WikiImages

该数据集是从WebQA提供的多模态语料库构建而成的，其中包含了从维基媒体共享中抓取的图像-文本对。为了提高数据质量，我们还进行了筛选以去除噪声数据。该数据集的任务是对文本生成图像的能力进行评估。

This dataset is constructed from the multimodal corpus provided by WebQA, which contains image-text pairs crawled from Wikimedia Commons. To improve data quality, we conducted filtering to remove noisy data. The task of this dataset is to evaluate the capability of text-to-image generation.