The Implementation of Deep Learning Technique in Mobile Application as a Preservation and Learning Media of Javanese Letter

Indonesia as the largest archipelagic country in the world has many local heritages. One of the valuable heritages is local language, which is 718 local languages in total from 37 provinces in Indonesia. Javanese language is a local language used by many people in Java Island, the most crowded mainland in Indonesia. This language has specific form of letter, differ from Latin alphabet. Because of that, nowadays people rarely learn how to write Javanese letter because of its difficulty. In this research, mobile application is developed using deep learning technique with VGG-16 convolutional neural network (CNN) architecture, as the cutting-edge method of artificial intelligence, to recognize Javanese letter and convert it into Latin alphabet. By using this application, younger generation can learn Javanese letter easily and in attractive way. Three main features are provided, i.e., scanning the letter, writing the letter, and converting the letter. Users are able to learn how to read and write Javanese letter, also receive feedback score from the system. This application is aimed as a preservation and learning media for those who want to learn Javanese letter using information technology. Based on testing result, this application has 89% of accuracy to recognize Javanese letter.