T2V视频质量评估数据集

本数据集由都柏林城市大学的数据分析中心创建，包含超过1000个由5种最新文本到视频（T2V）模型生成的视频。数据集旨在评估这些视频的自然性和与文本提示的语义匹配度，通过应用多种质量评估指标和人类评估来比较不同模型的表现。数据集的应用领域主要集中在AI视频生成质量评估，旨在解决如何有效评估T2V模型输出质量的问题。

This dataset was developed by the Data Analytics Center at Dublin City University, and contains over 1,000 videos generated by five state-of-the-art text-to-video (T2V) models. It aims to assess the naturalness of these videos as well as their semantic alignment with given text prompts, by utilizing multiple quality assessment metrics and human evaluation to compare the performance across different models. The dataset is primarily targeted at the field of AI video generation quality assessment, with the goal of addressing the challenge of effectively evaluating the output quality of T2V models.