A Real-World Hibiscus and Tea Leaf Image Dataset for Classification

The Combined Hibiscus and Tea Leaf Image Dataset is a comprehensive and well-balanced collection comprising 1,413 original high-quality leaf images captured under natural outdoor conditions across various regions of Bangladesh using a SONY α7 II DSLR camera and a OnePlus 7T smartphone. The dataset includes two major plant species—Hibiscus and Tea—and aims to facilitate research in plant disease detection, agricultural image analysis, and computer vision–based crop health monitoring. The Hibiscus subset consists of 1,165 images categorized into eight distinct classes: Healthy (473 images), Mild Edge Damage (226 images), Citruspot (150 images), Slightly Diseased (109 images), Early Mild Spotting (83 images), Wrinkled (56 images), Senescent (40 images), and Fungal Infected (28 images). The Tea subset contains 248 images divided into five disease categories: Algal Leaf Spot (54 images), Brown Blight (48 images), Grey Blight (53 images), Healthy (49 images), and Red Leaf Spot (44 images). These class distributions capture a diverse range of leaf conditions, disease severities, and environmental variations, providing a realistic foundation for machine learning and deep learning applications. To overcome class imbalance and enrich the dataset, extensive image augmentation was performed using both PIL and OpenCV techniques, including brightness and contrast adjustment, color enhancement, rotation, flipping, scaling, cropping, shifting, zooming, and Gaussian noise addition. Through this process, each class was expanded to 1,000 images, resulting in a total of 13,000 augmented images evenly distributed across the 13 classes. All images are stored in .JPG format and organized into separate folders per class, maintaining a consistent structure and naming convention. Overall, this dataset offers a rich and diverse resource for developing robust models for leaf disease classification, precision agriculture, and automated plant health monitoring, making it a valuable contribution to the fields of computer vision and agricultural research.

木槿与茶叶联合图像数据集（Combined Hibiscus and Tea Leaf Image Dataset）是一套全面且均衡的数据集，包含1413张原始高质量叶片图像。这些图像于孟加拉国多个地区的自然户外环境下采集，采集设备为索尼α7 II数码单反相机（SONY α7 II DSLR camera）与一加7T智能手机（OnePlus 7T smartphone）。该数据集涵盖木槿与茶树两大植物品类，旨在为植物病害检测、农业图像分析以及基于计算机视觉的作物健康监测相关研究提供支持。其中木槿子集包含1165张图像，划分为8个独立类别：健康叶片（473张）、轻度叶缘损伤（226张）、柑桔斑病（150张）、轻度染病（109张）、早期轻度斑点症（83张）、皱叶症（56张）、衰老叶（40张）以及真菌感染病（28张）。茶叶子集包含248张图像，分为5类病害：藻斑病（54张）、褐色叶枯病（48张）、灰色叶枯病（53张）、健康叶片（49张）以及红斑叶病（44张）。各类别样本分布涵盖了多样化的叶片状态、病害严重程度与环境差异，为机器学习与深度学习应用提供了贴合实际的研究基础。为解决样本类别不平衡问题并丰富数据集规模，研究团队采用PIL库（PIL）与OpenCV库（OpenCV）开展了大规模图像增强操作，具体包括亮度与对比度调整、色彩增强、旋转、翻转、缩放、裁剪、平移、变焦以及添加高斯噪声。通过上述增强流程，每个类别的样本量均扩充至1000张，最终得到13000张增强后的图像，均匀分布于全部13个类别中。所有图像均以.JPG格式存储，并按类别分别存放于独立文件夹中，整体结构与命名规则保持统一。总体而言，该数据集为叶片病害分类、精准农业以及自动化植物健康监测领域的高性能模型开发提供了丰富多样的资源，为计算机视觉与农业研究领域贡献了极具价值的研究成果。