opendiffusionai/laion2b-23ish-woman-solo

这个数据集包含了大约15,000张图片，每张图片中都有一个女性，比例为2:3或稍宽。所有图片都经过人工审核，无可见水印，质量和清晰度良好，适合AI训练使用。图片的体型多样，从尺寸0到18都有。数据集提供了三种标题选择：质量较差的“替代文本”标题，使用“moondream”模型生成的自然语言摘要，以及使用wd-large-tagger-v3模型生成的标签风格标题。数据集来源于LAION-2b-aesthetic集，经过筛选和人工审核。婚纱图片在数据集中占比较高，使用时需要注意内容平衡。

This dataset contains approximately 15,000 images, each featuring a woman at an approximately 2:3 aspect ratio or slightly wider. All images have been curated by humans, free of visible watermarks, with good quality and focus suitable for AI training. The dataset includes a variety of body sizes from size 0 to 18. There are three types of captions available: the poorly quality alt text, a natural language summary generated by the moondream model, and a tagged style using the wd-large-tagger-v3 model. The dataset is an initial extraction from the LAION-2b-aesthetic set, filtered for size and aspect ratio, and manually curated. There is a high proportion of bridal images, so attention should be paid to content balance when using the dataset.