Pointnet2/Pointnet++ PyTorch

• Implemention of Pointnet2/Pointnet++ written in PyTorch.
• Supports Multi-GPU via nn.DataParallel.
• Supports PyTorch version >= 1.0.0. Use v1.0 for support of older versions of PyTorch.

See the official code release for the paper (in tensorflow), charlesq34/pointnet2, for official model definitions and hyper-parameters.

The custom ops used by Pointnet++ are currently ONLY supported on the GPU using CUDA.

Setup

• Install python -- This repo is tested with 2.7, 3.5, and 3.6

• Install dependencies

  pip install -r requirements.txt
  

• Building _ext module

  python setup.py build_ext --inplace
  

• Optionally, you can also install this repo as a package

  pip install -e .
  

Example training

Two training examples are provided by pointnet2/train/train_sem_seg.py and pointnet2/train/train_cls.py. The datasets for both will be downloaded automatically by default.

They can be run via

python -m pointnet2.train.train_cls

python -m pointnet2.train.train_sem_seg

Both scripts will print training progress after every epoch to the command line. Use the --visdom flag to enable logging to visdom and more detailed logging of training progress.

Contributing

This repository uses black for linting and style enforcement on python code. For c++/cuda code, clang-format is used for style. The simplest way to comply with style is via pre-commit

pip install pre-commit
pre-commit install

Citation

@article{pytorchpointnet++,
      Author = {Erik Wijmans},
      Title = {Pointnet++ Pytorch},
      Journal = {https://github.com/erikwijmans/Pointnet2_PyTorch},
      Year = {2018}
}

@inproceedings{qi2017pointnet++,
    title={Pointnet++: Deep hierarchical feature learning on point sets in a metric space},
    author={Qi, Charles Ruizhongtai and Yi, Li and Su, Hao and Guibas, Leonidas J},
    booktitle={Advances in Neural Information Processing Systems},
    pages={5099--5108},
    year={2017}
}