MODIFIKASI ARSITEKTUR UNET-INCEPTION DENGAN PENAMBAHAN BATCH NORMALIZATION DALAM SEGMENTASI EKSUDAT PADA CITRA RETINA PENYAKIT DIABETIC RETINOPATHY

RAYANI, IRA and Desiani, Anita and Irmeilyana, Irmeilyana (2023) MODIFIKASI ARSITEKTUR UNET-INCEPTION DENGAN PENAMBAHAN BATCH NORMALIZATION DALAM SEGMENTASI EKSUDAT PADA CITRA RETINA PENYAKIT DIABETIC RETINOPATHY. Undergraduate thesis, Sriwijaya University.

	Text RAMA_44201_08011181924115.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (1MB) | Request a copy
	Text RAMA_44201_08011181924115_TURNITIN.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (6MB) | Request a copy
	Text RAMA_44201_08011181924115_0011127702_0017057403_01_front_ref.pdf - Accepted Version Available under License Creative Commons Public Domain Dedication. Download (1MB)
	Text RAMA_44201_08011181924115_0011127702_0017057403_02.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (624kB) | Request a copy
	Text RAMA_44201_08011181924115_0011127702_0017057403_03.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (319kB) | Request a copy
	Text RAMA_44201_08011181924115_0011127702_0017057403_04.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (858kB) | Request a copy
	Text RAMA_44201_08011181924115_0011127702_0017057403_05.pdf - Accepted Version Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (192kB) | Request a copy
	Text RAMA_44201_08011181924115_0011127702_0017057403_06_ref.pdf - Bibliography Restricted to Repository staff only Available under License Creative Commons Public Domain Dedication. Download (333kB) | Request a copy

Abstract

Diabetuc retinopathy is an effect of diabetes which is characterized by exudate. The segmentation process is needed to separate the exudate on the retinal image, to get clear exudate objects. The algorithm that is widely used in the segmentation process is the Convolutional Neural Network (CNN). The popular CNN architecture used for image segmentation process is UNet. However, the UNet architecture uses a large kernel size resulting in a large number of parameters. The UNet architecture consists of encoder lines and decoder lines connected by bridges. The bridge section usually uses a dense layer, which has a skip connection feature which results in missing feature information. Unlike the Inception module which does not implement a skip connection but replaces it with a parallel structure, the Inception module also implements a kernel with a smaller size to handle the large number of parameters. In this study, the exudate segmentation process on retinal images used a modification of the UNet-Inception architecture with the addition of batch normalization to make the weights converge more quickly. The results of the study yielded accuracy, sensitivity, f1-score, and Intersection over Union (IoU) values above 90%, which indicated that the model performance, prediction of exudate pixels, and prediction of segmentation results with ground truth were very good, and the specificity value was still below 85% is also in the good category in predicting background pixels.

Item Type:	Thesis (Undergraduate)
Uncontrolled Keywords:	Segmentation, image
Subjects:	Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation.
Divisions:	08-Faculty of Mathematics and Natural Science > 44201-Mathematics (S1)
Depositing User:	Ira Rayani
Date Deposited:	21 Aug 2023 02:07
Last Modified:	21 Aug 2023 02:07
URI:	http://repository.unsri.ac.id/id/eprint/126866

Actions (login required)

View Item