Pooling Comparison in CNN Architecture for Javanese Script Classification

Abstract

Javanese script is evidence of the past culture, which contains various current language learning, including script recognition. However, learning traditional scripts becomes less attractive to the students. Thus, we propose a learning method to enable character recognition among students to deal with the issues. We offer a novel CNN architecture and compare different pooling layers for Javanese script classification. We calculate the separate pooling layer to reduce extensive feature extraction of the image. We present the model comparison results in Javanese character classification to convince our development.