Development of Batik Pattern Classification Application Using Convolutional Neural Network Algorithm Using Android-Based Camera

Batik is a cultural heritage of Indonesia that has been recognized internationally and has received an award as a cultural heritage from UNESCO. However, the vast number of batik patterns in Indonesia makes it difficult for people, especially the general public, to identify batik motifs. To address this issue, a mobile Android-based application was developed to help the public gain a deeper understanding of batik motifs and their origins. This application was developed using the SDLC Waterfall model, the Kotlin programming language, Room for data storage, and the MVVM architecture. The model in this application was developed using PyTorch with the pre-trained MobileNetV3 Large model, ONNX, and TensorFlow Lite. The application was tested through black box testing, compatibility testing, confusion matrix analysis, and a T-Test. In the black box testing, all features were validated with a 100% success rate. Confusion matrix testing of the model was conducted on Android devices with both low-end and high-end specifications across 14 classes, showing fairly good results: 87% accuracy on high-end Android devices and 72% on low-end devices. These test results indicate that both types of Android devices perform well in predicting batik motifs with singular patterns. However, high-end Android devices are superior in predicting batik motifs with mixed patterns.