	en	en_US
dc.rights	Attribution-ShareAlike 4.0 International	*
dc.rights.uri	http://creativecommons.org/licenses/by-sa/4.0/	*
dc.subject	convolutional neural networks	en_US
dc.subject	thin section	en_US
dc.subject	petrographic classification	en_US
dc.thesis.degree	Master of Science	en_US
dc.title	Petrographic analysis with deep convolutional neural networks	en_US
ou.group	Gallogly College of Engineering	en_US
shareok.orcid	0000-0002-6747-1291	en_US