Yan-Xia / ASFM-Net

The codes for ACM MM 2021 paper 'ASFM-Net: Asymmetrical Siamese Feature Matching Network for Point Completion'

ASFM-Net

🔥🔥🔥 This repository is the official implementation for ACM Multimedia 2021 paper 'ASFM-Net: Asymmetrical Siamese Feature Matching Network for Point Completion'.

🔥🔥🔥 ASFM-Net achieves the 1st place in the leaderboard of Completion3D from April, 2021.

1. Getting Started Instructions.

• Clone this project

git clone --recursive https://github.com/Yan-Xia/ASFM-Net.git

Step-1: Environment Setup

• Install Docker Engine with NVIDIA GPU Support Toturial. We use the following commands

curl https://get.docker.com | sh \
  && sudo systemctl --now enable docker

distribution=$(. /etc/os-release;echo $ID$VERSION_ID) \
   && curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add - \
   && curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

sudo apt-get update
sudo apt-get install -y nvidia-docker2
sudo systemctl restart docker

• cd envs
• Run sudo ./build.sh, it may takes a while (around 10 minutes) for building the container.
• Run sudo ./launch.sh then it will bring up an new interactive command line interface.

• if your enounter problem below,

docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].

then you need to

sudo apt-get update && sudo apt-get install -y nvidia-container-toolkit
sudo systemctl restart docker

• The defualt GPU is all gpus. If you want to run on different deivce. Consider using this command in ./launch.sh, e.g., using device:1

docker run -itd --rm --gpus device=1 -p 6006:6006 -v $PWD/../:/ASFM-Net --name asfm-net asfm-net /bin/bash

Step-2: Download the point cloud datasets.

• We use the same dataset as PCN: Point Cloud Completion Network.
• Download the ShapeNet dataset in the shapent folder on Google Drive provided by PCN and unzip the dataset to folder data.

Step-3: ShapeNet Completion.

Download our trained model from Googe Drive for testing.
Run ./shapenet_test to test the accuracy of the model.
Run ./shapent_completion.sh to train and test ASFM-Net automatically. More specifically, the script will

• compile the point cloud distance ops, grouping ops and sampling ops
• train ASFM-Net
• test the accuracy of the saved model

2. Completion Results.

	Input	TopNet	PCN	RFA	ASFM-Net	Ground Truth
Airplane
Cabinet
Car
Chair
Lamp
Sofa
Table
Watercraft

3.Citation

---

If you find our work useful in your research, please consider citing:

@inproceedings{xia2021asfm ,
  title={ASFM-Net: Asymmetrical Siamese Feature Matching Network for Point Completion},
  author={Yaqi, Xia and Yan, Xia and Wei, Li and Rui, Song and Kailang, Cao and Uwe, Stilla},
  booktitle={Proceedings of the 29th ACM International Conference on Multimedia},
  year={2021}
}