RCVPose 3D

(3DV Poster) Keypoint Cascade Voting for Point Cloud Based 6DoF Pose Estimation, a pure point cloud solution for 6DoF pose estimation.

Initalization

Dependencies

Dataset

Download LINEMOD, OCCLUSION_LINEMOD and YCB-Video from BOP, follow the [BOP toolkit}(https://github.com/thodan/bop_toolkit.git) to set up the datasets and then run

python gt_bop.py --root $(DATASET_ROOT)$

to generate the point cloud and radii for training and validation. The pre-generated point cloud and GT Radii can be downloaded here here.

Note that the minmax values are required for standrization when training the model in this step.

Train

The semantic segmentation and radii regression are trained saperately. This means that both of the networks can be trained simultaneously. In order to train the segmentation part, simply run

python main.py --root_dataset $(DATASET_ROOT)$ --globalmin 0 --globalmax 1 --obj_idx 1

Test

Pending...

Citation

Please cite us if our work is helpful:

@inproceedings{wu2022keypoint,
      title = {Keypoint Cascade Voting for Point Cloud Based 6DoF Pose Estimation},
      author = {Wu, Yangzheng and Javaheri, Alireza and Zand, Mohsen and Greenspan, Michael},
      booktitle = {2022 International Conference on 3D Vision (3DV)},
      year = {2022},
      organization={IEEE}
  }

Acknowledgements

This work is supported by Bluewrist Inc. and NESERC.