40K自然主义6自由度机器人抓取演示数据集

本数据集由滑铁卢大学创建，包含约四万次成功的人控机器人抓取演示，使用RightHand Robotics的三指ReFlex夹持器，涵盖109种不同刚性物体。数据集通过结合人手动作跟踪和机器人手动控制的方法快速收集，支持非商业用途的自由使用、共享和修改。该数据集旨在为深度网络训练提供丰富的物理机器人抓取实例，以提高抓取规划的效率和成功率，同时帮助理解人类抓取策略。

This dataset was created by the University of Waterloo, containing approximately 40,000 successful human-controlled robotic grasping demonstrations using RightHand Robotics' three-fingered ReFlex gripper and covering 109 distinct rigid objects. The dataset was rapidly collected by combining human hand motion tracking and manual robotic control methods, and permits free use, sharing, and modification for non-commercial purposes. This dataset is designed to provide rich physical robotic grasping instances for deep network training, enhancing the efficiency and success rate of grasping planning, while also facilitating the understanding of human grasping strategies.