opendiffusionai/cc12m-2mp-squareish
收藏Hugging Face2025-01-09 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/opendiffusionai/cc12m-2mp-squareish
下载链接
链接失效反馈官方服务：
资源简介：
这是一个从 CC12m 数据集衍生出的子集，包含大约37k张分辨率在2mp到4mp之间的图片。图片宽高比主要在4:5到5:4之间，适合自动裁剪处理。数据集经过AI模型筛选去除了带水印和黑白图片，但未经过手工筛选，可能包含一些质量较低的照片。数据集提供长格式或短格式LLaVA标注。
This is a subset derived from the CC12m dataset, containing about 37k images with resolutions between 2mp and 4mp. The images mainly have aspect ratios ranging from 4:5 to 5:4, suitable for auto cropping. The dataset has been filtered using AI models to remove images with watermarks and black-and-white photos, but it has not been hand curated, so it may contain some lower-quality images. The dataset provides long-form or short-form LLaVA captions.
提供机构：
opendiffusionai
