ThirdToFirst

ThirdToFirst数据集由中佛罗里达大学计算机视觉研究中心创建，包含531对同时录制的自我中心和非自我中心视频，旨在探索这两种视角之间的关系。数据集分为合成和真实两部分，涵盖了广泛的动作和运动，每对视频都进行了时间上的对齐。该数据集的创建过程涉及使用游戏引擎生成合成数据，以及通过佩戴身体摄像机和静态摄像机录制真实数据。数据集的应用领域包括视频检索、视频合成、动作识别、姿态估计和3D重建等，旨在解决自我中心和非自我中心视频之间的信息转移问题。

The ThirdToFirst Dataset was developed by the Computer Vision Center at the University of Central Florida. It contains 531 pairs of simultaneously recorded egocentric and non-egocentric videos, aiming to explore the relationship between these two perspectives. The dataset is divided into synthetic and real subsets, covering a wide range of actions and motions, with each pair of videos temporally aligned. The creation process of this dataset involves generating synthetic data using game engines, as well as recording real-world data via body-worn cameras and static cameras. Its applicable domains include video retrieval, video synthesis, action recognition, pose estimation, 3D reconstruction, and more, and it is designed to solve the problem of information transfer between egocentric and non-egocentric videos.

原始信息汇总

ThirdToFirst 数据集概述

数据集类型

• 真实数据集
• 合成数据集

数据集链接

• 真实数据集：Real Dataset
• 真实数据集分割：Real Splits
• 真实数据集标注：Real Data Annotations
• 合成数据集分割：Synthetic Splits
• 合成数据集生成器：Synthetic Dataset Generator

其他资源

• 最终结果：Final Results