The pctma_net from one-of-student

pctma_net's Introduction

PCTMA-Net: Point Cloud Transformer with Morphing Atlas-based Point Generation Network for Dense Point Cloud Completion

Introduction

PCTMA-Net is a point transformer based shape completion method which directly maps a partial point cloud to a dense, complete point cloud without any voxelization. Its based on our IROS2021 [paper] more details please refers to our paper

Usage

1) Prerequisite

Install dependencies via pip3 install -r requirments.txt.
Follow this guide to install Open3D for point cloud I/O.
Build the point cloud distance metric in the folder pointComNet\pytorch_utils\components\externs_tools by running build.sh
Download dataset from [Completion3D] and [Kitti]

This code is built using pytorch 1.7 with CUDA 10.0 and tested on Ubuntu 18.04 with Python 3.7

Configuration

The configuration file for defining the PCTMA-Net parameter is located in pointComNet/config/pct_ma.yaml

Train

run the train demo in the pCTMA-Net by setting the value ``` --train True --evaluate False````

cd script
bash run_pctma_net.sh

Evaluate

run the evaluate demo in the pCTMA-Net by setting the value ``` --train False --evaluate True````

cd script
bash run_pctma_net.sh

For evaluating the kitti dataset, you can easily set the value --evaluateKitti True

Citation

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

@InProceedings{linpctma2021,
  author    = {Lin, Jianjie and Rickert, Markus and Perzylo, Alexander and Knoll, Alois},
  booktitle = {Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems},
  title     = {{PCTMA}-{N}et: Point Cloud Transformer with Morphing Atlas-based Point Generation Network for Dense Point Cloud Completion},
  year      = {2021},
}