WM33333/NCS-STR
收藏资源简介：
NCS-STR数据集包含两个子集：hands_on和advanced。每个示例包括一个打乱的图像序列、一个注释以及正确的顺序。该数据集用于视觉问答和图像到文本任务，语言为英语。字段包括：video_id（从原始vid路径提取的文件夹ID）、annotation（任务注释）、image_filenames（打乱的帧文件名）、images（打乱的图像序列，以Parquet格式存储为图像字节）、answer_order（从JSON文件保留的正确答案顺序）和subset（标识为hands_on或advanced）。
NCS-STR contains two subsets: hands_on and advanced. Each example contains a shuffled image sequence, an annotation, and the correct order. The dataset is used for visual-question-answering and image-to-text tasks, in English. Fields include: video_id (folder id extracted from the original vid path), annotation (task annotation), image_filenames (shuffled frame filenames), images (shuffled image sequence, stored as image bytes in Parquet), answer_order (correct order, preserved exactly from the JSON files), and subset (either hands_on or advanced).
