LVHaystack/LongVideoHaystack_tiny

该数据集包含了视频ID、视频剪辑ID、问题、选项、答案、帧索引和视频元数据等特征。视频元数据进一步包含了如视频时长、分辨率、帧率等信息。数据集分为三个部分：tiny、val和test，分别用于不同的训练和测试目的。每个部分包含了不同数量的视频和视频剪辑。此外，数据集还提供了关于视频、剪辑、帧和问答对的统计信息。

The dataset includes features such as video ID, video clip ID, question, options, answer, frame indexes, and video metadata. The video metadata further includes details like video duration, resolution, frame rate, etc. The dataset is split into three parts: tiny, val, and test, each serving different purposes for training and testing. Each part contains a different number of videos and video clips. Additionally, the dataset provides statistical information about videos, clips, frames, and QA pairs.