AIML-TUDA/t2i-diversity-captions

该数据集包含超过3900万个为100万张精选图像生成的合成图像标题，这些图像来自LAION美学。图像具有美学评分>6，最低分辨率为512p，并经过NSFW、CSAM和水印的筛选。我们还移除了完全相同的重复项。数据集结构包括样本的唯一标识符、图像URL和最多10个合成标题。目前，大约74%的图像URL仍然可以访问。

This dataset consists of over 39 million synthetic image captions generated for 1 million curated images from LAION Aesthetics. Images have an aesthetics score >6, at a minimum resolution of 512p, and have been screened for NSFW, CSAM, and watermarks. We also removed exact duplicates. The data structure includes unique identifiers for samples, image URLs, and up to 10 synthetic captions. As of June 2025, approximately 74% of the image URLs were still accessible.