ManiSkill-2021
Tongzhou Mu, Zhan Ling Zhan Ling, Fanbo Xiang, Derek Yang Derek Yang, Xuanlin Li Xuanlin Li, Stone Tao, Zhiao Huang, Zhiwei Jia Zhiwei Jia, Hao Su Hao Su · Science Data Bank Datasets · 2022
At home, office, or other environments, we see the 3D world and manipulate objects surrounding us. Computer scientists today are interested in building artificial intelligence that can behave and interact with 3D objects as we humans do. To this end, we introduce SAPIEN Manipulation Skill Challenge, or ManiSkill Challenge for short, a challenge focusing on acquiring generalizable manipulation skills.ManiSkill challenge is built on the ManiSkill Benchmark, which utilizes a rich and diverse set of articulated objects to build 4 tasks. All tasks are equipped with high-quality demonstrations for learning-from-demonstrations approaches and we provide common baseline algorithms. We believe this challenge will encourage the robot learning community to explore generalizable object manipulation skills.This dataset contains the training datasets and pre-trained baseline models of ManiSkill Challenge 2021( https://sapien.ucsd.edu/challenges/maniskill2021).