zhangbowen-wx / Shape-Completion-with-Points-in-the-Shadow

[SIGGRAPH Aisa 2022 Conference Paper] Shape Completion with Points in the Shadow
https://zhangbowen-wx.github.io/projects/SCPS/SCPS.html
MIT License
9 stars 3 forks source link
3d-vision deep-learning point-cloud-completion siggraph-asia-2022

Shape Completion with Points in the Shadow

Bowen Zhang, Xi Zhao, He Wang, Ruizhen Hu


This repository contains the source code for the paper Shape Completion with Points in the Shadow, The project page is here.

Datasets

We use the Multi-View Partial point cloud dataset (MVP) in our experiments. As there is no camera configuration in the origin MVP dataset, we recompute the MVP dataset to record the camera configuration. The dataset we use can be avaliable here.

The pretrained models of our method are avaliable as follows:

Install Python Dependencies

git clone https://github.com/zhangbowen-wx/Shape-Completion-with-Points-in-the-Shadow.git
cd Shape-Completion-with-Points-in-the-Shadow
pip install -r requirements.txt

Build Pytorch Extensions

cd utils/ChamferDistancePytorch/chamfer3D
python setup.py install

cd ../../../

cd utils/pointnet2_ops_lib
python setup.py install

Usage

To train our models, you can use following commands:

python train.py # train our model(with GT viewpoints)
python train_VP.py # train our model(without GT viewpoints)

The trained model parameters will be placed in the log/model/ folder.

To test our models, you can use following commands:

python test.py # test our model(with GT viewpoints)
python test_VP.py # test our model(without GT viewpoints)

License

This project is open sourced under MIT license.