Arabic Generated Text in Images - Methods, Challenges, and Enhancements
Supervisor Name
Radi Jarrar
Supervisor Email
rjarrar@birzeit.edu
University
Birzeit University
Research field
Computer Science
Bio
I am an assistant professor of Computer Science at the Faculty of Engineering and Technology, Birzeit University, Palestine. Currently, I am the director of the PhD program in Computer Science at Birzeit University. I obtained my B.Sc. in Computer Information Technology from the Arab American University in 2007 and the Ph.D. from Monash University in 2012. My research interests include machine learning, data science, and computer vision, with applications in health informatics and computer security.
Description
This project aims at addressing a major limitation in current image-generation systems characterized in the inability to correctly and consistently render Arabic text inside generated images. While existing models perform well for Latin-based scripts, they often fail to produce valid Arabic characters and proper word structures. To address this issue, this work aims to leverage a model that could generate readable Arabic text in images. We aim to achieve this through fine-tuning a diffusion-based model and integrate an OCR-guided discriminator to improve both the visual clarity and readability of gen- erated Arabic text. In addition, a diverse dataset was created by collecting real-world Arabic images, advertisements, and printed materials to better represent the variabil- ity of Arabic text in practical settings. The proposed system aims to support a wide range of applications that involve images that contain Arabic text, including posters and advertisements.
