METODE ENSEMBLE PADA ARSITEKTUR VGG-16 DAN VISION TRANSFORMER DALAM KLASIFIKASI PENYAKIT DIABETIC RETINOPATHY

SALAMAH, FITRI and Desiani, Anita and Suprihatin, Bambang (2023) METODE ENSEMBLE PADA ARSITEKTUR VGG-16 DAN VISION TRANSFORMER DALAM KLASIFIKASI PENYAKIT DIABETIC RETINOPATHY. Undergraduate thesis, Sriwijaya University.

	Text RAMA_44021_08011182025003.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (2MB) | Request a copy
	Text RAMA_44021_08011182025003_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (8MB) | Request a copy
	Text RAMA_44021_08011182025003_0011127702_0026017102_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (8MB)
	Text RAMA_44021_08011182025003_0011127702_0026017102_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (6MB) | Request a copy
	Text RAMA_44021_08011182025003_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_44021_08011182025003_0011127702_0026017102_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (9MB) | Request a copy
	Text RAMA_44021_08011182025003_0011127702_0026017102_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (254kB) | Request a copy
	Text RAMA_44021_08011182025003_0011127702_0026017102_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (3MB) | Request a copy

Abstract

Diabetic retinopathy (DR) is a complication of the retina caused by diabetic disease. DR can be identified through images of the retina of the eye. One of the architectures that is often used to identify DR is VGG-16. VGG-16 is capable of handling large amounts of data and large attributes. However, the overfitting often occurs due to deep layers and multiple convolutions. Another method in deep learning that does not use convolution operations is the Vision Transformer (ViT). ViT has the advantage of not using convolution operations in the training process but using self attention. ViT tends to ignore minority classes so that the performance of the model in classifying the minority class is less optimal. The weaknesses of each architecture can be overcome by using ensemble learning. The study proposed an ensemble method that combines the results of VGG-16 and Vision Transformer using weighted voting techniques with ResNet. ResNet was used to help the model learn the weight patterns on each data and avoid over training in the classification of DR disease. The accuracy, precision, recall, and F1-score results in this study were excellent, above 95%. The ensemble method training graphs in the study showed that the study was able to cope with overfitting. The results show the ensemble method in this study is excellent and strong for the classification of diabetic retinopathy disease based on the severity of retinal image behavior, implement in 5 classes: Non Proliferative Diabetic Retinopathy (NDR), Mild, Moderate, Severe dan Proliferative Diabetic Retinopathy (PDR)

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	VGG-16, Vision Transformer MobileNet, ResNet, Weighted Voting
Subjects:	Q Science > QA Mathematics > QA8.9-QA10.3 Computer science. Artificial intelligence. Computational complexity. Data structures (Computer scienc. Mathematical Logic and Formal Languages
Divisions:	08-Faculty of Mathematics and Natural Science > 44201-Mathematics (S1)
Depositing User:	Fitri Salamah
Date Deposited:	25 Jan 2024 03:29
Last Modified:	25 Jan 2024 03:29
URI:	http://repository.unsri.ac.id/id/eprint/139613

Actions (login required)

View Item