<![CDATA[Indonesia, as an agrarian country, faces serious challenges in agriculture, particularly plant leaf diseases that may lead to crop failure. Early detection is crucial to prevent their spread. This study aims to develop a leaf disease classification model using a Convolutional Neural Network (CNN) and implement it in an Android application called Leaf Lives. The model was developed using the EfficientNetB0 architecture and optimized with Stochastic Gradient Descent (SGD), trained for 30 epochs and fine-tuned for an additional 10 epochs. The dataset consisted of 44,030 leaf images covering 11 plant species and 60 disease or leaf-condition classes, obtained from public datasets available through platforms such as Kaggle and Mendeley Data. Four data split scenarios were evaluated (75:15:10, 70:20:10, 65:25:10, and 60:30:10). The best performance was achieved with the 70:20:10 split, yielding 92.03% accuracy, 91.12% precision, 89.89% recall, and 89.96% F1-score. The resulting model was then integrated into an Android application that classifies leaf images captured using the camera or selected from the gallery and stores classification history offline. The developed application is expected to assist farmers in identifying potential leaf diseases quickly and independently; however, further validation using real-field images and user-based evaluation is required to assess its practical effectiveness.]]>
Copyrights © 2026
