models are now classes. this abstracts their functionality from the eye commander class and is useful for changing and testing models.
face/eye detection is now done with tensorflow's mediapipe face mesh instead of their face detection algo. this is more precise
the baseline model has been significantly improved by training on a larger and better filtered dataset. this baseline model is a CNN and its class is called CNNModel.
calibration can now be triggered by hitting the 'c' key at anypoint during use
The significant changes to the main branch are: