Image Dataset of Ten Durian Diseases Captured in Real-Field Conditions from a Family Orchard in Vinh Long, Vietnam

This dataset contains 4,000 raw JPEG images of durian leaves, stems, fruits, and roots affected by ten common durian diseases, captured in real-field conditions from a family-owned durian orchard and four neighboring farms in Vĩnh Long, Vietnam. The images are organized into ten folders, each corresponding to one disease class: Thrips disease, Stem blight, Canker disease, Pink disease, Stem cracking and gummosis, Mealybug infestation, Anthracnose, Fruit rot, Sooty mold, and Yellow leaf disease.

Each class includes approximately 400 images, photographed using an iPhone 14 under natural lighting and farm conditions to simulate how farmers typically capture images. As a result, the dataset exhibits realistic noise, such as blurry focus, uneven lighting, varied angles, and complex backgrounds.

After collection, images were cropped using Preview on macOS (MacBook with Apple M4 chips), and saved in both JPEG (original) and PNG (cropped) formats. Disease symptoms were verified by plant pathology experts, ensuring the reliability of the class labels.

This dataset is intended for use in machine learning and computer vision applications, particularly in plant disease classification, mobile-based diagnostics, and evaluation of models under noisy, real-world conditions.

本数据集收录4000张原始JPEG图像，拍摄对象为感染十种常见榴莲病害的榴莲叶、茎秆、果实及根系，采集自越南永隆省一处家庭自营的榴莲园及周边四家农场的实地场景。
图像被划分为10个文件夹，每个文件夹对应一类病害：蓟马病害、茎枯病、溃疡病、粉红病、茎裂流胶症、粉蚧侵染、炭疽病、果腐病、煤污病、黄化叶病。
每类病害约含400张图像，均使用iPhone 14在自然光及农场实地环境下拍摄，以模拟农户日常的图像采集流程。因此，该数据集带有真实场景下的典型噪声特征，包括对焦模糊、光照不均、拍摄角度多样及背景复杂等。
采集完成后，使用搭载Apple M4芯片的MacBook设备上的macOS系统自带预览工具对图像进行裁剪，并分别保存为原始JPEG格式与裁剪后的PNG格式。所有病害症状均经植物病理学专家核验，确保类别标签的可靠性。
本数据集旨在应用于机器学习与计算机视觉相关研究，尤其适用于植物病害分类、移动端病害诊断，以及针对真实噪声场景下的模型性能评估。