Rapidata/text-2-video-human-preferences-wan2.1

Alibaba Wan2.1 Human Preferences数据集是一个包含约45000个人类注释的视频生成模型评估数据集。该数据集用于评估Alibaba Wan 2.1视频生成模型在特定基准上的表现。数据集包含成对的视频比较，包括视频链接、各种评分和详细结果。评分包括对齐度、连贯性和偏好度，用于衡量视频与提示的匹配程度、逻辑一致性以及视觉吸引力。

The Alibaba Wan2.1 Human Preferences dataset is a video generation model evaluation dataset containing approximately 45,000 human annotations. This dataset is used to assess the performance of the Alibaba Wan 2.1 video generation model on a specific benchmark. The dataset includes paired video comparisons with video links, various scores, and detailed results. The scores include alignment, coherence, and preference, which are used to measure the match between videos and prompts, logical consistency, and visual appeal.