JunJiaGuo/VidComposition_Benchmark
收藏Hugging Face2025-04-08 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/JunJiaGuo/VidComposition_Benchmark
下载链接
链接失效反馈官方服务：
资源简介：
VidComposition是一个评估多模态大型语言模型（MLLMs）对视频构成理解能力的数据集。它由982个经过精心挑选的视频和1706个多项选择题组成，旨在测试模型在理解视频元素组合和交互方面的能力。这些视频和问题涵盖了摄像机移动、角度、镜头大小、叙事结构、角色动作和情感等多个方面。
VidComposition is a benchmark for evaluating the video composition understanding capabilities of Multimodal Large Language Models (MLLMs). It consists of 982 carefully curated videos and 1706 multiple-choice questions designed to test the models ability to understand the combination and interaction of visual elements within the videos. The videos and questions cover various aspects such as camera movement, angle, shot size, narrative structure, character actions, and emotions.
提供机构：
JunJiaGuo
