-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Keras is an API written in python that runs on top Tensorflow, CNTK & Theano. It is a high-level neural network API designed to allow for easy and fast prototyping that runs seamlessly on CPU and GPU.
The model that is used to train the neural network from the train and test NMIST dataset is a Keras model and in this project, their model was adapted to train the network. This can be found here
Tensorflow is an open-source platform for machine learning, it ties in with Keras as Keras sits on top of it.
Flask which is a light web framework written in python is used in this project for serving the web application where a user can navigate to localhost:5000 and draw a number between 0-9 into a canvas and click send. This will make a request to the Flask backend API which will then process the data to then use it to make a prediction of what the number could be from the trained model saved within this project.
Jupyter is another light web application that allows the creation of documents that contains live code, visualization, markdown, and equations. This was great for this project as documenting code segments was needed as well as chart visualization. It allows for cells that were code to be executed one by one and over and over again without affecting other cells.
This python package was used to save models after training as well as allowing them to be loaded into memory when the model needed to be used to make a prediction from data that was sent from the Flask web application.