EgoExoLearn 跨视角技能学习数据集

该数据集是由上海人工智能实验室、南京大学、中科院深圳先进技术研究院等领头单位，与东京大学、复旦大学、浙江大学、中国科学技术大学等多所高校的学生和研究者联合发布的 EgoExoLearn，旨在为机器人赋予观察他人学习新动作的能力。

The dataset named EgoExoLearn was jointly released by leading institutions including Shanghai AI Laboratory, Nanjing University, and Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, as well as students and researchers from multiple universities such as The University of Tokyo, Fudan University, Zhejiang University, and University of Science and Technology of China. It aims to endow robots with the ability to learn new actions by observing human demonstrators.

搜集汇总

背景与挑战

背景概述

EgoExoLearn 是一个跨视角技能学习数据集，由上海人工智能实验室、南京大学等机构联合发布，旨在让机器人通过观察他人学习新动作。它包含120小时的第一视角和第三视角视频，覆盖日常和实验室场景，并提供注视数据与多模态标注，以支持机器在不同视角下的动作建模和基准测试。