Contact-GraspNet: Efficient 6-DoF Grasp Generation in Cluttered Scenes
Contact-GraspNet — predicts 6-DoF parallel-jaw grasps for a whole cluttered scene from one depth view by treating observed points as grasp contacts, which shrinks the problem to 4 DoF per point.Contact-GraspNet：从单帧深度图为整个杂乱场景预测 6 自由度平行夹爪抓取；把观测到的点当作抓取接触点，使每个点只需学习 4 个自由度。
BibTeX
@misc{sundermeyer2021contact,
title = {{Contact-GraspNet: Efficient 6-DoF Grasp Generation in Cluttered Scenes}},
author = {Martin Sundermeyer and Arsalan Mousavian and Rudolph Triebel and Dieter Fox},
year = {2021},
eprint = {2103.14127},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2103.14127},
}Manipulation & Dexterous Hands操作与灵巧手Perception, SLAM & 3D Mapping感知、SLAM 与三维建图graspingpoint-cloud6-dofclutter