📄 中文摘要
视频质量评估（VQA）正从单一的平均意见分数向更丰富的多维度视频内容评估演变。研究提出了一个大规模的多维VQA数据集UltraVQA，该数据集涵盖了多样化的用户生成内容（UGC），并在五个关键质量维度上进行了标注：运动质量、运动幅度、美学质量、内容质量和清晰度质量。数据集中每个视频均由超过三位人类评审者在这些维度上进行评分，并附有细致的子属性标签，同时根据集体人类判断生成了解释性理由。为更好地利用这些丰富的标注并改善离散质量评分评估，提出了分析评分优化的方法。
多维视频质量评估的分析评分优化
出处: Analytic Score Optimization for Multi Dimension Video Quality Assessment
发布: 2026年2月20日
📄 English Summary
Analytic Score Optimization for Multi Dimension Video Quality Assessment
Video Quality Assessment (VQA) is evolving from a single-number mean opinion score to richer, multi-faceted evaluations of video content. This study presents a large-scale multi-dimensional VQA dataset, UltraVQA, which encompasses diverse User-Generated Content (UGC) annotated across five key quality dimensions: Motion Quality, Motion Amplitude, Aesthetic Quality, Content Quality, and Clarity Quality. Each video in the dataset is scored by over three human raters on these dimensions, with fine-grained sub-attribute labels, and is accompanied by an explanatory rationale generated by GPT based on collective human judgments. To better leverage these rich annotations and improve discrete quality score assessment, Analytic Score Optimization is introduced.