KLASIFIKASI AKASARA JAWA DENGAN CNN

Abstract

It is common knowledge that CNN is a significant method in image classification. This is because CNN can classify Latin letters with a high degree of accuracy. Lenet5 in CNN is tasked with converting 2D features from an image into a convolutional network continuously. CNN architecture consists of several layers, including the Convolution Layer, Relu layer, Subsampling layer, Fully Connected Layer. In this research, CNN is used to classify Javanese script images into 20 classes. These classes include ha, na, ca, ra, ka, da, ta, wa, la, pa, dha, ja, yes, nya, ma, ga, ba, tha, nga. Javanese script used in this research is Ngalena Javanese script. The precision values for each class range from 0.5 to 0.6.