yzwang/X2I-subject-driven

X2I数据集是一个大规模的统一图像生成数据集，首次为统一图像生成任务创建。该数据集旨在训练OmniGen模型，以实现强大的多任务处理能力。X2I数据集涵盖了多种任务类型，包括多模态指令、主题驱动编辑、上下文学习、计算机视觉和文本到图像生成。其中，X2I-subject-driven子数据集包括Web-Image、GRIT-Entity和GRIT-Entity-New，分别包含36,316、45,425、111,734、1,708,742和676,603个样本。这些数据集用于主题驱动编辑任务，并提供了详细的文件处理指令。

The X2I dataset is a large-scale unified image generation dataset designed to support various tasks including multi-modal instruction, subject-driven editing, in-context learning, computer vision, and text-to-image generation. The dataset includes multiple subsets such as X2I-mm-instruction, X2I-subject-driven, X2I-in-context-learning, X2I-computer-vision, and X2I-text-to-image, each with specific sample counts and file structures.