Digital-Make-Up-Transfer-and-Stylization

Implementation of Digital face makeup by example with the following modifications -

1. Face key point detector is dlib's implementation with 69 points. Forehead points are manually adjustable and automatically added first using 3rd party skin detector implementation
2. Bilateral filtering is used instead of weighted-least-square (WLS)
3. XDoG of subject along with makeup is displayed. weighted sum of LAB values is taken.

How to run?

1. Install required dependencies using pip install -r requirements.txt

2. Place shape_predictor_68_face_landmarks.dat in src directory in not already there. You can download it from dlib official site

3. The program takes two inputs (i.e. Subject Image nd Example Image) and can be run as - python digital_makeup.py -lc -lf input/subject1.png input/example1.png

4. The program has option for 3 flags:

   • -si or show interactive to display output of intermediary steps like face alignment and various face masks extraction
   • The other two flags -lc and -lf which imply:
     • -lf or light foundation. Skin details aren’t hidden when foundation is light. Weight of subjects’ for transferring skin detail (δI ) is 1 instead of 0
     • -lc or light makeup color. Weight while doing color transfer are more in favor of subject image, γ = 0.5 instead of γ = 0.8 or weight of subject color 0.5 instead of 0.2

Design

The entire process can be broken into following steps -

• Face Alignment
• Layer Decomposition
• Skin Detail Transfer
• Color Transfer
• Highlight and Shading Transfer
• Lip Makeup

(To understand the complete approach of the Design go through - Report)

File Structure

1. digital_makeup.py
   1. Main py file, provides command line interface to call the entire digital makeup transfer process
2. src/face_alignment.py
   1. Face landmark points extraction
   2. Face Warping
   3. Face mask extraction for various components - skin, lips, eyes, mouth
3. src/makeup_util.py
   1. Layer Decomposition
   2. Color and Detail Transfer
   3. Highlight and Shading Transfer
   4. Lip makeup special function
4. src/xdog.py
   1. Cartoonize an image, black and white output
5. src/skin_detector.py
   1. Skin detection for face used in automatic detection of forehead points
6. input
   1. Contains image files to test code with
7. output
   1. Output of some experiment on input images

References

1. Face Warping Tutorial
2. Bilateral filtering for layer decomposition
3. Face landmark point extraction
4. XDoG implementation help
5. Similar Digital Makeup implementation