Affordable Deep learning-based Leaf Disease Detection and Localization for Precision Agriculture

Accurate detection of plant diseases is essential for effective crop treatment and improving yield and quality. The traditional manual method to monitor plant disease is time-consuming, labor-intensive, and requires specialized knowledge. In this context, deep learning techniques have been developed as advanced methods for detecting and classifying plant diseases. This paper contributes to the advancement of technology in smart agriculture by presenting a method for detecting and localizing diseases in plant leaves. Our approach used YOLOV5x model and a self-generated dataset, collected from Tunisian fields, to train and validate the model. The results demonstrate that our method achieves a mAP of 81%. Additionally, the proposed model operates efficiently at 17 FPS with a computational demand of 203.9 GFLOPs and a memory size of 173 MB. These operational parameters highlight the model's accuracy, speed, and computational efficiency, which significantly advance smart agricultural technology.