Transforming Tradition: Utilizing MUNIT for Sasirangan Fabric’s Color Pattern Generation
Nurul Fathanah Mustamin, Mutia Nadra Maulida, Yuslena Sari, Andreyan Rizky Baskara, Eka Setya Wijaya, Irvan Aulia Luthfi · 2023
The Sasirangan fabric is a significant cultural relic in South Kalimantan, Indonesia. These fabrics have patterns with particular purposes, meanings, and colors that offer therapeutic effects. Over time, the role of Sasirangan evolved into a valuable commodity among the local community. As mentioned earlier, the statement underscores the imperative of revitalizing the production and utilization of Sasirangan fabric to safeguard and perpetuate this precious cultural legacy. This study employed the multimodal unsupervised image-to-image translation (MUNIT) technique to generate color patterns in Sasirangan images characterized by a white background and black patterns. The present study employs the MUNIT model with the most recent implementation configuration. Certain parameters were notably modified, including a reduction in the weight assigned to picture reconstruction and adjustments to the batch size. The experimental findings demonstrate that the MUNIT model effectively decreases the weight. The research results show that the MUNIT model with loss weight: image recon 10 and batch size 1 produces the best results with an FID value of 170.062 and an LPIPS value of 0.23587.