Point Pattern Synthesis via Irregular Convolution

by Peihan Tu, Dani Lischinski, Hui Huang

Project page

The project page is available at https://vcc.tech/research/2019/PointSyn.

Introduction

This repository contains an implementation for Point Pattern Synthesis via Irregular Convolution. This method takes an input point pattern and generate visually similar output point pattern by optimization with a neural network. The implementation is in Python and Pytorch and Matlab.

You will need CUDA-compatible GPUs for generating results within several minutes.

If you have questions, please feel free to contact Peihan Tu (phtu@cs.umd.edu).

Starting

The current released codes are tested on Ubuntu 16.04. To train this network properly, please install the follow dependencies:

• Python 3.6
• CUDA 8.0
• Cudnn 6.0
• Pytorch
• Numpy

Clone our repo

git clone https://github.com/tph9608/Point-Synthesis.git

Run

You can simply generate results shown in the paper by runing

./exp.sh

Please go into the results_final folder the check all of the results.

Cite

If you use our code, please cite our paper:

@article{PointSyn19,
title = {Point Pattern Synthesis via Irregular Convolution},
author = {Peihan Tu, Dani Lischinski and Hui Huang},
journal = {Computer Graphics Forum (Proceedings of SGP 2019)},
volume = {38},
number = {5},
year = {2019},
} 

License

MIT License