GreenPulse-T: A Comprehensive Tea Leaf Image Dataset Featuring Field-Captured Samples Across Different Health and Disease Classes

The GreenPulse-T dataset consists of high-quality images of tea leaves collected from two prominent tea estates in Sreemangal, Moulvibazar, Bangladesh: M.R. Khan Tea Estate (coordinates: 24.27257, 91.75938) and Finlay Tea Estate (coordinates: 24.30334, 91.74249). Data collection took place over the course of eight days, from December 9 to December 16, 2024. The estates were carefully selected to represent diverse cultivation environments, ensuring the dataset includes a wide variety of tea leaf conditions observed in natural settings.

Images in the dataset depict both healthy and diseased tea leaves captured under real-world field conditions. Photographs were taken at different times of day, resulting in variations in lighting, background, and leaf orientation. This variability enhances the dataset’s practical value, making it suitable for robust machine learning and computer vision applications such as classification, object detection, and disease segmentation. The images were compressed to 480x640 pixels at 96 DPI to make them more accessible and manageable for researchers while preserving the key details necessary for disease detection and analysis.

The dataset is categorized into six classes based on expert assessment: Algal Leaf (301 images), Grey Blight (302 images), Healthy Leaf (435 images), Helopeltis (321 images), Looper Infested (332 images), and Red Spider (316 images), totaling 2007 images. Each image is stored in JPG format and organized into folders according to its respective class label.

This dataset is intended to support researchers, developers, and practitioners working in the fields of plant pathology, agriculture, and artificial intelligence. It can be used for academic research, the development of machine learning models, mobile application development for field diagnosis, and other agricultural technology innovations. Researchers using this dataset are kindly requested to cite it appropriately to acknowledge the effort invested in its collection and curation.

GreenPulse-T数据集包含高质量茶叶图像，采集自孟加拉国毛尔维巴扎尔县斯里曼加尔的两座知名茶园：M.R. Khan茶园（坐标：24.27257, 91.75938）与Finlay茶园（坐标：24.30334, 91.74249）。数据采集工作于2024年12月9日至16日开展，共计8天。两座茶园经过精心遴选，以覆盖多样化的种植环境，确保数据集涵盖自然场景下观测到的各类茶叶状态。
本数据集内的图像均拍摄于真实田间环境，涵盖健康与染病茶叶样本。拍摄时段覆盖全天不同时刻，因此图像存在光照、背景与叶片朝向的差异。这种多样性提升了数据集的实用价值，使其适用于鲁棒性较强的机器学习与计算机视觉任务，例如分类、目标检测以及病害分割。为便于研究人员使用与管理，所有图像均被压缩至480×640像素、分辨率96 DPI，同时保留了病害检测与分析所需的关键细节。
本数据集经专家评估后分为6个类别：藻斑病（Algal Leaf）样本301张、灰枯病（Grey Blight）样本302张、健康叶片（Healthy Leaf）样本435张、茶盲蝽为害叶（Helopeltis）样本321张、尺蠖为害叶（Looper Infested）样本332张以及红蜘蛛为害叶（Red Spider）样本316张，总计2007张图像。所有图像均以JPG格式存储，并按照对应的类别标签分文件夹整理。
本数据集旨在为植物病理学、农业以及人工智能领域的研究人员、开发者与从业者提供支持，可应用于学术研究、机器学习模型开发、田间诊断移动应用开发以及其他农业技术创新项目。使用本数据集的研究人员请适当引用该数据集，以肯定其采集与整理工作付出的心血。

搜集汇总

背景与挑战

背景概述

GreenPulse-T是一个包含2007张茶叶图像的数据集，涵盖六种不同健康状态和病害类别，图像在自然环境下拍摄，适用于机器学习和计算机视觉应用。数据集来自孟加拉国两个茶园，图像分辨率为480x640像素，存储为JPG格式。