Scholarlink

This paper offers a state-of-the-art method to classify and identify intensity of tomato leaf diseases with the YOLOv8m deep learning model. The research looks the necessary requirement of effective and precise crop health monitoring in tomato crop agriculture, where classical methods are typically time-consuming and subject to human bias. The YOLOv8m model, one of the prominent single-stage object detection architectures, was chosen due to its excellent trade-off between speed and precision, suitable for real-time use in tomato crop disease intensity identification. The task is to detect and classify five different classes: Powdery Mildew, Bacterial Spot, Early Blight, Late Blight, and Healthy leaves. The dataset consists of 5114 images where each class consists of approximately 1000 images which makes the dataset balanced and efficient for training. The model was trained and tested on a wide dataset of labelled images, with the main goal of estimating the precise bounding box location of targeted regions on the leaf and accurately classifying the resultant disease category. The excellent performance of the YOLOv8m model is evidenced by a train mAP@0.5 of 99.2% and test mAP@0.5 of 96.2%. The more stringent mAP@(0.5–0.95) metric, accounting for both intensity identification and classification precision, achieved 92.7% on the test set and 87.03% on the validation set. These metrics confirm the high efficacy of the model and its potential to significantly improve disease regulation, reduce pesticide application, and crop yields.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Deep Learning Approach to Predict the Disease in Tomato Leaf with Bounding Box

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • V. Sahiti Yellanki,
  • Sai Vardhan Babu Gunda,
  • Geethika Rao Kalakonda,
  • Golivi Manoj Kumar,
  • Cherukuri Hemanth Kumar

摘要

This paper offers a state-of-the-art method to classify and identify intensity of tomato leaf diseases with the YOLOv8m deep learning model. The research looks the necessary requirement of effective and precise crop health monitoring in tomato crop agriculture, where classical methods are typically time-consuming and subject to human bias. The YOLOv8m model, one of the prominent single-stage object detection architectures, was chosen due to its excellent trade-off between speed and precision, suitable for real-time use in tomato crop disease intensity identification. The task is to detect and classify five different classes: Powdery Mildew, Bacterial Spot, Early Blight, Late Blight, and Healthy leaves. The dataset consists of 5114 images where each class consists of approximately 1000 images which makes the dataset balanced and efficient for training. The model was trained and tested on a wide dataset of labelled images, with the main goal of estimating the precise bounding box location of targeted regions on the leaf and accurately classifying the resultant disease category. The excellent performance of the YOLOv8m model is evidenced by a train mAP@0.5 of 99.2% and test mAP@0.5 of 96.2%. The more stringent mAP@(0.5–0.95) metric, accounting for both intensity identification and classification precision, achieved 92.7% on the test set and 87.03% on the validation set. These metrics confirm the high efficacy of the model and its potential to significantly improve disease regulation, reduce pesticide application, and crop yields.