Perbandingan Kinerja Backpropagation dan Convolutional Neural Network untuk Klasifikasi Citra Batik Lampung

Abstract

one of Indonesia's cultures on October 2, 2009. Lampung initially did not have a batik tradition but there is a legacy that is referred to as the first batik worn by Lampung people, namely sembagi cloth. Batik Siger is a business that produces batik typical of Lampung which originated from a course and training institution and was established in 2008. LKP Batik Siger provides services to the community in the field of written batik. This research discusses the performance of backpropagation and convolutional neural networks that will be used for the classification of Lampung batik image patterns. The Lampung batik motifs used are sembagi, pakjimo, granitan, soga, siger tangkup betik, jung agung, kembang cengkih and siger ratu agung. The stages that will be carried out are scaling, grayscale, thresholding and classification. The comparison of training data, testing data and validation used is 70:20:10 with the needs of backpropagation and convolutional neural network, namely epoch = 100, learning rate = 0.01. Backpropagation classification resulted in an accuracy of 96.25% and a classification error of 3.75%. The convolutional neural network classification resulted in an accuracy of 99.37% and a classification error of 0.63%. The performance of the CNN method has 3.12% higher accuracy compared to the performance of convolutional neural network.