Face swap with OpenCV

This is a small project aim to implemet a swapping faces method, in which we use module dlib landmark detection.

Installation and setup

If you don't want to run .sh file you can run line by line in .sh file

1. Installation & setup

I recommend you use python 3.7, 3.8 or 3.9.

1.1 With conda

bash run/setup_conda.sh

1.2 With virtualenv

After you create an enviroment by using virtualenv. You have to remove line dlib==19.22.99 in requirements.txt. Then run command bellow in virtualenv enviroment.

bash run/setup_no_conda.sh

Note: To download dlib for python when not using conda you need to follow these instructions:

First: following this solution, you need to see this issue before running

2. Install model for dlib

You need put model file weights in folder models by downloading the shape-predictor-68-face-landmarks.dat here

Usage

1. Face swapping picture to picture

Source image: https://blog.icons8.com/articles/face-swapper-swap-faces-like-never-before-for-free/

• To use this function, you need at least 2 pictures, each contains only 1 face.
• Put your pictures into the 'images' folder to use them.

python utils/face_swap_picture.py --source_image_name=<SOURCE_IMAGE_NAME> --des_image_name=<DES_IMAGE_NAME>

In this:

• source_image_name: the name of picture you take the face from(E.g. Mickey.png)
• des_image_name: the name of picture you stick the face into(E.g. Donald.jpg)

2. Face swapping real-time with camera

• To use this function, you need at least 1 picture which contains only 1 face.
• A webcam(of course)
• Run the command below:

python utils/face_swap_realtime.py --image_name=<IMAGE_NAME> --video_name=<VIDEO_NAME> 

In this:

• image_name: the name of picture you take the face from(E.g. Jack.png)
• video_name: the name you choose for the output video file (E.g. Funny_face_swaping_clip)

The loop would run until you press the key 'p'