Design of an Android Application for Leaf Disease Detection in Plants

Abstract

Agriculture plays a strategic role in Indonesia's economy, with approximately 29,342,202 individual agricultural enterprises recorded in 2023, according to Statistics Indonesia (BPS). Golokan Village, located in Sidayu District, Gresik Regency, is one of the agrarian areas where 23.22% of the population works as farmers, and it has a total agricultural land area of 385 hectares. However, between 2019 and 2023, there was a significant decline in the production of three main commodities: corn decreased from 302.5tons to 275.6tons, tomatoes from 810tons to 585 tons, and cassava from 1,000tons to 832tons. One of the contributing factors is the difficulty in early detection of plant diseases. To address this challenge, this study designed and developed an Android application called AgroAI utilizing deep learning technology, specifically a Convolutional Neural Network (CNN) model based on the MobileNet architecture optimized with TensorFlow Lite for mobile devices. The development was carried out using the Scrum methodology in two sprints. The first sprint included needs analysis, dataset collection, interface design, and model training. The second sprint implemented the core features such as leaf disease detection via camera or gallery, classification results with recommended solutions, analysis history management, educational articles, and user authentication via Firebase. Black box testing confirmed that all features functioned as intended, while model validation achieved an accuracy of 94.74%. This application is expected to enhance farmers' efficiency in crop management and support the sustainability of both local and national agricultural sectors.