FashionMnist repository contains code to classify the fashion-mnist dataset through kaggle and
their dataset. NOTE: This means we do not have the labels for the test set, which makes this project a little different than
most fashion mnist projects, since they can test their model much easier, we get 5 submissions per day and the accuracy result
is only from 30% of the actual test data. Right now the training is being done in python using keras and a tensorflow
backend. Plans to experiment with Matlab’s gpu computing toolkit are in the future. See below for more information on the project.

• Keras Documentation
• Competition Leaderboard

Here are some things you’ll need to get started.

1. Cuda 9.0 Toolkit
2. Python 3.6.2
3. cuDNN v7.4.1 for CUDA 9.0
4. Pycharm IDE

1. numpy: pip3 install numpy
2. tensorflow with gpu support: pip3 install tensorflow-gpu
3. keras: pip3 install keras
4. python imaging library: pip3 install Pillow
5. Process and system utilities library pip3 install psutil
6. Kaggle API library pip3 install kaggle

The csv files are too large for us to put in the repository.

1. Create a folder called Data
2. Download the dataset here
3. Put train.csv and test.csv in the Data folder you created.

Here are some of the descriptions for each of our classes.

This class is used to read in all the files necessary for the training including the training images, testing images, and CSV files, which are the tasks handled by each function from within the class. Once the training has been completed, this class generates the CSV file containing the resulting data from the training session.

The functions for processing the images utilize matrix operations to generate the new training sets through normalization, shifting flipping, and shuffling the images, and adding contrast to the given data sets.

Used to save data augmentation and analyze the data.

Holds the code to train, load, and build networks.