recursal/SuperWikiImage-7M

SuperWikiImages-7M数据集是一个包含约15TiB（约700万张）图像的大规模数据集，这些图像主要从维基媒体共享资源中抓取。数据集支持多种语言，适用于图像分类、图像到文本、文本到图像、图像到图像等多种任务。数据集由KaraKaraWitch创建，由Recursal.ai资助，并通过WebDataset格式提供使用示例。数据集的许可信息复杂，包含多种允许的许可证类型，并排除了某些黑名单许可证。数据集的管理者KaraKaraWitch在多个AI社区活跃，并鼓励用户推广该数据集。

The SuperWikiImages-7M dataset is a large-scale collection of approximately 15TiB (~7 million) images, primarily scraped from Wikimedia Commons. The dataset supports multiple languages and is suitable for various tasks such as image classification, image-to-text, text-to-image, and image-to-image. Created by KaraKaraWitch and funded by Recursal.ai, the dataset is provided in WebDataset format with usage examples. The licensing information is complex, including a variety of permitted licenses and excluding certain blacklisted licenses. The dataset curator, KaraKaraWitch, is active in several AI communities and encourages users to promote the dataset.