opendiffusionai/laion2b-mixed-1024px-human

这是一个由其他数据集选择性合并而成的数据集，主要包括以人类为中心的真实世界照片，图片数量接近47000张，大小约为32G。图片的比例是混合的，其中最小的高度至少为1024px。数据集的标签是由AI自动生成的，而不是由人工输入。此外，大约有四分之一的数据集图片经过了人工审核。

This dataset is a selective merge of other datasets, mainly consisting of real-world, human-centered photos, with nearly 47,000 images taking up about 32G of space. The images have a mixed aspect ratio, with the smallest height being at least 1024px. The captions are auto-generated by AI rather than entered by humans. Additionally, about one-fourth of the images in the dataset have been manually reviewed.