Skripsi

TRANSFORMASI ORIENTASI TEKS BALON PERCAKAPAN VERTIKAL PADA CITRA MANGA BERBAHASA JEPANG UNTUK OPTIMASI EASY OCR MENGGUNAKAN RF-DETR

---

Text in Japanese manga speech balloons is typically written vertically, whereas modern OCR technology is optimized for horizontal text. As a result, standard OCR engines like EasyOCR struggle to accurately read vertically oriented manga text. To address this issue, a method is proposed that uses character detection and text orientation transformation to improve OCR accuracy on vertical text. Specifically, an RF-DETR (Roboflow Detection Transformer) model is utilized to detect individual characters in the manga speech balloon text. The detected characters are then rearranged from a vertical orientation into a horizontal layout, producing an image that conforms to the conventional OCR reading order. The transformed image is subsequently processed with EasyOCR to extract the text. The results of this study demonstrate a significant improvement in OCR accuracy for vertical manga text. The average Character Error Rate (CER) dropped sharply from 0.9549 (before transformation) to 0.2888 (after transformation). Additionally, the RF-DETR model achieved an average Intersection over Union (IoU) of 0.8281 in character detection, indicating good localization performance. This drastic reduction in CER confirms that the orientation transformation approach can significantly enhance OCR accuracy on vertically oriented manga text.

---

Availability

Detail Information

Series Title

-

Call Number

T1930522026

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xv, 104 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.420 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Teknik Informatika
Pengenalan Karakter Bahasa

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment