AI-generated video dataset
收藏资源简介:
该数据集是由中国科学院计算技术研究所构建的,包含由两种最先进的开源视频生成模型生成的13000个视频,旨在评估AI生成内容对视频检索模型的影响。数据集包括9000个训练视频和四个各含1000个视频的测试集。这些视频是基于MSR-VTT数据集生成的,并且通过多种策略确保生成的视频在语义上与原始视频相同。该数据集用于研究AI生成视频在检索模型中的排名偏置问题。
This dataset, developed by the Institute of Computing Technology, Chinese Academy of Sciences, contains 13,000 videos generated by two state-of-the-art open-source video generation models. It is designed to evaluate the impact of AI-generated content on video retrieval models. The dataset comprises 9,000 training videos and four test sets, each with 1,000 videos. All videos are generated based on the MSR-VTT dataset, and multiple strategies are employed to guarantee that the generated videos are semantically consistent with their original counterparts. This dataset is utilized to investigate the ranking bias issue of AI-generated videos in retrieval models.




