tieredImageNet

该数据集是从ILSVRC-12数据集中提取的一个更大的子集，名为tieredImageNet，包含了超过779,000张图像，分为608个类别。这些类别根据ImageNet的分类层次结构，每个类别都源自34个超类别中的一个。我们遵循了张等人在2020年以及姜等人在2021年的设置，将此数据集分为20/351、6/97和8/160个超类别/类别，分别用于训练、验证和测试。该数据集的规模为608个类别，图像数量超过779,000张，任务类型为少样本图像分类。

This dataset, named tieredImageNet, is a larger subset extracted from the ILSVRC-12 dataset, containing over 779,000 images divided into 608 categories. Each category belongs to one of 34 super-categories based on the ImageNet classification hierarchy. We follow the experimental settings proposed by Zhang et al. (2020) and Jiang et al. (2021), and split this dataset into training, validation, and test subsets, which respectively include 20 super-categories and 351 categories, 6 super-categories and 97 categories, and 8 super-categories and 160 categories. With a total of 608 categories and over 779,000 images, this dataset is designed for few-shot image classification tasks.