Requirements

• Install python packages

• pip install hydra-core --upgrade --pre  
  pip install torch
  pip install plyfile
  pip install seaborn

• Note: It is tested with:

  python 3.6.10
  torch==1.2.0
  hydra-core==0.11.3
  plyfile==0.7.1
  seaborn==0.10.0

Download dataset

• Download ModelNet40

  python examples/download.py

How to use

• This repository use hydra, so have configs (args) in examples/configs folder and outputs data in YYYY-MM-DD/HH-MM-SS folder.
• training
  • train PointNet AutoEncoder and sklearn.svm.OneClassSVM

    python examples/train_w_svm.py dataset_root=data/modelnet40_normal_resampled/

    • examples/train_w_svm.py args was written on examples/configs/config.yaml.
• evaluation
  • evaluate PointNet AutoEncoder with sklearn.svm.OneClassSVM, T-SNE and ply data

    python examples/eval_w_svm.py dataset_root=data/modelnet40_normal_resampled/ resume=outputs/YYYY-MM-DD/HH-MM-SS/model.pth.tar

    • examples/eval.py args was written on examples/configs/config.yaml.
• test
  • extract global features using PointNet AutoEncoder

    python examples/tests/test2.py

references

• yanx27. Pointnet_Pointnet2_pytorch. In Github repository, 2019. (url:https://github.com/yanx27/Pointnet_Pointnet2_pytorch) (access:2020/7/14)
• fxia22. pointnet.pytorch. In Github repository, 2017. (url:https://github.com/fxia22/pointnet.pytorch) (access:2020/7/20)
• charlesq34. pointnet-autoencoder. In Github repository, 2018. (url:https://github.com/charlesq34/pointnet-autoencoder/blob/master/part_dataset.py) (access:2020/7/25)
• chrdiller. pyTorchChamferDistance. In Gihub repository, 2019. (url:https://github.com/chrdiller/pyTorchChamferDistance) (access:2020/7/25)