Deep Learning-Based Detection of Plant Leaf Diseases and Wilting Using a Mobile Application

Abstract

Plant diseases and wilting are major factors contributing to crop yield reduction and economic losses in agriculture. Early and accurate diagnosis is essential for effective disease management and sustainable crop production. However, manual identification of plant diseases is often time-consuming, labor-intensive, and prone to human error due to the visual similarity of symptoms among different diseases and wilting conditions. This study proposes a deep learning-based mobile application for the detection and classification of plant leaf diseases and wilting using a lightweight MobileNetV2 architecture. Transfer learning was employed to enhance classification performance while reducing computational requirements for mobile deployment. The model was trained using TensorFlow and subsequently converted into TensorFlow Lite (TFLite) format for integration into an Android mobile application. Experimental results demonstrated a testing accuracy of 94.17%, with healthy leaves achieving the highest classification accuracy of 98.12%. The proposed system enables farmers to capture or upload leaf images using a smartphone and obtain real-time disease predictions, while supporting cloud-based data storage and monitoring. Compared with conventional deep learning models that require high computational resources, MobileNetV2 provides an effective balance between accuracy, efficiency, and mobile deployment capability. The proposed solution contributes to precision agriculture by facilitating early disease diagnosis, timely intervention, and improved crop management.