Implementasi Metode Convolutional Neural Network (CNN) Untuk Mendeteksi Pengenalan Pola Huruf Korea (Hangul) (Studi Kasus: Hangul Jamo)

Vidia, Syahputri (2021) Implementasi Metode Convolutional Neural Network (CNN) Untuk Mendeteksi Pengenalan Pola Huruf Korea (Hangul) (Studi Kasus: Hangul Jamo). Undergraduate Thesis thesis, Institut Teknologi Telkom Purwokerto.

	Text Cover.pdf Download (823kB)
	Text Abstract.pdf Download (124kB)
	Text Abstrak.pdf Download (177kB)
	Text Bab I.pdf Download (421kB)
	Text Bab II.pdf Download (2MB)
	Text Bab III.pdf Download (2MB)
	Text Bab IV.pdf Restricted to Registered users only Download (5MB) | Request a copy
	Text Bab V.pdf Download (155kB)
	Text Daftar Pustaka.pdf Download (239kB)
	Text Lampiran.pdf Restricted to Registered users only Download (576kB) | Request a copy

Abstract

Korean fever or commonly called the Hallyu Wave is the impact of liking the entertainment industry such as KPOP, Kdrama from Korea and it has spread in Indonesia. From this, a study was carried out, namely classifying letters about Korean letters, namely Hangul Jamo which can recognize these characters. Based on previous research, the author conducted a study using the Convolutional Neural Network method by M.Zatsepin on Hangul letters and Kim Yeon Gyu both getting 0.99 accuracy, which is a very good method in performing image recognition performance. To find out a good final result, a calculation is carried out using the calculation of the number of correct images divided by the total image and a Classification Report based on the images in the form of Hangul Jamo characters with 213 classes. Each class consists of 2500 images from Unicode and the three models are distinguished by 3x3, 5x5, and 7x7 kernel sizes. After testing, the accuracy results are 99.791549, 99.440376, and 98.880751. using the average measurement, while for the classification report, the accuracy results are 1.00, 0.99, and 0.99 for each model. Then testing using images from the internet resulted in 7 correct images from 30 images, namely with the correct number of 4 images in the first model, for the second model 2 images were correct and the third model 1 image was correct. So first model more better and that from this research it is hoped that it can be developed with a wider hangul letter later and then can read other hangul letter. Keywords: Hangul Jamo, Convolutional Neural Network, Classification Report.

Actions (login required)

View Item