PENERAPAN ARSITEKTUR VISION TRANSFORMER DALAM PENENTUAN JENIS TANAH PADA CITRA DIGITAL

WIBOWO, ILHAM TRI and Desiani, Anita and Suprihatin, Bambang (2022) PENERAPAN ARSITEKTUR VISION TRANSFORMER DALAM PENENTUAN JENIS TANAH PADA CITRA DIGITAL. Undergraduate thesis, Sriwijaya University.

	Text RAMA_44201_08011281823024.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_44201_08011281823024_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (8MB) | Request a copy
Preview	Text RAMA_44201_08011281823024_0011127702_0026017102_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (1MB) | Preview
	Text RAMA_44201_08011281823024_0011127702_0026017102_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_44201_08011281823024_0011127702_0026017102_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_44201_08011281823024_0011127702_0026017102_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_44201_08011281823024_0011127702_0026017102_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_44201_08011281823024_0011127702_0026017102_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy

Abstract

Soil is a thin layer of tiny particles formed by the natural process of weathering rocks on the earth's surface. Soil characteristics vary greatly from place to place. Soil classification currently plays an important role in agriculture to determine the right plants for certain soil types. So it is necessary to have a technology that is able to classify soil types in order to minimize planting errors and increase agricultural production. Therefore, this study was conducted to classify soil images in determining soil types. Broadly speaking, soil can be classified into 5 classes,namely Black Soil, Cinder Soil, Laterite Soil, Peat Soil and Yellow Soil. Deep learning is often used to perform various tasks in machine learning such as classification. One architecture that is very well classified is the Vision Transformer (ViT). The advantages of ViT is that it performs better on small datasets and has patching techniques in its architecture. In this study, soil classification was carried out to determine the type of soil using the ViT architecture. The research stages are data collection, data pre-processing, data augmentation, application of ViT, training, testing, and performance evaluation as well as analysis and interpretation.The study used soil image data sourced from Kaggle by obtaining performance evaluations, namely accuracy, sensitivity, specificity, F1-score, and Cohen's kappa respectively 97.17%, 93.20%, 97.62%, 92.99% , and 89.85%. Based on these results, it shows that the ViT Small Variant architecture is able to perform classification tasks to determine the type of soil from the image data used.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Klasifikasi, Jenis Tanah, Vision Transformer
Subjects:	Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation. Q Science > QA Mathematics > QA1-43 General
Divisions:	08-Faculty of Mathematics and Natural Science > 44201-Mathematics (S1)
Depositing User:	Ilham Tri Wibowo
Date Deposited:	11 Nov 2022 03:12
Last Modified:	11 Nov 2022 03:12
URI:	http://repository.unsri.ac.id/id/eprint/81699

Actions (login required)

View Item