Real?Time Detection and Segmentation of Tomato Pests with YOLOv8

Abstract

Tomato (Solanum lycopersicum L.) is vital for global nutrition and economic stability, yet it is threatened by pests such as Tuta absoluta, Helicoverpa armigera, and Bemisia tabaci. Effective pest management is crucial to prevent significant crop losses. Traditional pest detection methods relying on human observation are labor-intensive, time consuming, and prone to errors. In contrast, artificial intelligence (AI)based models such as YOLO provide timely and accurate pest identification, enhancing pest management practices. In this study, images captured throughout the tomato plant’s development, from seedling to fruit stage, were used for model training. The capabilities of the YOLOv8 model in detecting and segmenting tomato pests were evaluated. The results demonstrated significant improvements in both detection and segmentation tasks, with precision and recall reaching 98.91% and 98.98% for detection, and 97.47% and 98.81% for segmentation, respectively. These findings underscore the accuracy and robustness of the YOLOv8 model in monitoring diverse pest species, highlighting its potential to improve agricultural pest management practices. Although YOLO-based detectors have recently been tested on a limited set of pest species, comprehensive field-scale evaluations remain scarce. By assessing YOLOv8 across eleven pest taxa under commercial field conditions, this study delivers among the more comprehensive practice-oriented benchmarks to date for multi-species pest monitoring. This research suggests that integrating AI models like YOLOv8 into pest monitoring systems can contribute to more efficient and sustainable agricultural practices by minimizing human error and labor demands. Furthermore, future applications could extend this approach to other crops and pest species, validating the model’s versatility and supporting long-term farming sustainability.