A Convolutional Neural Network for Detecting Tuta Absoluta Effects in Tomato Crops using YOLOv8

Abstract

 Tuta absoluta, a devastating pest of tomato crops, poses a significant threat to global food security and agricultural economies. Early and accurate detection of tuta absoluta is crucial for effective pest management and to minimize yield losses. This paper presents a deep learning approach to detect tuta absoluta using the state-of-the-art YOLOv8. We curated a comprehensive dataset of tomato leaf images, capturing various stages of infestation by tuta absoluta. Our methodology involved training the YOLOv8 model to identify and localize the pest’s presence and its characteristic damage patterns on tomato leaves. The results demonstrate that YOLOv8 achieves high precision and recall rates, resulting in an accuracy of 97% of identifying healthy tomato leaves and 85% of infected tomato leaves. This study underscores the potential of advanced deep learning techniques in plant disease detection and offers a promising tool for farmers and agronomists to combat tuta absoluta infestations. The implications of this research extend to the broader field of precision agriculture, where real-time monitoring and automated response to crop health threats are increasingly vital. This study is a contribution to the realm of agricultural pest detection, underscoring the promise of deep learning approaches in boosting the accuracy and trustworthiness of object detection algorithms.