Tomato (Solanum lycopersicum L.) leaf disease detection using computer vision

Abstract

Tomato cultivation is a key component of global agriculture, but is significantly threatened by pests and diseases that impact yield and trade. To address this, the present study investigates the application of YOLOv9, a state-of-the-art object detection model, for automated disease detection in tomato leaves. Using a dataset of 4,323 images with 15,135 annotations and a modified PlantVillage dataset, YOLOv9 models were trained and evaluated. Among the evaluated models, YOLOv9e yielded the highest mean average precision (mAP) at 0.964, establishing a benchmark for accuracy. In contrast, the YOLOv9t model provided an optimal balance for practical applications, achieving a competitive mAP of 0.95 with a rapid inference time of 8.8 ms. Furthermore, this work contributes a public version of the PlantVillage dataset with bounding box annotations, providing a valuable resource for object detection research and extending the use of the original classification-focused dataset. The results indicate that YOLOv9 models are effective for real-time and accurate detection of various diseases in complex agricultural settings.