Making a Batik Dataset for Text to Image Synthesis Using Generative Adversarial Networks

Abstract

Batik is a cultural heritage as well as the identity of the Indonesian nation that needs to be preserved. The use of deep learning allows the process of making batik patterns done by computer through the mechanism of text-to-image synthesis without humans needing to make it directly. The main contribution of this research is to produce a synthetic batik pattern that is similar to the original without removing the characteristics possessed by each batik pattern. This process of text synthesis to images uses the Generative Adversarial Networks (GAN) by first creating a system that can learn from a datasets. A varied and structured dataset can make it easier for the system to learn faster. In this study, a batik dataset was created for the synthesis of text into images.