AIM-SCU/MAVEN_Multicultura_Text-to-Video_Generation
收藏资源简介:
这是一个全面的AI生成视频数据集,展示不同文化、动作类型和地点的多元文化内容。数据集包含使用多种管道方法生成的视频,包括原始和精炼提示。总共有972个条目(243个独特视频 × 4种管道),涵盖三种文化(中国、美国、罗马尼亚)、三种动作类型(食物、音乐、舞蹈)以及四种生成管道(基线单代理、单代理、多代理顺序、多代理并行)。数据集支持文化感知视频生成和表示的研究,每个视频为5秒MP4格式,并包含详细元数据。
A comprehensive dataset of AI-generated videos showcasing multicultural content across different cultures, action types, and locations. The dataset includes videos generated using multiple pipeline approaches with both original and refined prompts. Total entries: 972 (243 unique videos × 4 pipelines). Cultures: Chinese, American, Romanian. Action types: food, music, dance. Pipelines: base, sa (single-agent), mas (multi-agent sequential), map (multi-agent parallel). The dataset supports research into culturally-aware video generation and representation, with each video being 5 seconds in MP4 format and containing detailed metadata.




