LfED-6D — Learning from Experience and Demonstration for 6-DOF Grasping Dataset

The LfED-6D dataset is a collection of 6D grasp annotations acquired through experience (with a robot platform) or by human demonstration. For known objects, the annotated grasps can be directly applied given the pose of the object model is correctly computed. For unknown objects, the grasps can be generalized using methods for shape matching, for example the Dense Geometrical Correspondence Network.

Fig.1. Robot grasping an object to collect experience data.

Download

LfED-6D Dataset

Research paper

If you found our dataset useful, please cite the following paper:

@article{patten_dgcm_net,
  author = {Patten, Timothy and Park, Kiru and Vincze, Markus}, 
  title = {{DGCM-Net: Dense Geometrical Correspondence Matching Network for Incremental Experience-based Robotic Grasping}}, 
  journal={arXiv}, 
  year={2020}
}

Contact & credits

For any questions about the LfED-6D dataset, please contact:

  • Tim Patten – email: patten@acin.tuwien.ac.at
  • Kiru Park – email: park@acin.tuwien.ac.at