AIR-Act2Act
收藏arXiv2020-09-04 更新2024-06-21 收录
下载链接：
https://github.com/ai4r/AIR-Act2Act
下载链接
链接失效反馈官方服务：
资源简介：
AIR-Act2Act数据集是由电子通信研究院创建的，专门用于教授机器人非言语社交行为的人-人互动数据集。该数据集包含5000个互动样本，由100名老年人和2名大学生在室内环境中执行10种互动行为。数据集通过三台Microsoft Kinect v2相机捕捉，提供了深度图、身体索引和3D骨骼数据。此外，数据集还包括了将人类行为转换为NAO机器人关节角度的数据。AIR-Act2Act数据集不仅可用于教授机器人社交技能，还可用于动作识别算法的基准测试。
The AIR-Act2Act dataset, developed by the Electronics and Telecommunications Research Institute, is a human-human interaction dataset specifically designed for teaching robots non-verbal social behaviors. It contains 5,000 interaction samples collected from 100 elderly individuals and 2 college students who performed 10 types of interactive behaviors in an indoor environment. Captured using three Microsoft Kinect v2 cameras, the dataset provides depth maps, body index data, and 3D skeletal data. Additionally, the dataset includes data that converts human behaviors into joint angles of the NAO robot. The AIR-Act2Act dataset can be used not only for teaching social skills to robots but also for benchmarking action recognition algorithms.
提供机构：
电子通信研究院
创建时间：
2020-09-04
