Sim-Grasp-Dataset
收藏资源简介:
Sim-Grasp-Dataset是由普渡大学机械工程学院的多尺度机器人与自动化实验室创建的大型合成数据集,专注于复杂环境中的6自由度双指抓取。该数据集包含1550个对象,分布在500个场景中,共有790万个标注标签,旨在通过高级语言模型增强对象操作能力。数据集的创建过程涉及碰撞检查和刚体动力学,以生成全面的6D抓取标签,支持开发针对平行颚夹具的强大抓取算法。该数据集的应用领域包括智能机器人系统的抓取策略,旨在解决复杂和动态环境中的对象抓取问题。
Sim-Grasp-Dataset is a large-scale synthetic dataset developed by the Multi-Scale Robotics and Automation Lab, School of Mechanical Engineering, Purdue University, focusing on 6-degree-of-freedom (6-DoF) two-finger grasping in complex environments. This dataset includes 1550 objects distributed across 500 scenes, with a total of 7.9 million annotated labels, and is designed to enhance object manipulation capabilities via advanced large language models. The dataset creation process incorporates collision detection and rigid-body dynamics to generate comprehensive 6D grasping labels, supporting the development of robust grasping algorithms for parallel jaw grippers. Its application areas cover grasping strategies for intelligent robotic systems, aiming to solve object grasping problems in complex and dynamic environments.

- 1Sim-Grasp: Learning 6-DOF Grasp Policies for Cluttered Environments Using a Synthetic Benchmark普渡大学 · 2024年



