Classification of West Java Batik Motifs Using Convolutional Neural Network

Abstract

The difference in the types of batik in West Java Province for the general public will not be seen significantly, because all the motifs at first glance look the same. Classification of batik motifs needs to be done to overcome the difficulties of different types of batik, in order to provide information and make it easier to distinguish the name of a batik motif and can tell the name of a batik motif for ordinary people who do not know the name of a batik motif with a fairly accurate level of accuracy. Classification of batik needs to be done to determine the type of batik from each region to make it easier to distinguish motifs from each region. The method used in this study is the classification of batik types in West Java Province using the Convolutional Neural Network (CNN) method. The results carried out for the classification of West Java batik image types using the Convolutional Neural Network (CNN) method that the feature extraction process can be carried out outside the process contained in the CNN algorithm or using feature learning depending on the needs of the research itself, and the results of the classification at 20 epochs and a learning rate value of 0.001 obtained an accuracy of 90% with a precision of 90% and a recall of 90%. This result is quite good considering the quality and amount of data obtained is not so good and the amount is not much.