Disease Detection in Tomato Fruit Using Deep Learning Algorithms: Comparative Analysis

Abstract

The agricultural sector is increasingly turning to advanced technologies to enhance productivity and meet the challenges of disease management. In this context, deep learning-based image processing techniques have become critical for disease detection, especially in tomato fruits. The main objective of this research is to evaluate the performance of the YOLOv8 model in tomato disease detection by comparing it against the well-established YOLOv5 model. The results show that YOLOv8 achieves higher accuracy in detecting diseased tomato fruits compared to YOLOv5 (98.0% vs. 97.2%), as well as superior precision (97.5% vs. 96.8%), recall (98.5% vs. 97.6%), and F1-score (97.8% vs. 97.0%). YOLOv8 also demonstrated a faster inference time (35 ms) than YOLOv5 (45 ms). In detailed comparisons by disease type, YOLOv8 outperformed YOLOv5 in every category – notably on Early Blight, where YOLOv8 attained 99.0% accuracy and a 98.8% F1-score. In summary, YOLOv8 provides overall superior performance, speed, and training efficiency over YOLOv5 in tomato disease detection. These advantages of YOLOv8 have the potential to increase productivity and reduce losses in agriculture by enabling early disease detection and intervention. The study also highlights that the success of deep learning models depends on the quality and quantity of labeled data, providing insights for the future development of AI-driven agricultural disease detection technologies.