yzwang/X2I-computer-vision

X2I数据集是一个大规模的统一图像生成数据集，旨在训练OmniGen模型以实现多任务处理能力。该数据集首次以统一格式创建，包含多个子数据集，分别针对多模态指令、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任务。X2I-computer-vision子数据集包括ADE、Rain100H、LOL、MuLan、ReasonSeg、WeatherStream & GoPro等具体数据集，涵盖了图像分割、去雨、低光增强、多层图像编辑、图像推理分割、去雨、去雪和去模糊等多种计算机视觉任务。

The X2I dataset is a large-scale unified image generation dataset designed to train the OmniGen model for robust multi-task processing capabilities. This dataset is the first to be created in a unified format and includes multiple sub-datasets targeting tasks such as multi-modal instruction, subject-driven editing, in-context learning, computer vision, and text-to-image generation. The X2I-computer-vision sub-dataset includes specific datasets like ADE, Rain100H, LOL, MuLan, ReasonSeg, WeatherStream & GoPro, covering various computer vision tasks such as image segmentation, deraining, low-light enhancement, multi-layer image editing, image reasoning segmentation, deraining, desnowing, and deblurring.