Development of a Progressive Web Apps with Artificial Intelligence for Rice Leaf Disease Classification

Abstract

This research presents the development of RiceCare PWA, an application for classifying rice leaf diseases using Artificial Intelligence (AI) integrated with Progressive Web Apps (PWA) technology. The system is designed to provide farmers with a convenient and rapid tool for analyzing diseases from leaf images, even in areas with limited internet connectivity. The AI model was trained on a 6-class dataset from the rice leaf disease dataset using the Google Teachable Machine platform. The model's performance was evaluated through eight statistical metrics: precision, recall, F1-score, specificity, NPV, FPR, FNR, and MCC. The results demonstrate high classification accuracy, with precision values ranging from 0.70 to 0.78, recall between 0.72 and 0.82, and MCC between 0.65 and 0.76, reflecting strong stability and correlation between predictions and actual ground truth data. By leveraging PWA technology, ricecare enables offline processing, which is crucial for field operations in remote areas. System testing across six Android smartphone models revealed an average image processing time of only 1.46 seconds and a high operational success rate of 99.4%, with only minor initial latency during the first-time model download. Field testing on 200 rice leaf images showed a 90.5% agreement rate between the application's results and expert farmers' assessments, with nearly 100% accuracy in identifying healthy leaves. Furthermore, a user satisfaction survey with 10 participants yielded an excellent rating of 4.65. These findings confirm that RiceCare PWA is an effective tool for rice disease detection, reducing diagnostic time, increasing accuracy, and serving as a sustainable framework for applying AI and PWA in community-level smart agriculture.