Rapidata/image-to-video-human-preference-seedance-1-pro

该数据集包含约6,000名人类标注者的约2,000个响应，用于评估Seedance 1 Pro视频生成模型。数据集包括提示、图像和视频资产等特征，以及与对齐、一致性和偏好相关的加权结果。数据集还包括对数据集结构的解释、评分系统，并对对齐、一致性和偏好提供了示例。此外，还提到了用于收集人类反馈的Rapidata平台，并列出了其他相关数据集。

This dataset includes approximately 6,000 responses from 2,000 human annotators, collected to evaluate the Seedance 1 Pro video generation model. The dataset features include prompts, image and video assets, and various weighted results related to alignment, coherence, and preference. The README also explains the structure of the dataset, the scoring system, and provides examples for alignment, coherence, and preference. Additionally, it mentions the Rapidata platform for collecting human feedback and lists other related datasets.