Detection of Image-based Plant Leaf Diseases Using Convolutional Neural Networks

Abstract

The application of Deep Learning (DL) and Machine Learning (ML) has significantly improved the processing of images in the past few years over traditional approaches. With the ever-growing world population and increased demand for food, global food production depends heavily on plants. Plant diseases are caused by a variety of environmental variables, which significantly reduce productivity. This poses an urgent threat to food security, yet in several parts of the world, it is still challenging to quickly identify them. Traditional methods of detecting and preventing the spread of plant disease have become less efficient, inaccurate, and time-intensive, thus an automated and reliable approach to detecting these diseases is imperative. By enabling early detection of plant diseases, adopting cutting-edge technologies like ML and DL can assist in overcoming these difficulties. This study aimed to design a Convolutional Neural Network (CNN) model for detecting diseases in plants using publicly available plant leaf datasets obtained from Kaggle. The datasets contain Tomato, Potato, and Pepper bell leaves that are diseased and healthy. The efficacy of the model was assessed utilising precision, recall, and F1 score. CNN had an accuracy of 94%. Thus, the detection and proper diagnosis of plant disease is principal for successful plant growth and this reduces the threats it poses to food security in the world. It is recommended that the proposed deep learning model be translated into a system suitable for plant disease detection.

Plant phenotyping relevance

植物葉画像から病害状態を推定するCNNモデルの設計と性能評価が研究の中心であり、植物病害という表現型状態を直接推定しているため。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.