PPR-Net++: Accurate 6D Pose Estimation in Stacked Scenarios

This is the code of pytorch version for our IROS2019 paper and TASE2021 journal paper: PPR-Net: point-wise pose regression network for instance segmentation and 6d pose estimation in bin-picking scenarios; PPR-Net++: Accurate 6D Pose Estimation in Stacked Scenarios.

Environment

Ubuntu 16.04/18.04

python3.6, torch 1.1.0, torchvision 0.3.0, opencv-python, sklearn, h5py, nibabel, et al.

Our backbone PointNet++ is borrowed from pointnet2.

Dataset

Siléane dataset is available at here.

Fraunhofer IPA Bin-Picking dataset is available at here.

Evaluation metric

The python code of evaluation metric is available at here.

Citation

If you use this codebase in your research, please cite:

@inproceedings{pprnet19IROS,
  title={PPR-Net: point-wise pose regression network for instance segmentation and 6d pose estimation in bin-picking scenarios},
  author={Dong, Zhikai and Liu, Sicheng and Zhou, Tao and Cheng, Hui and Zeng, Long and Yu, Xingyao and Liu, Houde},
  booktitle={2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={1773--1780},
  year={2019},
  organization={IEEE}
}

@article{zeng2021ppr,
  title={PPR-Net++: accurate 6-D pose estimation in stacked scenarios},
  author={Zeng, Long and Lv, Wei Jie and Dong, Zhi Kai and Liu, Yong Jin},
  journal={IEEE Transactions on Automation Science and Engineering},
  volume={19},
  number={4},
  pages={3139--3151},
  year={2021},
  publisher={IEEE}
}