SynCamVideo-Dataset
收藏资源简介:
SynCamVideo-Dataset是由快手科技和浙江大学联合创建的多视角同步视频数据集,旨在支持多摄像头视频生成任务。该数据集包含500组同步视频,每组视频由36个摄像头从不同视角拍摄,涵盖了城市环境中的动态场景。数据集的创建过程结合了单摄像头视频、多视角图像和Unreal Engine渲染的多视角视频,确保了数据的几何和视觉一致性。该数据集主要应用于虚拟拍摄和多视角视频生成,旨在解决多视角视频生成中的动态一致性和视角多样性问题。
SynCamVideo-Dataset is a multi-view synchronized video dataset jointly developed by Kuaishou Technology and Zhejiang University, targeting multi-camera video generation tasks. The dataset contains 500 sets of synchronized videos, with each set captured by 36 cameras from different viewpoints, covering dynamic scenes in urban environments. Its creation process combines single-camera videos, multi-view images, and multi-view videos rendered via Unreal Engine, ensuring both geometric and visual consistency across the data. This dataset is mainly applied to virtual shooting and multi-view video generation, aiming to address the challenges of dynamic consistency and view diversity in multi-view video generation.

- 1SynCamMaster: Synchronizing Multi-Camera Video Generation from Diverse Viewpoints浙江大学, 快手科技, 清华大学, 香港中文大学 · 2024年



