Tea Leaf Disease Image Dataset for Automated Classification of Tea Garden Diseases

This dataset contains 2,500 high-quality images of tea leaves collected under natural field conditions in tea gardens across Habiganj, Sylhet, Bangladesh. It is designed to support machine learning and computer vision research for the automatic detection and classification of tea leaf diseases, contributing to innovations in precision agriculture and AI-based crop monitoring. Data Collection Details: ⦁ Collection Period: April 7, 2025 – June 23, 2025 ⦁ Total Duration: Approximately 2.5 months ⦁ Collection Locations: Chāndpur Tea Garden, Deundi Tea Garden, Nalua Tea Estate, and Amo Tea Estate (Chunarughat, Habiganj, Sylhet, Bangladesh) ⦁ Environmental Conditions: Images were captured outdoors under natural daylight, during various weather conditions (sunny, cloudy, and humid) to ensure dataset diversity. ⦁ Equipment Used: All photographs were taken using an iPhone 12 Pro Max, featuring a 12 MP triple-camera system (Ultra Wide, Wide, and Telephoto) with image resolutions ranging from 1080×1080 to 3024×3024 pixels. ⦁ Data Validation: All images were manually reviewed, cleaned, and labeled by agricultural experts and plant pathologists to ensure class consistency and labeling accuracy. Dataset Composition: The dataset is organized into four distinct classes based on disease symptoms and healthy leaf conditions. ⦁ Blight (555 images): Leaves showing symptoms of fungal blight infection such as brown lesions and decayed edges. ⦁ Healthy Leaf (650 images): Fresh, disease-free green tea leaves. ⦁ Red Rust (620 images): Leaves infected with red-orange fungal rust spots. ⦁ Helopeltis (675 images): Leaves damaged by the Helopeltis insect pest, characterized by small dark punctures and yellowing areas around bites. Annotation and Preprocessing: ⦁ All images were manually inspected and cleaned to remove duplicates, blurred samples, and inconsistent labels. ⦁ Labeling was carried out under expert supervision, following standard plant disease identification guidelines. ⦁ No artificial data augmentation was applied, preserving real-world variability in lighting, leaf texture, and background conditions. Applications: This dataset can be effectively used for: ⦁ Image classification and disease recognition tasks ⦁ Deep learning model development (CNNs, transfer learning, etc.) ⦁ Comparative benchmarking for plant pathology research ⦁ AI-based agricultural monitoring and tea yield optimization File Information ⦁ Total Images: 2,500 ⦁ File Format: JPEG (.jpg) ⦁ Resolution Range: 1080×1080 – 3024×3024 pixels ⦁ Average File Size: 0.67 – 7.8 MB per image ⦁ Folder Structure: Each class stored in a separate labeled directory本数据集包含2500张高质量茶叶图像，采集自孟加拉国锡尔赫特市哈比甘杰地区各茶园的自然田间环境。其旨在支撑机器学习（Machine Learning）与计算机视觉（Computer Vision）领域的研究，用于茶叶病害的自动检测与分类，助力精准农业（Precision Agriculture）与基于人工智能的作物监测技术创新。 数据采集详情： ⦁ 采集周期：2025年4月7日 – 2025年6月23日 ⦁ 总时长：约2.5个月 ⦁ 采集地点：孟加拉国锡尔赫特市哈比甘杰地区楚纳鲁哈特的昌德普尔茶园、当迪茶园、纳卢阿茶叶庄园与阿莫茶叶庄园 ⦁ 环境条件：所有图像均在室外自然光下拍摄，涵盖晴天、阴天与潮湿等多种天气条件，以保障数据集的多样性。 ⦁ 采集设备：所有照片均采用iPhone 12 Pro Max拍摄，该设备搭载1200万像素三摄系统（超广角、广角与长焦镜头），图像分辨率范围为1080×1080至3024×3024像素。 ⦁ 数据验证：所有图像均由农业专家与植物病理学家进行人工审核、清洗与标注，以确保类别一致性与标注准确性。 数据集构成： 本数据集根据病害症状与健康叶片状态分为4个明确类别： ⦁ 茶枯病（Blight，555张）：显示真菌性枯病感染症状的叶片，表现为褐色病斑与叶缘腐烂。 ⦁ 健康叶片（Healthy Leaf，650张）：新鲜无病害的绿色茶叶叶片。 ⦁ 茶红锈病（Red Rust，620张）：感染红橙色真菌锈斑的叶片。 ⦁ 茶盲蝽虫害（Helopeltis，675张）：受该虫害破坏的叶片，特征为细小深色穿刺痕迹与叮咬周围的泛黄区域。 标注与预处理： ⦁ 所有图像均经过人工检查与清洗，以移除重复图像、模糊样本与标注不一致的样本。 ⦁ 标注工作在专家监督下开展，遵循标准植物病害识别规范。 ⦁ 未采用人工数据增强手段，保留了光照、叶片纹理与背景条件的真实世界变异性。 应用场景： 本数据集可有效应用于以下场景： ⦁ 图像分类与病害识别任务 ⦁ 深度学习模型开发（卷积神经网络（Convolutional Neural Network，CNNs）、迁移学习（Transfer Learning）等） ⦁ 植物病理学研究的对比基准测试 ⦁ 基于人工智能的农业监测与茶叶产量优化 文件信息： ⦁ 总图像数：2500张 ⦁ 文件格式：JPEG（.jpg） ⦁ 分辨率范围：1080×1080 – 3024×3024像素 ⦁ 单张图像平均文件大小：0.67 – 7.8 MB ⦁ 文件夹结构：每个类别均存储于独立的标注目录中