This is the code we used so far in Drowsiness Detection. For the code to run properly you need to feed it with the flags -o, which means output (it gives the path where the data will be droped after the main file runs), and then also provide -d with the path to the database.
You also need to download shape_predictor_68_face_landmarks.dat file which is a pre - pre trained model for facial landmark detection. The model can be found in the link below as well as a detailed explanation of what it does :
https://github.com/davisking/dlib-models
Then, once you have the data, you can run the notebook to get the results from the models.