Arsitektur Convolutional Neural Network untuk Model Klasifikasi Citra Batik Yogyakarta

Abstract

Batik is an Indonesian culture that has been recognized as a world heritage by UNESCO. Indonesian batik has a variety of different motifs in each region. One area that is famous for its batik motifs is Yogyakarta. Yogyakarta has a variety of batik motifs such as ceplok, kawung, and parang which can be differentiated based on the pattern. Yogyakarta batik motifs need to be preserved so they do not experience extinction, one way is by introducing Yogyakarta batik motifs. The recognition of Yogyakarta batik motifs can utilize technology to classify images of Yogyakarta batik motifs based on patterns using the Convolutional Neural Network (CNN). The Yogyakarta batik motif images used for classification totaled 600 images consisting of 3 different motifs such as ceplok, kawung, and parang. Image classification using CNN depends on the architectural model used. The CNN architecture consists of two stages, namely Convolutional for feature extraction and Neural Network for classification. The CNN architectural models made for the introduction of Yogyakarta batik motifs totaled 7 models which were distinguished at the feature extraction stage. The highest accuracy results in the classification of Yogyakarta batik motif images using CNN were obtained in the 6th model. The 6th model has an accuracy of 87.83%, an average precision of 88.46% and an average recall of 87.66%. The accuracy, precision, and recall values ​​obtained by the 6th model are above 80%, which means that the 6th model can classify Yogyakarta batik motifs quite well.