Web-Based Sustainable Detection and Treatment Recommendation System for Wheat Plant Diseases Using Convolutional Neural Networks.

Abstract

Wheat, being a major staple crop worldwide, is often attacked by rust diseases, which cause severe yield losses. The early detection and diagnosis of fungal infections, yellow rust, and brown rust are critical in minimizing their consequences. A web-based system based on a Convolutional Neural Network (CNN) was developed for the quick identification and classification of wheat plant diseases. The diseases that we examine in wheat plants are brown rust (BR) and yellow rust (YR), and healthy plants are classified in the third category. A dataset of labeled images of YR, BR, and healthy wheat plants was used to train the CNN. The model achieved a remarkable 96% classification accuracy. In addition to disease diagnosis, a recommendation module that gives advice on proper treatment based on disease names or symptoms is also provided. This twofold functionality allows for timely disease management and identification and facilitates the treatment of other wheat diseases besides rust diseases. Integrating the trained CNN model into an intuitive web application makes it user-friendly for end users, notably farmers, to have a practical tool in protecting wheat crops.

Plant phenotyping relevance

小麦植物画像から病害状態を分類するCNN手法を開発・評価しており、植物病害表現型の取得・推定が中心。治療推薦機能もあるが、画像ベース病害診断が主要な技術的貢献である。

Code and data availability

The paper's wheat disease image dataset (YR, BR, healthy; 3679 images) is a publicly available Kaggle dataset explicitly used for the CNN training, with an authors-provided URL matching an allowed URL.