A comprehensive dataset of rice leaf images for disease detection using machine learning.

Abstract

This manuscript presents a comprehensive, expert-annotated dataset comprising 19,000 rice leaf images, including 2,753 original images and 16,247 augmented images, sourced from the Bangladesh Rice Research Institute (BRRI). The dataset includes seven disease classes: Healthy (603 original images), Rice Blast (696 original images), Scald (421 original images), Leaf-folder Injury (247 original images), Insect Infestation (281 original images), Rice Stripes (266 original images), and Tungro Disease (239 original images). These images, captured under varying environmental conditions using smartphone cameras, accurately reflect real-world conditions. The images have been meticulously annotated by agronomy experts for reliable disease labeling. To enhance dataset diversity, data augmentation methods such as rotation, scaling, brightness adjustment, and horizontal flipping were systematically applied, expanding the dataset by creating additional variants from the original images. The dataset serves as a rich resource for developing machine learning models for the automatic detection of rice diseases. This initiative aims to enable early disease detection, promote sustainable farming practices, and improve food security, particularly in rice-dependent developing countries.

Plant phenotyping relevance

イネ葉画像を用いて病害状態を表現型として扱う、専門家注釈付きデータセットの構築・提供が中心であり、植物病害フェノタイピング手法の基盤となる。

Code and data availability

The paper is a Data in Brief article describing a rice leaf disease image dataset (19,000 images, 7 classes) publicly deposited on Mendeley Data with DOI 10.17632/vwv3nry3wr.1. This is the paper's own phenotyping image dataset and is directly actionable. No separate analysis code or trained model checkpoint is reported