<![CDATA[The Batak Toba script is one of Indonesia’s cultural heritages that has become increasingly rare and less recognized among younger generations. This research aims to develop a handwriting recognition system for Batak Toba characters using the Convolutional Neural Network (CNN) method, capable of accurately recognizing characters, transliterating them into Latin script, and translating them into Indonesian. The dataset was self-generated using the Noto Sans Batak font and character combinations, totaling 113 labels, which were processed into 64×64 grayscale images. The CNN model was designed with several convolutional and pooling layers and compiled using the Adam optimizer and categorical cross-entropy loss function. Training results achieved a validation accuracy of 98.36% and a testing accuracy of 98.12%, with respective loss values of 0.0268 and 0.0295. The system was then integrated into a web-based application built as a Progressive Web App (PWA), supporting both online transliteration and translation features. These results demonstrate that the CNN approach is highly effective in recognizing Batak Toba characters. In the future, the system can be further developed into a full sentence-level OCR, integrated into a native Android application, and expanded with datasets from real handwritten samples.]]>
