Real-Time Plant Disease Identification via AI-Enabled Mobile App: A Comprehensive Review

Abstract

Plant diseases pose a major threat to agricultural productivity, especially in tomato cultivation, which is susceptible to a wide range of infections. Traditional identification techniques, based on manual inspection, are often time-consuming, error-prone, and unsuitable for large-scale deployment. With the evolution of deep learning and edge computing, real-time detection systems using mobile applications have emerged as a promising solution. This review explores a mobile-based system for disease in tomato leaf using a lightweight deep learning model, MobileNetV3, optimized for on-device inference. The integrated mobile app not only detects diseases from leaf images captured via smartphone cameras but also provides farmers with practical recommendations through a chatbot and fertilizer advisory module. The system design incorporates data preprocessing, transfer learning, and model optimization to improve accuracy and ensure generalization across various disease types. We also use Grad-CAM visualizations along with performance metrics to get a better sense of how well the classification is working and to interpret the results more clearly. The review compares different CNN architectures, hybrid models, and explainable AI frameworks to offer some useful insights. By combining AI tools with mobile access, this approach really helps push forward smart farming, especially in rural areas where resources are tight. Overall, this review aims to show how real-time AI applications on smartphones can really change how we handle crop diseases, flagging the way for smarter, more efficient agricultural practices in the future.