FruitVeg-81 Dataset 水果和蔬菜数据集

该杂货数据集已在MANGO（营养增强和食品意识的移动增强现实）项目中收集。它包含15737张图像（最长的一面调整为512px），总下载大小为3.2GB。 数据集由带有分层标签的水果和蔬菜项目组成。它的结构如下： 第一层描述了一般的食品种类（苹果，香蕉，...） 第二层收集外观相似的食品品种（红苹果，青苹果...） 不同品种（Golden Delicious，Granny Smith等）或包装类型（盒装，托盘等）之间的第三级目的地 这导致文件夹层次结构具有：53个粗分类（例如苹果），81个细分类（例如青苹果）和125个品种（例如Golden Delicious）。每个品种文件夹中的文件夹对应于相同水果或蔬菜的特定记录，另外还提供了一个标签，指示使用的智能手机（请参见表3）。因此，这些文件夹中包含的图像（每个记录4-5个）应被视为训练和测试的一个样本。 BibTeX reference for convenience: @INPROCEEDINGS{waltner17b, author={Georg Waltner and Michael Schwarz and Stefan Ladstätter and Anna Weber and Patrick Luley and Meinrad Lindschinger and Irene Schmid and Walter Scheitz and Horst Bischof and Lucas Paletta}, title={{Personalized Dietary Self-Management using Mobile Vision-based Assistance}}, booktitle={Proc. of ICIAP Workshop on Multimedia Assisted Dietary Management (MADIMA)}, year={2017}, }

This grocery dataset was collected within the MANGO (Mobile Augmented Reality for Nutrition Enhancement and Food Awareness) project. It contains 15,737 images (with the longest side resized to 512px), with a total download size of 3.2 GB. The dataset consists of fruit and vegetable items with hierarchical labels. Its structure is as follows:
The first level describes general food categories (e.g., apples, bananas, ...)
The second level groups food varieties with similar appearances (e.g., red apples, green apples, ...)
The third level covers different cultivars or packaging types (e.g., boxed, tray-packed, ...). This results in a folder hierarchy with 53 coarse categories (e.g., apples), 81 fine-grained categories (e.g., green apples), and 125 cultivars (e.g., 'Golden Delicious'). Each folder within a cultivar folder corresponds to a specific record of the same fruit or vegetable, with an additional label indicating the smartphone used (see Table 3). Therefore, the images contained in these folders (4-5 per record) should be treated as one sample for training and testing.

BibTeX reference for convenience:
@INPROCEEDINGS{waltner17b,
author={Georg Waltner and Michael Schwarz and Stefan Ladstätter and Anna Weber and Patrick Luley and Meinrad Lindschinger and Irene Schmid and Walter Scheitz and Horst Bischof and Lucas Paletta},
title={{Personalized Dietary Self-Management using Mobile Vision-based Assistance}},
booktitle={Proc. of ICIAP Workshop on Multimedia Assisted Dietary Management (MADIMA)},
year={2017},
}

搜集汇总

背景与挑战

背景概述

FruitVeg-81 Dataset是一个包含15737张水果和蔬菜图像的数据集，总大小为3.2GB，图像最长边调整为512px。数据集具有分层标签结构，包括53个粗分类、81个细分类和125个品种，适用于图像分类任务。