SpatialVID/SpatialVID

SpatialVID是一个大规模视频数据集，包含空间标注，适用于文本到视频、文本到3D、图像到3D、图像到视频等任务。数据集包含每个视频剪辑的元数据、动态掩码、相机内参和位姿、运动指令以及可选的深度图。README文件详细介绍了数据集的大小、任务类别、语言、目录结构、使用指南、许可协议、引用格式以及如何下载和使用数据集。

SpatialVID is a large-scale video dataset with spatial annotations, suitable for tasks like text-to-video, text-to-3D, image-to-3D, image-to-video, and other related tasks. The dataset includes metadata for each video clip, annotations with dynamic masks, camera intrinsics and poses, motion instructions, and optional depth maps. The README file provides detailed information about the datasets size, task categories, language, directory structure, and usage guide. It also includes information about the license, citation format, and links to the datasets project page, paper, code, and dataset on Hugging Face and ModelScope.