gmongaras/CC12M_and_Imagenet21K_Recap
收藏Hugging Face2025-09-17 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/gmongaras/CC12M_and_Imagenet21K_Recap
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了大约1300万张图片和大约1.9万个类别的完整21K ImageNet数据集，以及整个CC12M数据集的再配字。ImageNet数据集的图片和类别字符串数量为19K，而不是21K。CC12M数据集于2021年创建，包含1200万张图片。数据集中的图片均为PNG格式，可以通过提供的代码示例进行解码。数据集特征包括图片、类别标签、ID以及原始和缩短的再配字。
This dataset includes the entire 21K ImageNet dataset with about 13 million images and 19 thousand classes as strings, as well as the entire CC12M dataset with recaptioned images. The ImageNet dataset has around 19K classes instead of the expected 21K. The CC12M dataset consists of 12 million images created in 2021. All images in the dataset are in PNG format and can be decoded using the provided code example. The dataset features include images, class labels, IDs, and both original and shortened recaptioned text.
提供机构：
gmongaras
