DeepFruit 水果图像分类数据集

DeepFruit 是穆罕默德·本·法赫德亲王大学等研究机构联合发布的水果图像分类数据集。该数据集包含基于 8 种不同水果集组合的 20 种不同水果的 21,122 张水果图像。每个类别的图像数量因类别而异。每种水果以 4 种或 5 种不同的组合出现，图像采集自不同尺寸、形状和颜色的盘子，拍摄角度、亮度水平和距离各异。还附上了所有水果组合的示例图像。训练和测试 CSV 文件根据图像文件名包含每个图像中每个相应水果类的标签。

DeepFruit is a fruit image classification dataset jointly released by research institutions including Prince Mohammad bin Fahd University. This dataset comprises 21,122 fruit images spanning 20 distinct fruit species, which are structured across 8 unique fruit set combinations. The number of images per class varies across categories. Each fruit species appears in 4 or 5 distinct combinations. All images were captured using plates of varying sizes, shapes and colors, with different shooting angles, brightness levels and distances. Sample images for all fruit set combinations are also provided. The training and test CSV files contain labels for each corresponding fruit class in every image, linked via their filenames.

搜集汇总

背景与挑战

背景概述

DeepFruit是一个水果图像分类数据集，包含21,122张图像，涵盖20种不同水果，基于8种水果集组合，图像采集自智能手机并经过预处理以确保一致性。数据集已划分为80%训练集和20%测试集，适用于水果检测、识别、分类及卡路里估算等研究应用。