SCARP: 3D Shape Completion in ARbitrary Poses for Improved Grasping

Bipasha Sen^*1, Aditya Agarwal^*1, Gaurav Singh^1*, Brojeshwar Bhowmick², Srinath Sridhar³, Madhava Krishna^1
1International Institute of Information Technology, Hyderabad, ²TCS Research India, ³Brown University

^*denotes equal contribution

This is the official implementation of the paper "SCARP: 3D Shape Completion in ARbitrary Poses for Improved Grasping" accepted at ICRA 2023.

This work was featured as "The Publication of the Week" by Weekly Robotics here

For more results, information, and details visit our project page and read our paper.

Creating the Anaconda/Miniconda environment

Make sure you have Anaconda or Miniconda installed before you proceed to load this environment.

conda env create -f environment.yml
conda activate SCARP

Pretrained Checkpoints

Pretrained checkpoints for the demo can be downloaded from here

Move the checkpoints to the directory checkpoints/.

The directory structure should be:

└── checkpoints
    ├── plane.pt
    .
    .
    .

    └── car.pt

Demo

To run a demo run the following command

CUDA_VISIBLE_DEVICES=0 python3 demo.py  \
--class_choice plane \
--ckpt_load checkpoints/plane.pt

The outputs are stored in the demo_data/<class_choice> directory

Real World Results

Acknowledgement

Some parts of the code are insipired and borrowed from ConDor,Sinv and Pointnet++. We thank the authors for providing the source code.

Citation

If you find our work useful in your research, please cite:

@INPROCEEDINGS{10160365,
  author={Sen, Bipasha and Agarwal, Aditya and Singh, Gaurav and B., Brojeshwar and Sridhar, Srinath and Krishna, Madhava},
  booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)}, 
  title={SCARP: 3D Shape Completion in ARbitrary Poses for Improved Grasping}, 
  year={2023},
  volume={},
  number={},
  pages={3838-3845},
  doi={10.1109/ICRA48891.2023.10160365}
}

Contact

If you have any questions, please feel free to email the authors.

Bipasha Sen: [email protected]
Aditya Agarwal: [email protected]
Gaurav Singh: [email protected]