opendiffusionai/cc12m-a_woman

该数据集是从opendiffusionai/cc12m-cleaned中提取的一个便利子集，通过搜索A woman并手动筛选得到。筛选标准包括去除水印、网站品牌标识等可能影响机器学习训练的元素，并选择主焦点清晰、锐利的高质量图像。目前数据集包含几千张图像，目标是达到10万张。获取图像的步骤包括下载.gz文件和crawl.sh脚本，并安装img2dataset工具来运行脚本。数据集也可以以parquet格式获取。

This dataset is a convenience subset of opendiffusionai/cc12m-cleaned, created by searching for A woman and hand-curating the results. The curation process involved removing watermarks, site branding, and other elements that could interfere with ML training, and selecting only images with clear, sharp camera focus on the main subject. Currently, the dataset contains a few thousand images, with a goal of reaching 100,000. To acquire the images, download both the .gz file and the crawl.sh file, and install the img2dataset utility to run the script. The dataset is also available in parquet format.