Medicinal Plant Leaf Disease Dataset

This dataset contains 2,547 high-resolution leaf images (4000 × 3000 px, 4:3 aspect ratio) of three medicinal plant species, Kalanchoe pinnata (PatharKuchi), Azadirachta indica (Neem), and Ocimum tenuiflorum (Tulsi), collected from Senbag Upazila, Noakhali, Bangladesh, during August 2025. Each species includes one healthy and three disease or symptom classes, forming twelve balanced categories verified by agricultural experts. All images were captured in natural daylight using a 50 MP Honor 200 smartphone camera, ensuring clear color and texture representation. Images were taken from multiple angles and distances to preserve natural variations in illumination and background. Leaf samples were collected directly from local fields and home-grown medicinal gardens. Class distribution: Kalanchoe pinnata (PatharKuchi): Healthy (209), Web Blight (213), Yellow (204), Yellow Blight (213) → Total 839 Azadirachta indica (Neem): Healthy (244), Leaf Spot (211), Web Blight (205), Yellow (225) → Total 885 Ocimum tenuiflorum (Tulsi): Healthy (213), Downy Mildew (200), Web Blight (205), Yellow Spot (205) → Total 823 Grand Total: 2,547 images across 12 classes. Each image is stored in .jpg format (4000 × 3000 px) and named following a class-based convention (e.g., Neem_Healthy, Neem_LeafSpot, Tulsi_Healthy, Patharkuchi_YellowBlight). All files are placed in a single folder, with class names embedded in filenames for convenient parsing and automatic label extraction during model training. The dataset is designed for AI-based plant disease detection, particularly hybrid CNN/Transformer models and Explainable AI (XAI) research. It enables studies in agricultural image classification, medicinal plant health monitoring, and digital pathology applications. Verification and labeling were conducted under the supervision of the Upazila Agriculture Officer, Senbag, on 19 October 2025, ensuring correct disease identification and class validity.本数据集包含2547张高分辨率叶片图像（分辨率为4000×3000像素，宽高比4:3），涵盖3种药用植物物种：落地生根（Kalanchoe pinnata，别名PatharKuchi）、印楝（Azadirachta indica，别名Neem）以及圣罗勒（Ocimum tenuiflorum，别名Tulsi），采集于2025年8月的孟加拉国诺阿卡利县森巴格乌帕齐拉。每个物种包含1个健康类别与3个病害/症状类别，共计形成12个均衡分类，经农业专家核验无误。 所有图像均在自然日光下使用5000万像素荣耀200（Honor 200）智能手机相机拍摄，可清晰呈现叶片的色彩与纹理特征。拍摄过程采用多视角、多距离的采集方式，以保留光照条件与背景环境的自然变化。叶片样本均直接采集自当地农田与家庭药用花园。 类别分布如下： 1. 落地生根（PatharKuchi）：健康组（209张）、纹枯病（Web Blight，213张）、黄化症（Yellow，204张）、黄化纹枯病（Yellow Blight，213张），总计839张 2. 印楝（Neem）：健康组（244张）、叶斑病（Leaf Spot，211张）、纹枯病（205张）、黄化症（225张），总计885张 3. 圣罗勒（Tulsi）：健康组（213张）、霜霉病（Downy Mildew，200张）、纹枯病（205张）、黄斑病（Yellow Spot，205张），总计823张 总样本量：12个类别共计2547张图像。 每张图像均以.jpg格式存储（分辨率4000×3000像素），命名遵循基于类别的规范（例如Neem_Healthy、Neem_LeafSpot、Tulsi_Healthy、Patharkuchi_YellowBlight）。所有文件均存放于单个文件夹中，文件名中嵌入了类别信息，便于模型训练阶段的便捷解析与自动标签提取。 本数据集专为基于人工智能的植物病害检测任务设计，尤其适配混合卷积神经网络（Convolutional Neural Network, CNN）/Transformer模型以及可解释人工智能（Explainable AI, XAI）相关研究，可支撑农业图像分类、药用植物健康监测以及数字病理学相关应用研究。 本数据集的标注与核验工作由森巴格乌帕齐拉农业官员于2025年10月19日监督完成，确保病害识别与类别划分的准确性与有效性。