<![CDATA[Manual identification of coffee leaf diseases remains challenging because it is time-consuming and highly dependent on the user's ability to recognize disease symptoms. In addition, most previous studies on plant disease classification have focused on model evaluation in desktop environments, while the implementation of on-device inference on mobile devices remains limited. This study aims to develop an Android-based coffee leaf disease detection application using a Convolutional Neural Network (CNN) with the MobileNetV3Small architecture. The dataset consists of 840 coffee leaf images categorized into three classes, namely healthy leaves, leaf spot, and leaf rust. The research stages include image preprocessing, data augmentation, model training, model evaluation, and TensorFlow Lite implementation on an Android application. The results show that the proposed model achieved an accuracy of 95% on the test dataset with satisfactory classification performance across all classes. The model was successfully deployed on Android devices using TensorFlow Lite, resulting in a model size of 1.1 MB and an average inference time of 0.37 seconds without requiring an internet connection. Furthermore, black-box testing confirmed that all application functions operated properly. The contribution of this study lies in the development of a MobileNetV3Small-based coffee leaf disease classification application implemented using TensorFlow Lite on Android devices. In addition to employing preprocessing and data augmentation to enhance the model's generalization capability, this study demonstrates that the on-device inference approach enables offline disease detection with low computational requirements, making it suitable for resource-constrained mobile devices.]]>
Copyrights © 2026
