E-Bench DB
收藏资源简介:
E-Bench DB是由北京大学电子与计算机工程学院和鹏城实验室联合创建的视频质量评估数据集,专门用于文本驱动视频编辑的质量评估。该数据集包含169个多样化的源视频,涵盖了真实世界、计算机图形渲染和AI生成内容等多种场景,以及多种编辑提示和编辑结果。数据集的创建过程中,邀请了24名人类注释者对每个视频进行主观评分,以确保评估的准确性和可靠性。E-Bench DB主要应用于AI生成内容的视频编辑质量评估,旨在解决现有评估方法与人类主观感知不一致的问题。
E-Bench DB is a video quality assessment dataset jointly created by the School of Electronic and Computer Engineering, Peking University and Peng Cheng Laboratory, which is specifically designed for text-driven video editing quality assessment. This dataset contains 169 diverse source videos covering various scenarios including real-world content, computer graphics renderings, and AI-generated content, as well as multiple editing prompts and their corresponding edited results. During the dataset construction process, 24 human annotators were invited to conduct subjective scoring for each video to ensure the accuracy and reliability of the assessment. E-Bench DB is primarily applied to the video editing quality assessment of AI-generated content, aiming to resolve the inconsistency between existing assessment methods and human subjective perception.
E-Bench 数据集概述
数据集名称
E-Bench
数据集描述
E-Bench 是用于评估文本驱动视频编辑质量主观对齐基准套件的官方实现。




