qcchong / scnet

SCNet: Point Cloud Completion by shape-aware convolution

Xiangyang Wu, Chongchong Qu, Haixin Zhou, Yongwei Miao

---

This repository contains the source code for the paper “SCNet: Point Cloud Completion by shape-aware convolution”.

---

Usage

1) Envrionment

• CUDA 11.1
• Python 3.7
• Pytorch 1.9.0
• Pytorch-lightning 1.3.8
• knn_cuda 0.2

2) Compile

Compile loss and PyTorch 3rd-party modules

    cd loss/emd
    python setup.py install

    cd utils/pointnet2_ops_lib
    python setup.py install

3) Download data (MVP dataset)

    cd data
    bash get_dataset.sh

4) Train or validation

1. train the model

   python train.py 

2. validate the model. Here is a pretrained model. Download them here, and unzip it under logs/.

   python evalution.py

Acknowledgement

This code is based on VRCNet and Pointnet2.Pytorch.