A small project for my graduate-level graphics course. 

​

I decided to build the "Hello World" of image classification by tackling the MINST dataset. The project was built using the keras library (in Python) using a Theano backend. The code consists of two different neural network implementations: a simple backprop and a convolutional neural net (CNN).

​

Both nets use one-hot encoding, as well as keras functionality including:

- Dense and Sequential model layers

- Dropout

- built-in SGD and activation functionality

- model fit & train functionality.

 

The CNN additionally uses: 

- Flatten, Convolutional2D and MaxPooling2D model layers

​

All the code is clean, well commented and tested.