DexArt Manipulation Dataset (DAM)
收藏资源简介:
DexArt Manipulation Dataset (DAM) 是由上海交通大学等机构创建的一个用于评估机器人操作多指手和关节对象能力的数据集。该数据集包含6000个点云观察,涵盖了机器人和关节对象的随机状态,用于支持机器人在模拟环境中进行复杂的操作任务。数据集的创建旨在解决机器人在日常环境中操作关节对象的挑战,特别是在处理高自由度的对象和手部时。通过此数据集,研究者可以评估和改进机器人的操作策略,使其更接近人类的行为,并提高在未知对象上的泛化能力。
The DexArt Manipulation Dataset (DAM) is a dataset developed by institutions including Shanghai Jiao Tong University for assessing robotic multi-fingered hand manipulation and articulated object manipulation capabilities. This dataset contains 6000 point cloud observations covering random states of robots and articulated objects, which supports complex manipulation tasks for robots in simulated environments. The creation of this dataset aims to address the challenges of robotic articulated object manipulation in daily environments, especially when handling objects and multi-fingered hands with high degrees of freedom. With this dataset, researchers can evaluate and refine robotic manipulation strategies to align them more closely with human-like manipulation behaviors, as well as enhance generalization capabilities to unknown objects.

- 1DexArt: Benchmarking Generalizable Dexterous Manipulation with Articulated Objects上海交通大学 · 2023年



