<![CDATA[Javanese script is an Indonesian cultural heritage rich in historical, aesthetic, and spiritual values, but it is now becoming marginalized. To reintroduce its use, this research develops a Javanese script recognition application based on an Artificial Neural Network (ANN). In this study, the Javanese script was divided into 120 classes (ha, hi, hu, he, hee, ho, up to nga, ngi, ngu, nge, ngee, ngo). Each class was represented by 40 sample images of the script handwritten by 40 different respondents, resulting in 4800 samples. The research began with preprocessing, which included adding padding to the top, bottom, left, and right sides of the script; downsizing the image to a 33x33 resolution by applying average pooling; image segmentation to separate the script characters from the background; converting the color image to grayscale; and converting the grayscale image to a binary image with the help of thresholding. A number of images that had undergone preprocessing were then structured into a ready-to-use dataset of 4800 samples. This dataset was then divided with an 80:20 ratio, where 80% of the data was used to train the model and 20% was used to validate the model. An evaluation was conducted to measure the model's accuracy. Subsequently, the application was developed using PySide6 as the desktop interface. After the application development, the researchers provided an additional 600 images, where each class was represented by 5 samples, for real-world application testing. The evaluation results showed that the model achieved a validation accuracy of 70.21%. Meanwhile, testing with the application using the additional test images showed an accuracy of 73.83%.]]>
Copyrights © 2025
