An Extensive Image Dataset for Classifying Rice Varieties in Bangladesh

This dataset contains a carefully collected set of low-resolution images of 38 well-known rice varieties from BINA and IRRI. It helps in identifying the unique features of these rice types for accurate classification. #Dataset Composition# The dataset includes images of 38 different types of rice, specifically: BD33, BD30, BD39, BD56, BD93, BD91, BD49, BD51, BD52, BD76, BD95, BD57, BD87, BD70, BD85, BD72, BD79, BD75, Binadhan7, Binadhan8, Binadhan10, Binadhan11, Binadhan12, Binadhan14, Binadhan16, Binadhan17, Binadhan19, Binadhan20, Binadhan21, Binadhan23, Binadhan24, Binadhan25, Binadhan26, BR22, BR23, BRRI67, BRRI74, and BRRI102. There are 19,000 original JPG images and 76,000 augmented images in total. #Image Capture and Dataset Organization# Images were taken using high-power 1600x and 1000x digital microscope cameras between January 15 and February 28, 2024. The dataset is divided into two main parts: Original images and Augmented images. Each part has 38 folders, one for each rice variety. #Original Image Dataset# The main set includes 19,000 JPG images, each sized at 640x480 pixels. The uncompressed file size is 1.52 GB, reduced to 1.47 GB after compression. #Augmented Image Dataset# To increase the number of images for deep learning models, data augmentation techniques were used. These include rotations (90° left, 90° right, 180°), shear range and flips, creating 76,000 more images. These images are also in JPG format, sized at 640x480 pixels, and were initially 1.98 GB, reduced to 1.89 GB after compression. #Dataset Storage and Access# The raw and augmented datasets are stored in two separate zip files, 'Original.zip' and 'Augmented.zip'. Each zip file contains 38 folders, one for each rice variety mentioned above.

本数据集精心收集了来自孟加拉国核农业研究所（BINA）与国际水稻研究所（IRRI）的38个知名水稻品种的低分辨率图像，可用于识别各水稻品种的独有特征，从而实现精准分类。#数据集构成# 本数据集涵盖38种不同水稻的图像，具体包括：BD33、BD30、BD39、BD56、BD93、BD91、BD49、BD51、BD52、BD76、BD95、BD57、BD87、BD70、BD85、BD72、BD79、BD75、Binadhan7、Binadhan8、Binadhan10、Binadhan11、Binadhan12、Binadhan14、Binadhan16、Binadhan17、Binadhan19、Binadhan20、Binadhan21、Binadhan23、Binadhan24、Binadhan25、Binadhan26、BR22、BR23、BRRI67、BRRI74及BRRI102。总计包含19000张原始JPG图像与76000张增强图像。#图像采集与数据集组织# 图像采集于2024年1月15日至2月28日期间，使用高倍1600x与1000x数码显微镜相机拍摄。本数据集分为两大模块：原始图像集与增强图像集，每个模块均包含38个文件夹，分别对应上述38个水稻品种。#原始图像数据集# 原始图像集共计19000张JPG图像，单张分辨率为640×480像素。未压缩总容量为1.52 GB，压缩后容量降至1.47 GB。#增强图像数据集# 为扩充深度学习模型训练所需的图像数据量，本数据集采用了数据增强技术，具体包括旋转（左旋90°、右旋90°、180°旋转）、剪切变换与翻转操作，最终生成76000张增强图像。该增强图像同样采用JPG格式，分辨率为640×480像素，未压缩总容量为1.98 GB，压缩后容量降至1.89 GB。#数据集存储与获取# 原始数据集与增强数据集分别存储于两个独立的压缩包中，即'Original.zip'与'Augmented.zip'，每个压缩包均包含38个文件夹，分别对应前文提及的38个水稻品种。

搜集汇总

背景与挑战

背景概述

该数据集是一个用于孟加拉国水稻品种分类的广泛图像数据集，包含38种水稻品种的19,000张原始图像和76,000张增强图像，所有图像均为640x480像素的低分辨率JPG格式。数据采集于2024年初，通过高功率显微镜相机拍摄，并采用数据增强技术扩展图像数量，旨在帮助识别水稻的独特特征以实现准确分类。数据集分为原始和增强两部分，便于存储和访问，适用于深度学习模型训练。