Rapidata/sora-video-generation-physics-likert-scoring

这是一个名为t2v Sora Style Likert Scores的视频生成物理数据集。数据集包含了约6000名人类评估者对AI生成的视频进行评分，评分基于视频中的重力、运动、碰撞和交互是否具有物理意义。评估者使用了李氏量表进行评分，分数越低表示视频的物理表现越好。数据集的特征包括提示、视频、李氏量表分数、详细结果等，详细结果中包含了用户的选择类别和用户详情信息。数据集分为训练集，大小为507060字节，共有198个示例。数据集的许可为Apache-2.0。

This is the t2v Sora Style Likert Scores video generation physics dataset. It includes ratings from approximately 6000 human evaluators on AI-generated videos, scored based on the physical sense of gravity, movement, collision, and interaction within the videos. Evaluators used the Likert Scale for scoring, where lower scores indicate better physical performance. The dataset features include prompts, videos, Likert Scores, and detailed results with user selection categories and user details. The training split is 507060 bytes in size with 198 examples. The dataset is licensed under Apache-2.0.