Enhancing the Classification and Identification of Natural Rocks using Swin-Transformer Architecture

Bera, Subhashree (2022) Enhancing the Classification and Identification of Natural Rocks using Swin-Transformer Architecture. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration manual) Download (1MB) | Preview

Abstract

The identification and categorization of rock lithology are crucial topics in the geological survey. The reliability of the classification cannot be assured since the identification technique is based on rock-thin layers and seems to have a long identification duration and high cost. Additionally, the aforementioned approach is unable to offer a practical answer. The majority of geological survey employees carry devices, which are transceivers with several detectors. During the extraction of the rocks, natural calamities are always a risk therefore, in this study deep learning-based methods are discussed which can recognize the rocks by images and these images can be collected through drones. This method also ensures reliability and consumes less time in comparison to the traditional methods. Deep learning is succeeding in recognition and classification, which is simplifying the laborious process of categorization and recognizing images. This field of study is still fully unexplored. Four distinct deep learning models are used in this task which are VGG-19, Inception V3, Custom Model, and Swin Transformers. Important results have been discovered by employing the state-of-the-art design of deep learning models. Here, a bespoke model is created using convolutional blocks, and a swin transformer is a cutting-edge model. Vgg-19 and Inception V3 are based on a transfer learning technique. All models are evaluated on test data and assessed using various metrics after being trained on the rock’s image data. The categorization of images of rocks into several groups in real-world tasks may be accomplished by using the swin transformers model, which has proven to be superior to other models after assessment of the model’s using metrics like Accuracy, validation loss, Precision, and Recall.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	Swin-Transformer; Rock classification; VGG-19; Geographical; Inception-V3; Deep learning algorithms
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QE Geology Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	18 Jan 2023 17:16
Last Modified:	06 Mar 2023 16:48
URI:	https://norma.ncirl.ie/id/eprint/6086

Actions (login required)

View Item