[论文] AnimationBench: Are Video Models Good at Character-Centric Animation?

## 论文概要 **研究领域**: CV **作者**: Leyi Wu, Pengjun Fang, Kai Sun **发布时间**: 2025-04-17 **arXiv**: [2504.13082](https://arxiv.org/abs/2504.13082) ## 中文摘要 视频生成技术快速发展，近期方法产生了越来越令人信服的动画效果。然而，现有基准——主要为真实视频设计——难以评估具有风格化外观、夸张动作和以角色为中心的连贯性的动画风格生成。此外，它们还依赖于固定提示集和刚性流程，为开放域内容和自定义评估需求提供有限的灵活性。为解决这一差距，我们引入AnimationBench，首个用于评估动画图像到视频生成的系统基准。AnimationBench将动画十二基本原则和IP保护可操作化为可测量的评估维度，同时包含更广泛的质量维度，包括语义一致性、动作合理性和相机运动一致性。该基准既支持标准化闭集评估以进行可复现的比较，也支持灵活的开放集评估以进行诊断分析，并利用视觉-语言模型进行可扩展评估。大量实验表明，AnimationBench与人类判断高度一致，并揭示了面向真实感的基准所忽略的动画特定质量差异，从而对最先进I2V模型进行更具信息量和区分度的评估。 ## 原文摘要 Video generation has advanced rapidly, with recent methods producing increasingly convincing animated results. However, existing benchmarks-largely designed for realistic videos-struggle to evaluate animation-style generation with its stylized appearance, exaggerated motion, and character-centric consistency. Moreover, they also rely on fixed prompt sets and rigid pipelines, offering limited flexibility for open-domain content and custom evaluation needs. To address this gap, we introduce AnimationBench, the first systematic benchmark for evaluating animation image-to-video generation. AnimationBench operationalizes the Twelve Basic Principles of Animation and IP Preservation into measurable evaluation dimensions, together with Broader Quality Dimensions including semantic consistency, mot... --- *自动采集于 2026-04-18* #论文 #arXiv #CV #小凯