KLASIFIKASI TULISAN TANGAN AKSARA SUNDA MENGGUNAKAN EKSTRAKSI CIRI LBP DAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK

Dika Rahman Maulana, . (2024) KLASIFIKASI TULISAN TANGAN AKSARA SUNDA MENGGUNAKAN EKSTRAKSI CIRI LBP DAN ALGORITMA CONVOLUTIONAL NEURAL NETWORK. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.

	Text ABSTRACT.pdf Download (10kB)
	Text AWAL.pdf Download (3MB)
	Text BAB I.pdf Restricted to Repository UPNVJ Only Download (617kB) | Request a copy
	Text BAB II.pdf Restricted to Repository UPNVJ Only Download (1MB) | Request a copy
	Text BAB III.pdf Restricted to Repository UPNVJ Only Download (881kB) | Request a copy
	Text BAB IV.pdf Restricted to Repository UPNVJ Only Download (3MB) | Request a copy
	Text BAB V.pdf Download (595kB)
	Text DAFTAR PUSTAKA.pdf Download (1MB)
	Text DAFTAR RIWAYAT HIDUP.pdf Restricted to Repository UPNVJ Only Download (68kB) | Request a copy
	Text HASI PLAGIARISME.pdf Restricted to Repository staff only Download (1MB) | Request a copy
	Text ARTIKEL KI.pdf Restricted to Repository staff only Download (811kB) | Request a copy
	Text LAMPIRAN.pdf Restricted to Repository UPNVJ Only Download (1MB) | Request a copy

Abstract

The Sundanese script, a visual representation of the Sundanese language, has declined in use but remains a vital part of Sundanese cultural heritage (Nurwansyah, 2015). In the digital era, there is a need to integrate and promote the Sundanese script through advanced technology. This study proposes an innovative approach to classify handwritten Sundanese script by combining Local Binary Pattern (LBP) and Convolutional Neural Network (CNN). LBP is effective in texture extraction, and CNN excels in pattern recognition, expected to create an accurate and efficient classification model. The dataset consists of 1,600 images of handwritten Sundanese script in 40 classes. Data preparation includes resizing, grayscale conversion, thresholding, noise and grid removal, contour identification, region grid extraction, image cropping, data labeling, and data augmentation. The CNN model was trained with and without LBP for 30 epochs using early stopping monitoring 'val_accuracy' and 'val_loss'. Evaluation results show that the CNN model without LBP achieved 92.11% accuracy with a loss of 0.7429, while the model with LBP achieved 71.21% accuracy. The confusion matrix and classification report indicate that the model without LBP performs better in recognizing Sundanese script classes. Hypertuning parameters were conducted for optimal performance, showing that although the model without LBP showed the best performance, hypertuning remains essential. This research demonstrates that while LBP aids in identifying textures and patterns, models without LBP are superior in accuracy. This study aims to develop technological solutions to promote and preserve the Sundanese script in the digital era and significantly contribute to the scientific literature in pattern recognition and information technology.

Item Type:	Thesis (Skripsi)
Additional Information:	[No. Panggil: 2010511122] [Pembimbing 1: Mustofa Galih Pradana] [Pembimbing 2: Muhammad Panji Muslim] [Penguji 1: Ridwan Raafi'udin] [Penguji 2: Ika Nurlali]
Uncontrolled Keywords:	Sundanese Script, Convolutional Neural Network, Local Binary Pattern, Handwriting Classification, Hypertuning Parameter
Subjects:	P Language and Literature > P Philology. Linguistics Q Science > QA Mathematics > QA75 Electronic computers. Computer science T Technology > TR Photography
Divisions:	Fakultas Ilmu Komputer > Program Studi Informatika (S1)
Depositing User:	DIKA RAHMAN MAULANA
Date Deposited:	09 Sep 2024 08:07
Last Modified:	09 Sep 2024 08:07
URI:	http://repository.upnvj.ac.id/id/eprint/31283

Actions (login required)

View Item