opendiffusionai/laion2b-squareish-1024px

这是一个laion2b-en-aesthetic数据集的子集，包含大约25万张真实世界的摄影图片，用于模型训练。所有图片的分辨率至少为1024像素高，纵横比在1到5/4之间，可以安全地自动裁剪成正方形。数据集提供了两种风格的自动生成标题。总磁盘大小约为118G。

This dataset is a subset of the laion2b-en-aesthetic dataset, containing about 250,000 real-world photographs for model training. All images are at least 1024 pixels tall with an aspect ratio between 1 and 5/4, allowing for safe auto-cropping to squares. The dataset provides two styles of automatically generated captions and has a total disk size of approximately 118G.