ciFAIR-10, CUB, ISIC 2018, EuroSAT, CLaMM

本研究涉及五个不同的数据集，用于图像分类任务。ciFAIR-10数据集包含10个类别的日常物品低分辨率图像；CUB数据集是200种鸟类的细粒度分类数据集；ISIC 2018数据集包含7种皮肤病的医学图像；EuroSAT数据集是基于Sentinel-2卫星图像的多光谱数据集，包含10种土地覆盖类别；CLaMM数据集用于中世纪手稿的分类，包含12种手稿风格。这些数据集涵盖了从自然图像到专业领域图像的多样化应用，旨在解决小样本学习问题，特别是在数据稀缺情况下的图像分类挑战。

This study utilizes five distinct datasets for image classification tasks. The ciFAIR-10 dataset comprises low-resolution images of daily objects belonging to 10 categories; the CUB dataset is a fine-grained classification dataset covering 200 bird species; the ISIC 2018 dataset contains medical images of 7 types of skin disorders; the EuroSAT dataset is a multispectral dataset based on Sentinel-2 satellite imagery, which includes 10 land cover categories; the CLaMM dataset is intended for medieval manuscript classification, encompassing 12 manuscript styles. These datasets span diverse applications from natural images to professional domain images, and are designed to address few-shot learning problems, particularly the image classification challenges under data-scarce scenarios.