SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters. ECCV 2018
Yifan Xu, Tianqi Fan, Mingye Xu, Long Zeng, Yu Qiao.

This project is based on our ECCV18 paper. You can find the arXiv version here.

@article{xu2018spidercnn,
  title={SpiderCNN: Deep Learning on Point Sets with Parameterized Convolutional Filters},
  author={Xu, Yifan and Fan, Tianqi and Xu, Mingye and Zeng, Long and Qiao, Yu},
  journal={arXiv preprint arXiv:1803.11527},
  year={2018}
}

SpiderCNN is a convolutional neural network that can process signals on point clouds.

The code is based on PointNet， and PointNet++. Please install TensorFlow, and follow the instruction in PointNet++ to compile the customized TF operators.
The code has been tested with Python 2.7, TensorFlow 1.3.0, CUDA 8.0 and cuDNN 6.0 on Ubuntu 14.04.

1. Install Docker using these commands:

sudo apt update
sudo apt install apt-transport-https ca-certificates curl software-properties-common
curl -fsSL https://download.docker.com/linux/ubuntu/gpg | sudo apt-key add -
sudo add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/ubuntu focal stable"
sudo apt update
apt-cache policy docker-ce
sudo apt install docker-ce
sudo systemctl status docker

2. Copy raw This script to .txt file and save it as "Dockerfile".
3. Run Dockerbuild using these commands:

sudo docker build -t installation .
sudo docker run -w /SpiderCNN -it installation:latest

4. After running commands above, do unzip in the container:

unzip -q modelnet40_ply_hdf5_2048.zip -d data
unzip -q shapenetcore_partanno_segmentation_benchmark_v0_normal.zip -d data

5. Run this command:

python train.py

6. There were written several test scripts. They are located in part_seg directory and in utils directory. To run then use the following commands:

python -m unittest test

Preprocessed ModelNet40 dataset can be downloaded here.
To train a SpiderCNN model (with input XYZ coordinates and normal vectors) to classify shapes in ModelNet40:

python train.py

To train a SpiderCNN model (with input XYZ coordinates) with multi GPU to classify shapes in ModelNet40:

python train_xyz.py

Preprocessed ShapeNetPart dataset can be downloaded here. To train a model to segment object parts for ShapeNet models (with input XYZ coordinates and normal vectors):

cd part_seg
python train.py

This repository is released under MIT License (see LICENSE file for details).