Handwritten Katakana Character Recognition Using ResNet-18 with CLAHE-Based Image Enhancement

Christanti, Shella (2026) Handwritten Katakana Character Recognition Using ResNet-18 with CLAHE-Based Image Enhancement. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22081010058-cover.pdf Download (681kB)
	Text (Bab 1) 22081010058-bab1.pdf Download (190kB)
	Text (Bab 2) 22081010058-bab2.pdf Restricted to Repository staff only until 22 July 2029. Download (1MB)
	Text (Bab 3) 22081010058-bab3.pdf Restricted to Repository staff only until 22 July 2029. Download (1MB)
	Text (Bab 4) 22081010058-bab4.pdf Restricted to Repository staff only until 22 July 2029. Download (5MB)
	Text 22081010058-bab5.pdf Download (166kB)
	Text 22081010058-daftarpustaka.pdf Download (182kB)
	Text 22081010058-lampiran.pdf Restricted to Repository staff only until 22 July 2028. Download (812kB)

Abstract

Handwritten Katakana characters exhibit substantial variations in writing styles and high inter-character similarity, making their classification a challenging task. This study aims to analyze the performance of a transfer learning-based ResNet-18 model for handwritten Katakana character classification, evaluate the effect of applying Contrast Limited Adaptive Histogram Equalization (CLAHE), and compare the model performance before and after CLAHE enhancement. The research process consists of dataset collection, image preprocessing, dataset splitting, model training, hyperparameter tuning, and model evaluation using accuracy, precision, recall, and F1-score metrics. The best configuration was obtained using a dataset split ratio of 70:15:15, a learning rate of 0.001, a batch size of 16, the Adam optimizer, 50 epochs, a clip limit of 2.0, and a tile grid size of 8 × 8. The ResNet-18 model with CLAHE achieved an accuracy of 95.10%, precision of 95.43%, recall of 95.10%, and F1-score of 95.09%, whereas the model without CLAHE achieved an accuracy of 89.75%, precision of 91.19%, recall of 89.85%, and F1-score of 89.94%. These results demonstrate that CLAHE enhances image quality and improves the ability of the ResNet-18 model to classify handwritten Katakana characters. Therefore, the combination of CLAHE and ResNet-18 is effective in improving classification performance. Keywords: Handwritten Katakana Characters, Katakana, ResNet-18, Transfer Learning, CLAHE

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Puspita Sari, Anggraini	NIDN0716088605	anggraini.puspita.if@upnjatim.ac.id
Thesis advisor	Nugroho, Budi	NIDN0707098003	budinugroho.if@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)
Q Science > QA Mathematics > QA76.87 Neural computers
T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Shella Shella Christanti

Date Deposited:

22 Jul 2026 04:31

Last Modified:

22 Jul 2026 06:57

URI:

https://repository.upnjatim.ac.id/id/eprint/56515