Arabic Calligraphy Image Analysis with Using Traditional Machine Learning Algorithms and Transfer Learning

The The Art of Calligraphy, which is defined as the art of writing arising from Arabic writing by undertaking the task of conveying religious messages that came with Islam, contains a lot of information about Turkish-Islamic art and civilisation. These artefacts are part of many historical sites in Istanbul. Understanding the messages written in the works is important for both locals and tourists in terms of making sense of the places in them and keeping the spirit of the place alive. The interpretation of millions of artefacts in Turkey and in the world is very difficult with a manual process. For this reason, to automate this process, both traditional machine learning models and transfer learning models were built on the classification of 9 different calligraphy styles and the classification of 32 letters in these calligraphy styles. The highest F1 scores were obtained from support vector machines and transfer learning models.