An Open-Source Multi-Format Maize Leaf Image Dataset for Disease Detection and Classification

This dataset presents a comprehensive collection of maize (corn) leaf images captured from maize-growing regions in Bangladesh during June 2026. The dataset was developed to support research on automated plant disease diagnosis using machine learning (ML), deep learning (DL), and computer vision techniques. As maize has become an increasingly important crop in Bangladesh for food, livestock feed, and industrial applications, foliar diseases pose a significant threat to crop yield, grain quality, and farmers' income. Early and accurate disease identification is therefore essential for effective disease management and sustainable maize production. Existing maize disease datasets are often limited in terms of regional representation and data formats. Since disease occurrence and symptom characteristics vary according to climate, environmental conditions, and agricultural practices, a region-specific dataset is necessary for developing reliable diagnostic models. This dataset addresses these limitations by providing high-quality, well-annotated images collected under real field conditions in Bangladesh and made available in multiple formats to facilitate their use across different ML and DL frameworks. The dataset includes images belonging to ten disease categories and one healthy class: * Northern Corn Leaf Blight * Southern Corn Leaf Blight * Common Rust * Gray Leaf Spot * Leaf Blight * Eyespot * Curvularia Leaf Spot * Maize Streak Disease * Bacterial Leaf Streak * Healthy Leaves The diversity of disease classes and image conditions makes the dataset suitable for a wide range of applications, including image classification, object detection, disease severity assessment, transfer learning, and explainable artificial intelligence (XAI). It can also serve as a benchmark dataset for evaluating and comparing different machine learning and deep learning algorithms. The primary objective of this dataset is to assist researchers, data scientists, agricultural engineers, and plant pathologists in developing accurate and robust maize disease detection systems. In addition to supporting research in Bangladesh, the dataset may also be valuable for studies conducted in maize-growing regions with similar agroecological and climatic conditions. Ultimately, the use of this dataset can contribute to early disease diagnosis, improved crop management practices, reduced production losses, and enhanced maize productivity and profitability.本数据集收录了2026年6月于孟加拉国玉米（maize）种植区域采集的一套完整的玉米叶片图像集。本数据集的构建旨在支持利用机器学习（machine learning, ML）、深度学习（deep learning, DL）及计算机视觉技术开展自动化植物病害诊断的相关研究。 随着玉米在孟加拉国作为粮食、牲畜饲料及工业用途作物的重要性与日俱增，叶部病害已对作物产量、籽粒品质及农户收入构成严重威胁。因此，及早且精准的病害识别对于开展有效的病害管理与实现玉米可持续生产至关重要。 现有玉米病害数据集往往在区域代表性与数据格式方面存在局限。由于病害发生规律与症状特征会随气候、环境条件及农业生产模式的不同而发生变化，因此开发适配特定区域的数据集对于构建可靠的病害诊断模型至关重要。本数据集通过提供在孟加拉国真实田间环境下采集的高质量、精细标注图像，并以多种格式发布以适配不同机器学习与深度学习框架的使用需求，从而弥补了上述局限。 本数据集包含10类病害图像与1个健康叶片类别： * 北方玉米叶枯病（Northern Corn Leaf Blight） * 南方玉米叶枯病（Southern Corn Leaf Blight） * 普通锈病（Common Rust） * 灰斑病（Gray Leaf Spot） * 叶枯病（Leaf Blight） * 眼斑病（Eyespot） * 弯孢霉叶斑病（Curvularia Leaf Spot） * 玉米条斑病（Maize Streak Disease） * 细菌性叶条斑病（Bacterial Leaf Streak） * 健康叶片（Healthy Leaves） 病害类别与图像采集条件的多样性，使得本数据集可适用于多种研究场景，包括图像分类、目标检测、病害严重程度评估、迁移学习以及可解释人工智能（explainable artificial intelligence, XAI）。本数据集还可作为基准数据集，用于评估与对比不同机器学习及深度学习算法的性能。 本数据集的核心目标是协助研究人员、数据科学家、农业工程师及植物病理学家开发精准且鲁棒的玉米病害检测系统。除服务于孟加拉国内的相关研究外，本数据集对于具有相似农业生态与气候条件的其他玉米种植区域的研究也具有重要价值。最终，本数据集的应用可助力实现病害早诊断、优化作物管理方案、降低生产损失，并提升玉米产量与种植收益。