KwaiVGI/MultiCamVideo-Dataset

MultiCamVideo数据集是一个使用Unreal Engine 5渲染的多相机同步视频数据集，包含同步的多相机视频及其相应的相机轨迹。该数据集由13.6K个不同的动态场景组成，每个场景由10个相机捕获，总共产生136K个视频。每个动态场景包括四个元素：3D环境、角色、动画和相机。数据集涵盖了各种室内外场景，如城市街道、购物中心、咖啡馆、办公室和乡村等。角色和动画从Fab和Mixamo收集，相机轨迹和参数多样化，以模拟真实世界的拍摄情况。The MultiCamVideo Dataset is a multi-camera synchronized video dataset rendered using Unreal Engine 5, including synchronized multi-camera videos and their corresponding camera trajectories. The dataset consists of 13.6K different dynamic scenes, each captured by 10 cameras, resulting in a total of 136K videos. Each dynamic scene includes four elements: 3D environment, character, animation, and camera. The dataset covers a variety of indoor and outdoor scenes such as city streets, shopping malls, cafes, office rooms, and countryside. Characters and animations are collected from Fab and Mixamo, and camera trajectories and parameters are diversified to simulate real-world shooting conditions.

搜集汇总

背景与挑战

背景概述

MultiCamVideo-Dataset是一个由Unreal Engine 5渲染的多摄像头同步视频数据集，包含13,600个动态场景，每个场景由10个摄像头捕获，总共有136,000个视频和112,000个相机轨迹，视频分辨率为1280x1280。该数据集设计用于支持相机控制视频生成、同步视频制作和3D/4D重建等计算机视觉任务，提供了丰富的3D环境、角色、动画和多样化相机轨迹配置。