place aligned/unaligned face images in dataset folder under a folder name which will same as ur label.
then run encode_faces. (This will take time depending on GPU/CPU!)
This will train the model with ur face data and generate a new encodings.pickle.
A pickle file is already uploaded trained with dataset also upoaded.
If encodings.pickle exists then face_capture can be run.
It will capture video from primary camera and detect faces logging entering and leaving time.