yzwang/X2I-mm-instruction

X2I数据集是一个大规模的统一图像生成数据集，旨在支持从任何内容到图像的生成任务。该数据集首次以统一格式提供，用于训练具有强大多任务处理能力的OmniGen模型。数据集包含多个子任务，如多模态指令、主题驱动编辑、上下文学习、计算机视觉和文本到图像生成等。每个子数据集都有详细的样本数量和操作指南。

The X2I dataset is a large-scale unified image generation dataset designed to support multi-task processing capabilities, particularly in unified image generation tasks. This dataset includes multiple sub-datasets, such as multi-modal instruction, subject-driven editing, in-context learning, computer vision, and text-to-image generation. Each sub-dataset has its specific task and number of samples, for example, the FashionTryOn dataset contains 41,004 samples, the HR-VITON dataset contains 13,679 samples, the MagicBrush dataset contains 8,807 samples, the InstructPix2Pix dataset contains 1,000,032 samples, the SomethingSomethingv2 dataset contains 168,913 samples, and the StyleBooth dataset contains 11,325 and 14,766 samples.