Utility Scripts and systemd unit files for the Flunkey Mirror
In the folder face-rognition you will find the scripts to use flunkeys face recognition. The recognition works with a raspberry pi camera connected via the PI's camera interface.
The implementation is separated in two parts: Model training and Face Recognition itself. While the model training process the script takes all .jpg files in known_faces folder and generates numpy arrays with unique face encodings. It is important to give the image files correct names based on the schema Name.Surname.jpg.
The script will take the file name, split it in variables and publish it via flunkey-api as a user object with an initial config and the generated face_encoding as an attribute in the database. The face_encoding give us the opportunity to compare each other and on match to determine which face in front of the camera.
For debugging and as a break in protection the face detection service will take 8 Sek video and save it to /tmp if the face is unknown. In future releases it will do it only if the MONITORING_DEVICE is not online and the face is unknown.
All software dependencies are defined in debian.packages and python.packages. Please install them with apt-get and pip3 and be patient. The dlib will need on a Pi4 about 20-30 Minutes.
In system/unit_files you will find all unit files for running flunkey. The units require that all the flunkey components are checked out in /home/pi/flunkey/. For this purpose you could check out the global Flunkey Mirror Repo with all the components as git submodules.
Then you can copy the unit files from system/unit_files to /etc/systemd/system/, run the systemctl daemon-reload command and start the components in order that flunkey-api starts first.
Don`t forget to enable them for autostart.
Copy the the LXDE-pi/autostartfile to /home/pi/.config/lxsession/LXDE-pi/autostart to start flunkey glass automatically.