Convolutional Neural Network (CNN) Algorithm for Geometrical Batik Sade’ Motifs

Abstract

In Indonesia, batik was not popular among all socio-economic groups until the 20th century. Recently, batik has been considered an essential part of Indonesian culture and heritage. Geometric batik patterns are recognized by their symmetry, horizontal repetition, and vertical and diagonal angles between shapes. Sade is one village located south of Lombok island. Woven fabrics typical of Sade Village have distinctive motifs that differ from those of Sukarara Village, Central Lombok. Sade's batik mostly has geometric patterns that are almost similar. There are 5 motifs in Sade, namely Selolot, kembang komak, tapok kamalo, ragi genep and batang empat. The Sade village’s economy, which mostly relied on the sales of its fabric production, has been placed under an enormous burden by the COVID-19 pandemic. There must be a new and creative way in order to sustain its market penetration. One possible approach is by linking the community of Sade village fabric producers to the nationwide established marketplace. We propose an ML-based mobile web application that is supposed to be used by ordinary users, not only the tourists who visited Sade village. This mobile web main feature is to do the image classification of the aforementioned motifs and to provide a list of Sade village fabric sellers on the marketplace so that interested users may purchase the product. Models were created using the CNN algorithm to classify batik-sade images. CNN is one frequently used deep learning algorithm for image classification. Image datasets consist of training, testing, and validation datasets. The training datasets contain 2398 photos, while the testing and validation datasets each have 480 data. Ten epochs of experimental data revealed that the suggested CNN model has a training loss of 0.0560 and a training accuracy of 0.9805.