Face authorization

Authorization via face detection in python using some Machine Learning approaches and image processing

Dependencies

source setup_env.sh
python FaceApp.py

Description

The aim in our project is to give a user a valid face authorization tool. It should be user friendly and intuitive. Therefore, there is a GUI giving the instructions to follow. There are two different perspectives:

The user's perspective:

Choose in a GUI between options: 
	- Train the model
		- The program will ask the user to move his/her head in a specific way (e.g. straight/up/down/left/right)
	- Authorize using the trained model
		- The program will try to detect if your straight looking face is the authorized one and if it's a real person by a blinking test
		- If yes:
                - Ask the user to follow given head movements
                    - If passed, authorization successful
	- there is a live stream underneath (test if webcam is working)
	- there is a stream with some techniqued applied (head pose, orientation, etc.)

From the technical perspective:

- Region of interest (ROI) detection (dlib)
- Feature extraction strategies 
    - PCA and SVM in order to test face recognition when standing straight (many image samples takend and compared within a database)
    - SIFT to test left and right head orientations and how many features matched among them
    - FR: [face_recognition](https://pypi.org/project/face_recognition/)   tested with [lfw dataset](http://vis-www.cs.umass.edu/lfw/#download)
- Blink detection (OpenCV)
- Head orientation detection (deepgaze)