Batiknet: Batik Classification-based Management Application for Inexperienced User

Batik has significantly contributed to the Indonesian economy, is diverse, and is spread throughout cities. Currently, batik patterns are very diverse and spread from Sabang to Merauke. Each batik pattern holds distinct meanings, philosophies of life, and ancestral heritage and reflects the region where it was crafted. We introduce a new batik dataset containing five patterns: Kawung, Megamendung, Parang, Sekarjagad, and Truntum. The Convolutional Neural Network (CNN) method is an effective Deep Learning method for extracting image information. CNNs have become the state of the art for various image processing tasks, such as classification, segmentation, and object recognition. This study used several state-of-the-art architectures, including Xception, ResNet50V2, MobileNetV2, and DenseNet169. However, we chose EfficientNetV2 as the primary feature extractor due to its superior performance. Our results show that EfficientNetV2 outperformed other architectures in training, validation, and testing accuracy, making it the best choice for classifying batik patterns. The training process resulted in an accuracy of 98% for training, 97% for validation, and 96% for testing. To ensure the accessibility and practical application of this research, we developed a user-friendly, web-based interface with a RESTful API, making the tool accessible to a broader audience. The application is named "BatikNet," a name chosen to reflect the blend of traditional batik culture ("Batik") with neural network technology ("Net"). This research contributes a valuable dataset and a practical tool for future studies and applications in batik pattern recognition and supports the preservation and understanding of Indonesian cultural heritage