there are a lot of changes here but they can be broken down as follows:
structural: the repo is now structured as a proper python package with subpackages and modules. This means that for now the OpenCV pipeline has been moved to a directory within the main package.
docstrings: have been added and I will add more comments as I go
calibration: still done 'automatically' but users can now control the pace through keyboard input of esc key
keystrokes: added pynput method for returning keystrokes of predictions
model: the new baseline model I am calling eyenet3. It is based on MobileNet, takes in slightly larger color images but performs much better than the previous jupiter model.
classification: EyeCommander now makes a prediction on both the left and right eye but keeps only the prediction with the highest confidence. This seems to reduce a lot of noise and improves classification.
there are a lot of changes here but they can be broken down as follows: