PAC

该数据集名为PAC，是一个包含著名人物在特定情境中的图片及其对应描述的图像-字幕数据集。它特别强调在图像描述中使用非通用词汇。此外，该数据集包含了62位知名人士的图片，涵盖了著名运动员和政治人物，并具有独特的标注说明，强调专有名词的使用。规模上，该数据集包含了1,572张图片，每张图片配有3条描述。其任务重点在于图像字幕生成，尤其关注人名和OCR令牌。

The dataset named PAC is an image-caption dataset that pairs images of famous individuals in specific contexts with their corresponding descriptive captions. It specifically highlights the employment of non-general vocabulary within the image captions. Additionally, the dataset encompasses images of 62 prominent figures, including renowned athletes and political personalities, and incorporates unique annotation specifications that emphasize the use of proper nouns. In terms of scale, the dataset contains 1,572 images, each accompanied by three descriptive captions. Its primary task focuses on image caption generation, with particular attention paid to person names and OCR tokens.