Tea Leaf Diseases Dataset: Towards Accurate Field Diagnosis Using Image-Based Detection

The dataset comprises a total of 31,265 images, including 4,009 raw images and 27,256 augmented images, distributed across six categories. The raw dataset captures natural variability in leaf conditions under diverse environmental settings, while the augmented dataset enhances this variability by applying techniques such as rotation, scaling, flipping, and brightness adjustments to improve machine learning model generalizability.

Category Distribution:

Sunlight Scorching:
Raw: 1,712 images
Augmented: 4,009 images
Total: 5,721 images

Red Spider:
Raw: 410 images
Augmented: 4,746 images
Total: 5,156 images

Red Rust:
Raw: 184 images
Augmented: 4,849 images
Total: 5,033 images

Heliopeltis:
Raw: 281 images
Augmented: 4,825 images
Total: 5,106 images

Thrips:
Raw: 176 images
Augmented: 4,881 images
Total: 5,057 images

Normal (Healthy):
Raw: 1,246 images
Augmented: 3,946 images
Total: 5,192 images

Key Features:
Diversity: The dataset captures variability in backgrounds, lighting, and growth stages, ensuring robust training and testing data.
Augmentation: Includes 27,256 augmented images, generated using state-of-the-art techniques to simulate real-world variability.
Category Coverage: Spans six distinct leaf conditions—ranging from pest-related damage (e.g., Red Spider, Heliopeltis) to environmental stress (e.g., Sunlight Scorching) and healthy leaves.

Purpose:
This dataset is designed to aid in the development of machine learning models capable of:
Automated Leaf Condition Classification: Identifying specific conditions for improved crop health monitoring.
Ecological Research: Supporting biodiversity studies by cataloging leaf conditions.
Agricultural Applications: Enabling early detection of plant stressors, pests, and diseases to enhance sustainable farming practices.
This combined raw and augmented dataset serves as a valuable resource for researchers, ecologists, and machine learning practitioners, contributing to advancements in agriculture, botany, and AI-driven ecological monitoring.

本数据集共计包含31,265张图像，其中原始图像4,009张，增强图像27,256张，并分布于六大类别之中。原始数据集捕捉了叶片在多种环境条件下的自然变化，而增强数据集通过旋转、缩放、翻转和亮度调整等手段进一步增强了这种变化，以提升机器学习模型的泛化能力。

类别分布如下：

阳光灼烧：原始图像1,712张，增强图像4,009张，总计5,721张。
红蜘蛛：原始图像410张，增强图像4,746张，总计5,156张。
红锈：原始图像184张，增强图像4,849张，总计5,033张。
日轮叶点：原始图像281张，增强图像4,825张，总计5,106张。
蓟马：原始图像176张，增强图像4,881张，总计5,057张。
正常（健康）：原始图像1,246张，增强图像3,946张，总计5,192张。

关键特性：
多样性：数据集涵盖了背景、光照和生长阶段的多样性，确保了稳健的训练和测试数据。
增强：包含27,256张增强图像，利用最先进的技术生成，以模拟现实世界的多样性。
类别覆盖：涵盖了六种不同的叶片状况——从病虫害相关损伤（如红蜘蛛、日轮叶点）到环境压力（如阳光灼烧）以及健康叶片。

目的：本数据集旨在协助开发机器学习模型，使其能够：
自动化叶片状况分类：识别特定状况，以提升作物健康监测。
生态研究：支持生物多样性研究，通过编制叶片状况目录。
农业应用：实现植物压力源、病虫害的早期检测，以增强可持续农业实践。
该结合原始和增强的数据集为研究人员、生态学家和机器学习从业者提供了一个宝贵的资源，有助于农业、植物学和AI驱动的生态监测领域的进步。