Real-Time Tomato Leaf Disease Classification Using Efficientnet and UAV

Abstract

Tomato leaf diseases significantly hinder agricultural productivity, resulting in considerable yield losses when not detected and managed in a timely manner. Traditional disease detection methods, often reliant on manual inspection, are time-intensive, prone to errors, and unsuitable for large-scale monitoring. This study addresses these limitations by developing a real-time tomato leaf disease classification system leveraging a Tello drone integrated with a deep learning model based on EfficientNet-B5 and a Streamlit interface. The model was trained on the PlantVillage dataset, encompassing 9 distinct tomato leaf diseases and a healthy class, with data augmentation techniques applied to improve its generalization capacity. While similar models achieve high accuracy in controlled environments (99.8%), the novelty of this work lies in its adaptation for real-time field deployment, where the system achieved an accuracy of 96%. This research bridges the gap between laboratory-based deep learning systems and practical agricultural applications by enabling the drone to capture images dynamically, analyze them in real-time, and provide immediate feedback. The proposed system offers a scalable, efficient, and precise solution for early disease detection, emphasizing its transformative potential for precision agriculture and sustainable farming practices.