Deep Learning technique for predicting tomato leaf disease

The increased focus on smart agriculture in recent years can be attributed to various factors, including the adverse effects of climate change, extreme weather events, population growth, the need for food security, and the scarcity of natural resources. The Moroccan government has undertaken proactive measures to address plant diseases, particularly in relation to tomatoes, which are globally recognized as one of the most significant vegetable crops and are susceptible to numerous diseases resulting in substantial yield reduction. The identification of tomato leaf diseases has increasingly shifted towards the utilization of deep learning techniques. In this study, we have conducted a comparative analysis of various deep learning techniques, specifically focusing on CNN models. Our objective was to determine the most suitable, efficient, and accurate method for detecting diseases that affect tomato leaves by merging two publicly available datasets, namely PlantVillage and PlantDoc. This research examines the potential of utilizing cutting-edge deep learning techniques that are focused on YOLO models. We have selected four models, namely YOLOv5, YOLOX, YOLOv7, and YOLOv8, which fall under the category of "One-stage detectors." These models are renowned for their fast inference speed and exceptional accuracy. Based on the experimental findings, YOLOv5 demonstrates the highest level of accuracy, achieving a mAP of 93.1% after fine-tuning the hyperparameters. The ultimate model is implemented as a mobile application to enhance usability.