bloom-vist

该数据集包含了一系列有序的图像和标题序列，这些序列被组织成短故事，利用了Bloom图书馆的连续图像收藏，以支持多种语言的理解。该数据集遵循创意共享许可协议发布，并包含了一个用于数据质量的手动检查流程。规模上，该数据集共有11,407个故事，包含112,080对图像和标题。其任务领域涉及多模态故事讲述和图像标题生成。

This dataset consists of ordered image-caption sequences structured into short stories, which draws on the continuous image collection of the Bloom Library to enable multilingual understanding. Released under a Creative Commons license, this dataset incorporates a manual inspection procedure for data quality control. In terms of scale, the dataset encompasses 11,407 stories and 112,080 image-caption pairs in total. Its application domains cover multimodal storytelling and image caption generation.