Finds readout by searching for a rectangle of roughly the right size Uses fact that surrounding area is all white to improved acquisition
Divides readout into individual digits
Removes noise from digits and centers them
Classifies digits via tesseract ocr package and CNN trained on the MNIST dataset, using Tensorflow and Keras
TODO:
- Retrain network with more variation to better identify digits
- Add debug mode