<![CDATA[Abstract— Bali is one of the provinces with quite complex cultural diversity in Indonesia. One of them is the characterization of traditional masks. Traditional masks in Balinese tradition are not only intended as performance accessories, but also as symbols of characterization, social status in indigenous communities, rites, and certain primordial activities. Every detail in the curve of the carving on the Balinese mask indicates an aesthetic richness that is certainly measurable as an ontological entity. In this case, the magnitude of this aesthetic measurability can be assisted by using various computational methods. This study tries to create a machine learning model with supervised learning to create a classification system for Balinese mask characterization. The methods used include: processing mask images into a 3-dimensional vector, each representing the red, green and blue color indices. Then each vector will go through a training process to create a measurability model for each characterization. The models used are EfficientNetV2 and Xception which are developments of convolutional models. The performance measurement metrics used are Accuracy, Precision, Recall & F1-Score. The Xception model produced an accuracy of 97%, while the EfficientNetV2 model produced an accuracy of 99%. Keywords— Bali, Classification, EfficientNetV2, Mask, Xception]]>
Copyrights © 2024
