Enhancing Batik Classification Leveraging CNN Models and Transfer Learning

Batik is a traditional art originating from Indonesia and recognized by UNESCO. Batik motifs vary depending on the region of origin. The diverse batik motifs reflect the rich cultural heritage and unique traditions owned by each region in Indonesia. From Sabang to Merauke, each motif features a different story and values, depicting the beauty and diversity of nature and the lives of diverse local people. However, in the context of the modern era that continues to develop, batik motifs also experience renewal and creativity that always adapts to the times. As a result, the diversity of batik motifs is increasingly abundant in Indonesia. Thus, complicating efforts to identify and categorize batik motifs appropriately. Therefore, in the context of this study, we chose to combine the MobileNetV2 model with Transfer Learning to classify batik motifs. We used a dataset consisting of 3000 batik images and categorized them into three main classes, namely Kawung batik, Mega Mendung batik, and Parang batik. This approach not only leads to the introduction and understanding of traditional batik motifs but also applies the latest technology for a more in-depth and accurate analysis. The results of this model show an extremely high level of testing accuracy, reaching 0.9946%, and training accuracy of 0.8916%, and the time required by the model to train and test the entire dataset is 18 minutes 1 second. Future research can explore the integration of other technologies or new approaches to improve accuracy and efficiency in classifying batik motifs.