Fake Face Replace Space in Case of Chase we Race to Erase then Place, incorporated
aka FFRSCCREP, inc.
Unfortunately the model is too large to be stored on Github, so you'll have to download it from here and place it in models/FFRSCCREP/ for the program to work correctly.
To run the program:
main.py [path/to/img.jpg]
For example:
main.py TestImages/group.jpg
If you do not provide an argument path to an image, it will open the default image path defined on line 8 as defaultPath. If you provide an argument but it is not a valid path to an image, the program will display an error then exit.
After the program opens the input image, it will run facial detection and draw a green square over each detected face. Clicking in any of these rectangles will turn them blue. This will then become the selected face. After selecting a face, pressing G will generate a fake face and place it over the selected face. Note this will take a few seconds. You can press G as many times as you want until you get a face you like.
After a face has been generated, it will not be saved until you press the S key.
This will save the image in its normal path but with -modified added to the file name.
This will also clear the selected face.
After generating a face, but before saving it, you also have the following options:
- B: Increase brightness
- V: Decrease brightness
- C: Increase contrast
- X: Decrease contrast
- Y: Horizontally flip image
- R: Return generated face to before any modifications made
This project was inspired by and uses the same neural network (stylegan2) as thispersondoesnotexist.com.
We trained the model ourself with 70k input images over 150k iterations, taking over 10 days to complete.
Facial detection is achieved using a pre-trained Haar Cascade model.