QUB-PHEO
收藏资源简介:
QUB-PHEO是一个视觉为基础的二元多视角数据集,旨在推动装配操作中的人机交互(HRI)和意图推理研究。该数据集捕捉了两个参与者之间的丰富多模态互动,其中一个充当“机器人代理”,涵盖多种装配任务,细分为36个不同的子任务。数据集包含丰富的视觉注释,如面部标志、注视、手部动作、物体定位等,共有70名参与者。QUB-PHEO提供两个版本:50名参与者的完整视频数据和所有70名参与者的视觉提示。设计用于改进HRI的机器学习模型,QUB-PHEO能够深入分析细微的互动线索和意图,有望为该领域做出贡献。数据集的应用领域主要集中在复杂装配任务中的人机协作,旨在解决机器人如何准确理解和响应人类意图的问题。
QUB-PHEO is a vision-based two-participant multi-view dataset developed to advance human-robot interaction (HRI) and intent inference research in assembly operations. This dataset captures rich multimodal interactions between two participants, where one assumes the role of the "robot agent", covering various assembly tasks categorized into 36 distinct subtasks. Involving a total of 70 participants, the dataset provides comprehensive visual annotations including facial landmarks, gaze tracking, hand movements, object localization and other relevant cues. QUB-PHEO offers two variants: full-length video data from 50 participants and visual prompts derived from all 70 participants. Tailored to enhance machine learning models for HRI, QUB-PHEO enables in-depth analysis of subtle interaction cues and intents, and holds promising potential to contribute to the relevant research field. The dataset is primarily applied in human-robot collaboration for complex assembly tasks, aiming to address the problem of how robots can accurately understand and respond to human intentions.




