Symptom-Labeled Image Dataset of Rice Plants for Stem Borer Infestation Classification

Description: The Rice Stem Borer Infestation Dataset is an extensive image dataset of paddy plants with symptoms for the purposes of research in precision agriculture and AI based plant health monitoring. It emphasizes on three classes based on the symptoms: Healthy, Dead Heart, and White Head, corresponding to normal growth and two key damage stages caused by the Yellow Stem Borer (Scirpophaga incertulas) and Striped Stem Borer (Chilo suppressalis). The dataset consists of 2,096 high-resolution RGB images originally taken in the field, under natural lighting, across different crop fields in Bangladesh, and captured with a 64MP Realme 6 smartphone. Photos were captured under different weather, lighting, and time conditions to mimic actual environments. A certified agricultural expert confirmed the accuracy of symptom classification for each image. In order to enhance the variability and robustness of the dataset for machine learning purposes, a total of 14,672 augmented images were generated through transformations that include rotation, flipping, zoom, addition of noise, brightness, shearing, and height shifting, providing a total of images of 16,768. Dataset Details: 1. Original Dataset: Number of Images: 2,096 high-resolution RGB images ( .jpg format ) Classes: • Healthy: 1106 images • Dead Heart :439 images • White Head: 551 images 2. Augmented Dataset: Number of Augmented Images: 14,672 high-resolution RGB images ( .jpg format ) Classes: • Healthy: 7742 images • Dead Heart: 3073 images • White Head: 3857 images Use a data augmentation library or tool (e.g. TensorFlow ImageDataGenerator, OpenCV) to apply transformations. Define the Parameters Used for Data Augmentation: Horizontal Flip: Image flipped left-to-right. Rotation: Rotated by 45 degrees. Zoom: Random zoom between 0.8x and 1.2x. Add Gaussian Noise: Random noise with mean=0, std=25. Height Shift: Shift image vertically by 20% of height. Brightness Adjustment: Random brightness between 0.5x and 1.5x. Shearing: Shear (distort) the image with shear factor up to 0.5. Capture Details: • Device: Realme 6 (64MP camera) • Resolution: 475×635 pixels • Collection Period: January–March 2025 • Environmental Diversity: Varied weather (sunny, cloudy, windy), times (morning, noon, afternoon) Data Collection Location: • Satrujitpur, Magura Sadar, Khulna, Bangladesh (Latitude: 23°25'07.0"N, Longitude: 89°29'20.0"E) • Khajura, Bagherpara Upazila, Jessore, Khulana, Bangladesh (Latitude: 23.2764° N, Longitude: 89.2538° E) Purpose: The purpose of this dataset is to help automated, rapid identification of the level of infestation from images of rice stem borers through computer vision and artificial intelligence. It helps improve pest detection systems, reduces crop loss to pests by lessening reliance on pesticide use, and provides and encourages development of strong agricultural monitoring tools.### 数据集描述 稻螟虫害数据集（Rice Stem Borer Infestation Dataset）是一款面向精准农业与基于人工智能的作物健康监测研究的大型水稻植株症状图像数据集。该数据集基于受害症状划分为三大类别：健康（Healthy）、枯心苗（Dead Heart）与白穗（White Head），分别对应水稻正常生长状态，以及由二化螟（Yellow Stem Borer，Scirpophaga incertulas）和三化螟（Striped Stem Borer，Chilo suppressalis）引发的两个关键受害阶段。 该数据集包含2096张原始田间高分辨率RGB图像，均于孟加拉国不同稻田中在自然光照条件下拍摄，采集设备为6400万像素的Realme 6智能手机。为模拟真实田间环境，拍摄时涵盖了不同天气、光照与时段条件。 每一张图像的症状分类均经过认证农业专家确认其准确性。为提升数据集用于机器学习任务时的多样性与鲁棒性，研究人员通过旋转、翻转、缩放、添加噪声、亮度调整、剪切与高度偏移等变换，共生成14672张增强图像，使总样本量达到16768张。 ### 数据集详情 1. 原始数据集： 图像数量：2096张高分辨率RGB图像（格式为.jpg） 类别分布： • 健康：1106张 • 枯心苗：439张 • 白穗：551张 2. 增强数据集： 增强图像数量：14672张高分辨率RGB图像（格式为.jpg） 类别分布： • 健康：7742张 • 枯心苗：3073张 • 白穗：3857张 可通过数据增强库或工具（如TensorFlow ImageDataGenerator、OpenCV）执行上述变换。本次数据增强所用参数如下： 水平翻转：图像沿左右方向翻转 旋转：以45度为间隔进行旋转 缩放：随机缩放比例介于0.8倍至1.2倍之间 添加高斯噪声：添加均值为0、标准差为25的随机噪声 高度偏移：沿垂直方向偏移图像高度的20% 亮度调整：随机调整亮度至原亮度的0.5倍至1.5倍 剪切变换：以最大0.5的剪切因子对图像进行扭曲操作 ### 采集详情 • 采集设备：Realme 6（6400万像素摄像头） • 图像分辨率：475×635像素 • 采集时段：2025年1月至3月 • 环境多样性：涵盖不同天气（晴天、阴天、大风天）与不同时段（早晨、正午、下午） ### 数据采集地点 • 孟加拉国库尔纳专区马古拉县萨达尔乡Satrujitpur（纬度：23°25'07.0"N，经度：89°29'20.0"E） • 孟加拉国库尔纳专区杰索尔县巴格赫拉帕拉乌帕齐拉Khajura（纬度：23.2764° N，经度：89.2538° E） ### 数据集用途 本数据集旨在通过计算机视觉与人工智能技术，实现基于图像的稻螟虫害程度自动化快速识别。其有助于优化虫害检测系统，通过降低农药依赖度减少虫害引发的作物损失，并推动高性能农业监测工具的研发与应用。