Classification of coastal batik and inland batik using machine learning Ardha Ardea Prisilla (a), Yori Pusparani (b), Maftuhah Rahimah Rum (c), Shinta Lidwina Djiwatampu (a), Chi-Wen Lung (d*)
a. Department of Fashion Design, LaSalle College Jakarta, Jakarta, Indonesia,
b. Department of Visual Communication Design, Budi Luhur University, Jakarta, Indonesia
c. Department of Bioinformatics and Medical Engineering, Asia University, Taichung, Taiwan
d. Department of Creative Product Design, Asia University, Taichung, Taiwan
Abstract
Batik is a form of wax-resist dyeing that is applied to an entire piece of fabric. In Java, batik is classified into inland and coastal regions based on its spread and development. The incorrect classification of batik will adversely affect the income of certain Indonesian regions due to the significant demand for certain batik designs on the market. There are hundreds of batik cloth motifs throughout Indonesia, each with its name and symbolism. As a result of the large number of batik patterns in Indonesia, it has been difficult for ordinary people, especially those unfamiliar with batik patterns, to identify motifs. Due to the difficulty of distinguishing batik patterns with the human eye, computer scientists have researched the classification of batik patterns using machine learning. Image classification datasets are obtained from Google Images using keywords specific to distinguishing between inland and coastal batiks. The inland and coastal batik were then classified using three machine learning types: Resnet-50, Pretrained Resnet-50, and Pretrained Inception. Results showed that Pretrained ResNet-50 demonstrated the highest performance accuracy at 93%, Resnet-50 had a lower performance accuracy at 81%, and Pretrained Inception had the lowest performance accuracy at 79%. Therefore, this study demonstrated that machine learning has the potential to classify inland and coastal batik and can be applied across various fields by providing a basis for more advanced transfer learning using artificial intelligence, thus facilitating the preservation and understanding of traditional Indonesian clothing.
