LemonVarities: A comprehensive image dataset to classify rotten and multiple types fresh lemon

This dataset is prepared for publishing in Data in Brief in order to contribute to the AI and Agriculture research field. The dataset comprises 1,956 Original and 43,000 Augmented images. Two major classes, such as Fresh and Rotten Lemon available, where fresh lemon has 7different sub classes. The types of seven lemon are Meyer lemon, Eureka lemon, Citron lemon, Assam lemon, Desi kagzi type-1 and type-2 lemon, and Elachi lemon. The dataset is not balance which helps AI researchers to think differently, apply their idea and knowledge to overcome AI weaknesses. The pixel size of each image is 900*1200, which reflects the perfect quality to have detailed features about images.

本数据集专为投稿至《数据简报（Data in Brief）》而制备，旨在为人工智能（AI）与农业研究领域贡献相关研究数据。该数据集包含1956张原始图像与43000张增强图像，设有两大核心类别：完好柠檬与腐烂柠檬，其中完好柠檬下设7个子类别，分别为迈耶柠檬（Meyer lemon）、尤力克柠檬（Eureka lemon）、香柠檬（Citron lemon）、阿萨姆柠檬（Assam lemon）、德西卡齐1号与2号柠檬（Desi kagzi type-1 and type-2 lemon）以及伊拉奇柠檬（Elachi lemon）。本数据集存在类别不平衡问题，这可促使人工智能研究者跳出固有思维，运用自身的创意与专业知识来攻克人工智能模型的短板。每张图像的像素尺寸均为900×1200，可完整保留图像的细节特征，具备极佳的成像质量。