Rice Plant Disease Detection System based on Leaf Image using Web-based CNN Algorithm

Abstract

Rice (Oryza sativa) plays a crucial role as a major staple food commodity. However, diseases such as Bacterial Blight, Brown Spot, and Leaf Blast can cause significant crop losses. Current manual identification methods have limitations due to high subjectivity and long diagnosis time. This study proposes a web-based automatic detection system using a Convolutional Neural Network (CNN). The dataset was obtained from Kaggle and consisted of 2,800 images evenly distributed across four classes (700 images per class). The data were split using an 80:20 ratio for training and validation sets, followed by preprocessing steps including resizing to 224×224 pixels and data augmentation. The CNN architecture was designed with four convolutional blocks and optimized using the Adam optimizer. Training for 50 epochs achieved an accuracy of 77.50%, precision of 82.98%, recall of 77.50%, and an F1-score of 72.84%. Based on the confusion matrix analysis, the model performed very well in detecting Bacterial Blight and Brown Spot but still faced difficulties in identifying the Leaf Blast class. Overall, the developed system has the potential to serve as a decision-support tool for farmers, although further performance improvements are required, particularly for detecting specific disease variants.

Plant phenotyping relevance

イネ葉画像から病害状態をCNNで自動推定する手法の開発・評価が研究の中心であり、植物の病害表現型を直接測定しているため採用。

Code and data availability

The paper uses a public Kaggle rice leaf disease image dataset (2,800 images, four classes) but provides no dataset identifier/URL, no author code, model checkpoints, or supplementary deposit. The Kaggle dataset is a third-party source without a paper-specific link, and no authors' public analysis code or trained model

No evidence-backed public reproduction asset is currently recorded.