Communication is the exchange of ideas and information between individuals, societies, or locales. Language serves as the primary medium for communication, with writing representing a pivotal tool within linguistic frameworks. The art of Arabic calligraphy, which has its origins in the Arabic script and was historically employed for conveying religious messages associated with Islam, embodies a rich source of knowledge pertaining to Turkish-Islamic art and civilization. These artistic artifacts are integral components of numerous historical sites throughout Turkey. Interpreting the written content within these works holds significant importance for both local inhabitants and tourists, as it enhances the understanding of the historical context and preserves the essence of these locations. It is very difficult to convey the meaning of millions of artifacts to people with a manual process. For this reason, to make this process automatic, the problems of identifying the styles of the writings in the works and recognizing the letters in the writings as a step in the transition from image to text in the literature have been studied. In this study, as a technique that has not been tried in other studies, Arabic calligraphy style determination and letter classification were performed by transfer learning, and an f1 score of over 79% was obtained.

Index Terms—Arabic calligraphy, convolutional neural network, image processing, transfer learning.