An effective deep learning approach for the classification of Bacteriosis in peach leave.

Abstract

Bacteriosis is one of the most prevalent and deadly infections that affect peach crops globally. Timely detection of Bacteriosis disease is essential for lowering pesticide use and preventing crop loss. It takes time and effort to distinguish and detect Bacteriosis or a short hole in a peach leaf. In this paper, we proposed a novel LightWeight (WLNet) Convolutional Neural Network (CNN) model based on Visual Geometry Group (VGG-19) for detecting and classifying images into Bacteriosis and healthy images. Profound knowledge of the proposed model is utilized to detect Bacteriosis in peach leaf images. First, a dataset is developed which consists of 10000 images: 4500 are Bacteriosis and 5500 are healthy images. Second, images are preprocessed using different steps to prepare them for the identification of Bacteriosis and healthy leaves. These preprocessing steps include image resizing, noise removal, image enhancement, background removal, and augmentation techniques, which enhance the performance of leaves classification and help to achieve a decent result. Finally, the proposed LWNet model is trained for leaf classification. The proposed model is compared with four different CNN models: LeNet, Alexnet, VGG-16, and the simple VGG-19 model. The proposed model obtains an accuracy of 99%, which is higher than LeNet, Alexnet, VGG-16, and the simple VGG-19 model. The achieved results indicate that the proposed model is more effective for the detection of Bacteriosis in peach leaf images, in comparison with the existing models.

Plant phenotyping relevance

モモ葉の病徴状態を画像から分類する深層学習手法を開発・比較評価しており、植物病害表現型の取得・推定が中心です。

Code and data availability

The supplied blocks describe a self-collected peach leaf image dataset (1000 original images from a research farm, augmented to 10000) and a proposed LWNet CNN model, but contain no public deposit, repository, or download URL for the dataset, images, code, or trained model. The only URL present (tribune.com.pk) is a新闻/

No evidence-backed public reproduction asset is currently recorded.