Arabic Calligraphy Generation Through Image-to-Image Translation Using Generative Adversarial Networks (GANs)

Arabic calligraphy is a revered art form in the Arab world that has seen limited exploration in automated generation due to its complex nature and aesthetic principles and rules. However, recent advancements in deep learning provide promising avenues to address this challenge. In this work, we investigate the automatic generation of Arabic Calligraphy through image-to-image translation. Our approach utilizes Generative Adversarial Networks (GANs), a highly effective class of deep learning models recognized for their exceptional data generation abilities, to convert source font images (such as Arial) into desired calligraphy styles. Additionally, we propose a dataset comprising aligned images of conventional font and corresponding calligraphy styles. We evaluate the quality of the generated images on the proposed dataset. Our experiments yield promising results, demonstrating significant progress in the automatic generation of Arabic calligraphy images.