Mulberry Leaf Dataset

Data collection: We collected the mulberry leaf cultivars from three regions of Thailand (northern, central, and northeast) that included five provinces in total (Chiang Mai, Phitsanulok, Nakhon Ratchasima, Burriram, and Mahasarakham). DSLR and phone cameras were used to take images with different perspectives from ten mulberry leaf cultivars recorded in the natural environments with different perspectives. The mulberry leaf dataset includes 5,262 images of 10 mulberry leaf cultivars: King Red, King White, Taiwan Maechor, Taiwan Strawberry, Black Austurkey, Black Australia, Chiang Mai 60, Buriram 60, Kamphaeng Saen 42, and Mixed Chiang Mai 60+Buriram 60. Data description: The mulberry leaf dataset was curated in 2020 using digital single-lens reflex (DSLR) and smartphone cameras to collect 5,262 images categorized into ten classes existing in diverse environmental conditions. No seasonal considerations were factored in during the data collection; however, all data was collected on the sunny days. There is no existence of external plants in the samples of the mulberry leaves, but there are some natural backgrounds of the brown soil and the mulberry tree, which exist in small proportion relative to the mulberry leaves. The image format of the mulberry leaf dataset exists in JPEG format and has varying resolution sizes. The researcher captured mulberry leaf images from various regions and provinces in Thailand for eight months. The image captures of mulberry leaves were taken from five Thai areas, as shown in Fig. 1. Further, the dataset was annotated by a domain expert responsible for classifying each of the mulberry leaves into their respective classes or categories. The leaves with similar features or properties were stored in a specific folder (class), resulting in ten possible classes. Data source location: The mulberry leaf dataset was collected from three regions of Thailand: northern (Chiang Mai), central (Phitsanulok), and northest (Nakhon Ratchasima, Burriram, and Mahasarakham). Related research article: Chompookham, T. & Surinta, O. (2021). Ensemble methods with deep convolutional neural networks for plant leaf recognition. ICIC Express Letters, 15(6), 553-565. DOI: 10.24507/icicel.15.06.553

数据集采集：我们从泰国北部、中部及东北部三个区域的共计5个省份（清迈、彭世洛、呵叻、武里南及玛哈沙拉堪）采集桑树叶品种样本。采用数码单反相机（Digital Single-Lens Reflex, DSLR）与手机摄像头，在自然环境下针对10个桑树叶品种拍摄多视角图像。本桑树叶数据集共计包含5262张图像，涵盖10个桑树叶品种，分别为：红王（King Red）、白王（King White）、台湾Maechor、台湾草莓桑（Taiwan Strawberry）、黑奥斯特基（Black Austurkey）、澳大利亚黑桑（Black Australia）、清迈60号、武里南60号、甘烹盛42号以及混合品种清迈60号+武里南60号。

数据说明：本桑树叶数据集于2020年构建完成，通过数码单反相机及智能手机采集图像，共得到5262张图像，被划分为10个类别，涵盖多样化的环境条件。数据采集过程未纳入季节因素考量，但所有图像均拍摄于晴天。桑树叶样本中未混入其他外来植物，样本背景中存在少量棕褐色土壤与桑树本体的自然场景，占比低于桑树叶主体。本数据集的图像格式均为JPEG，分辨率尺寸各不相同。研究人员耗时8个月，从泰国各区域及省份采集桑树叶图像，拍摄地点涵盖5个泰国区域，详见图1。此外，数据集由领域专家完成标注，负责将每张桑树叶图像归类至对应类别。特征相似的样本被存放至同一文件夹（类别）中，最终形成10个类别。

数据源位置：本桑树叶数据集采集自泰国三个区域：北部（清迈）、中部（彭世洛），以及东北部（呵叻、武里南及玛哈沙拉堪）。

相关研究论文：Chompookham, T. & Surinta, O. (2021). 集成深度卷积神经网络的植物叶片识别方法. ICIC Express Letters, 15(6), 553-565. DOI: 10.24507/icicel.15.06.553

搜集汇总

背景与挑战

背景概述

该数据集包含5,262张泰国三个地区10个不同品种的桑叶图像，使用DSLR和智能手机相机在自然环境下拍摄，适用于图像分类和深度学习研究。