ShiZhuming / StyleTransfer

A real-time image style transfer project based on AdaIn and VGG network , python flask as frontend and HTML5 as backend , project of AI Introduction, a course of Peking Unicersity

Image Style Transfer

Introduction

This is a real-time image style transfer project based on AdaIn and VGG network , python flask as frontend and HTML5 as backend , project of AI Introduction, a course of Peking Unicersity

Our website is http://pkuszm.cn/

Authors

Zhuming Shi, Weibo Xu, Yiming Zhao

Dependence

Python 3 only

$ apt-get install python3.6

Use requirements.txt to install packages

$ pip3 install -r requirements.txt

Use start.py to start web application

$ python3 start.py

It is recommended to use virtualenv and gunicorn to deploy the website on the server

$ pip install virtualenv
$ virtualenv -p /usr/bin/python3 ENVE
$ gunicorn -c gunicorn.conf.py --error-logfile errorlog.txt start:app -D

Structure

StyleTransfer/
├── LICENSE
├── README.md
├── start.py
├── function.py
├── net.py
├── train.py
├── convert.py
├── thumb.py
├── favicon.ico
├── image
│   ├── upload
│   │   ├── content
│   │   ├── style
│   │   └── convert
│   └── thumb
│       ├── content
│       ├── style
│       └── convert
├── static
├── templates
│   ├── index.html
│   ├── show.html
│   ├── authors.html
│   └── submissions.html
├── test.py
├── .gitignore
└── requirements.txt

References

1. Gatys, Leon A., A. S. Ecker, and M. Bethge. "Image Style Transfer Using Convolutional Neural Networks." Computer Vision & Pattern Recognition 2016.

2. Huang, Xun, and S. Belongie. "Arbitrary Style Transfer in Real-time with Adaptive Instance Normalization." 2017.