This dissertation was written as part of the MSc in Mobile & Web Computing at the International Hellenic University. The purpose of this paper is the understanding of the deep learning theory, including terms like Neural Networks, activation functions, loss functions, optimizers, overfitting, data augmentation, and more, having as a final objective the development of a TensorFlow (TF) model that can recognize characters of the English alphabet.
This model is then converted into a compatible lite version in order to be imported into an Android application that was developed for this thesis. The user can then capture images through the app’s interface by using the phone’s camera, or by using a custom-made and programmed smart IoT remote camera (R-CAM) through Bluetooth, to feed into the TensorFlow model and present a prediction on the phone’s screen.
The thesis was completed successfully within the predetermined deadline. Of course, there is always room for improvement as determined in the last chapter of this thesis.
Collections
Show Collections
