Pembangkitan Pola Batik dengan Menggunakan Neural Transfer Style dengan Penggunaan Cost Warna

Abstract

In this research, the neural transfer styles technique was applied to transfer styles of pattern of Batik, a traditional Indonesian cloth painted using the wax-resist dyeing technique, to some certain images. The transfer was performed using a well-known convolutional neural network (CNN) architecture. Some neural transfer tests were done to produce solid color which originally came from color clustered images. The color cost function of the CNN was computed at every epoch of the iterative neural computation. The result of the transfer are images with clustered colors and a slightly apparent color gradient. The produced images can be classified as "Creative Batik".