Advanced YOLO models for real-time detection of tomato leaf diseases
Abdelaaziz Bellout, Mohamed Zarboubi, Azzedine Dliou, Rachid Latif, Amine Saddik
Abstract
The increasing focus on smart agriculture in the last decade can be attributed to various factors, including the adverse effects of climate change, frequent extreme weather events, increasing population, the necessity for food security, and the scarcity of natural resources. The government of Morocco adopts preventative measures to combat plant illnesses, specifically focusing on tomatoes. Tomatoes are widely acknowledged as one of the most important vegetable crops, but they are highly vulnerable to several diseases that significantly decrease their productivity. Deep learning algorithms are increasingly being used to identify tomato leaf diseases. In this study, we thoroughly examine different deep learning methodologies, with a specific emphasis on Convolutional Neural Network (CNN) models. Our study aims at identifying the optimal approach for detecting diseases that impact tomato leaves by combining two publicly accessible datasets, PlantDoc and PlantVillage. We focused on finding a strategy that is effective and efficient in accurately identifying these diseases. This study investigates the feasibility of employing state-of-the-art deep learning methods that are based on YOLO models. We have chosen five models, specifically YOLOv5, YOLOX, YOLOv7, YOLOv8, and YOLO-NAS, which belong to the category of "One-stage detectors." These models are widely recognized for their rapid inference speed and outstanding accuracy. According to the experimental results, YOLOv5 has the highest level of accuracy, reaching a mean average precision (mAP) of 93.1% after adjusting the hyperparameters. The final model is developed as a smartphone application to improve user-friendliness.