Skip to content
SC-FEGAN : Face Editing Generative Adversarial Network with User's Sketch and Color
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
imgs Add files via upload Feb 13, 2019
samples Add files via upload Feb 13, 2019
ui Update Feb 20, 2019
utils Add files via upload Feb 12, 2019
.gitignore Initial commit Feb 12, 2019
LICENSE Add LICENSE file Feb 21, 2019 Update Mar 20, 2019 Update clear & fix bug Feb 20, 2019
demo.yaml Update demo.yaml Feb 13, 2019 Update Feb 13, 2019 Update Mar 4, 2019
requirements.txt Update requirements.txt May 9, 2019


SC-FEGAN : Face Editing Generative Adversarial Network with User's Sketch and Color

Youngjoo Jo, Jongyoul Park


Teaser GUI


Edit face images using a a deep neural network. Users can edit face images using intuitive inputs such as sketching and coloring, from which our network SC-FEGAN generates high quality synthetic images. We used SN-patchGAN discriminator and Unet-like generator with gated convolutional layers.

Teaser Image


  • tensorflow
  • numpy
  • Python3
  • PyQt5
  • opencv-python
  • pyyaml


First, download the model from Google drive.

Run these commands to start the program.

mv /${HOME}/SC-FEGAN.ckpt.* /${HOME}/ckpt/

Select the number of GPUs you want to use by editing demo.yaml file (multi-GPUs are not supported).

GPU_NUM: 1 (the number you want to use)
#GPU_NUM: (if you want to use only CPU, erase the number)

How to Use

Edit face images using a simple GUI. Only erased regions of the image are filled in by the network.

Explanation of the buttons:

  • Open Image: Open the image you want to edit.
  • Mask: Draw a mask on the desired regions of the face on the left viewer.
  • Sketches: Sketch the desired lines on the left viewer.
  • Color: Draw colored lines. If you click this button the first time, you have to choose a color from the palette.
  • Palette: Change color. After choosing a color, click the Color button to apply the change.
  • Save Img: Save the results. It is saved as 'name.jpg'.
  • Arrange: Arrange the editing works.
  • Undo: Undo previous editing work.
  • Complete: Generate the image and show it on the right viewer.

We recommend using the following workflow:

1. Draw the sketch plausibly referring to the original image.
2. Draw the mask on the sketched region.
3. Click the `Arrange` button.
4. Draw the color on the masked region.
5. Click `Complete'.

Example Results

Face editing

Face editing

Edit earring


Face restoration


Face restoration (with only sketch and color)



CC 4.0 Attribution-NonCommercial International

The software is for educational and academic research purpose only.


  • This is developed on Linux machine running Ubuntu 18.04.1
  • Provided model and sample code is under a non-commercial creative commons license.


  title={SC-FEGAN: Face Editing Generative Adversarial Network with User's Sketch and Color},
  author={Jo, Youngjoo and Park, Jongyoul},
  journal={arXiv preprint arXiv:1902.06838},


  • Update training code


We acknowledge the official code DeepFillv1. We thank all the researchers at ETRI, especially Kimin Yun and Jinwoo Jung, for insightful discussions.

You can’t perform that action at this time.