DoFace | 抖脸

A package that supports Virtual Makeup function, including lip color transfer and cosmetic contact lenses. This repository is based on PyTorch.

一个支持包括唇彩试色和美瞳试戴功能的虚拟试妆软件包。该项目基于PyTorch。项目介绍见此MOOC链接。

Get started

Clone the repository and configurate the environment.

git clone [email protected]:ChineseYjh/DoFace.git
pip install -r requirements.txt

Use DoFace in your projects.

Your project
│   README.md
│   ...
│   foo.py
│
└───DoFace
│
└───directory1
│   
└───...

Write codes in foo.py like:

import DoFace
img_path='./imgs/test.jpg' # original face image path
style="niuxuese" # it supports "niuxuese", "nanguase" and "yinghuafen" for chuncai(唇彩) ,and "shenlanse", "yamajin" for meitong(美瞳). All styles are supported by doface function.
dst_path='./imgs/res-niuxuese.jpg' # makeup result image path
save=True # whether to save img to dst_path
img=DoFace.doface(img_path=img_path,style=style,dst_path=dst_path,save=save,
                  cp='./DoFace/facetools/parsing/cp/cp_parsing.pth') # cp points to the face-parsing weight path

This package now supports:

• chuncai(唇彩)
  • nanguase(南瓜色)
  • niuxuese(牛血色)
  • yinghuafen(樱花粉)
• meitong(美瞳)
  • shenlanse(深蓝色)
  • yamajin(亚麻金)

Choose one and assign it to style in foo.py.

Acknowledgement

This repository is based on facetools and pytorch-CycleGAN-and-pix2pix