SPARE: Simulated and Physical ARticulated Extendable dataset
收藏arXiv2018-03-30 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/1803.11147v1
下载链接
链接失效反馈官方服务：
资源简介：
SPARE数据集是密歇根大学创建的第一个结合模拟与物理实例的开放源代码数据集，专注于提供可扩展的关节对象（运动链）实例。该数据集通过Gazebo物理模拟环境捕获RGBD图像，提供完整的3D运动链描述，支持从模拟到真实世界的学习转移。SPARE数据集适用于机器人视觉领域，旨在解决复杂关节对象的动态属性估计问题，支持多视角和时间序列分析，为机器人理解和操作未知关节对象提供基础。
The SPARE Dataset, developed by the University of Michigan, is the first open-source dataset that integrates both simulated and physical real-world instances, focusing on providing scalable articulated object (kinematic chain) instances. This dataset captures RGBD images via the Gazebo physics simulation environment, offers complete 3D kinematic chain descriptions, and supports sim-to-real learning transfer. Tailored for the field of robotic vision, the SPARE Dataset aims to address the problem of dynamic attribute estimation for complex articulated objects, supports multi-view and time-series analysis, and provides a foundation for robots to understand and manipulate unknown articulated objects.
提供机构：
密歇根大学
创建时间：
2018-03-30
