SC-FEGAN : Face Editing Generative Adversarial Network with User's Sketch and Color (ICCV 2019)

Youngjoo Jo, Jongyoul Park

arXiv: https://arxiv.org/abs/1902.06838

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.

• 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/
python3 demo.py

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)

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'.

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.

@InProceedings{Jo_2019_ICCV,
  author = {Jo, Youngjoo and Park, Jongyoul},
  title = {SC-FEGAN: Face Editing Generative Adversarial Network With User's Sketch and Color},
  booktitle = {The IEEE International Conference on Computer Vision (ICCV)},
  month = {October},
  year = {2019}
}

• 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.