TIGER-Lab/Vript

这个数据集包含视频文件，每个视频被分割成帧序列。数据集中的特征包括视频文件名、帧序列、统一采样的索引、针对问答任务的帧索引以及问答样本的类型。数据集目前只有一个训练集划分，包含大约10935个示例，总大小约为52347亿字节。This dataset consists of videos, each divided into a sequence of frames. The features in the dataset include video filenames, frame sequences, uniformly sampled indices, frame indices for question-answering tasks, and the type of question-answering samples. Currently, the dataset has only one training split, containing about 10,935 examples, with a total size of approximately 52,347 billion bytes.