Songket Motif Classification using MobileNet Architecture for Cultural Heritage Preservation

Abstract

Cultural heritage is a nation's identity that must be preserved. In Bali, Indonesia, Songket is a very valuable cultural heritage. There are several famous Songket motifs, such as Bunga Pucuk, Naga, Cangkir, Bintang, and Wayang. Unfortunately, many people have difficulty distinguishing the existing motifs. In this case, the use of CNN is one of the most proven effective ways to overcome this problem. In this study, we utilized MobileNetV2 to train the detection model with K-Fold Cross Validation to validate the model’s performance. According to the experimental result, we can obtain the best performance with accuracy=99.98%. In addition, we conducted usability testing with the System Usability Scale (SUS) that showed an average score of 86.33 as the "Best Imaginable" category. This research is expected to support cultural preservation and support the digital promotion of local Songket.