robust2notation

Comp 551 Kaggle project.

Setting up

1. Make sure you have Python 3 installed, as well as pip3.
2. Run setup.sh to make sure you have the correct modules installed.
3. Create a directory called data/ in the root of this project.
4. Place all of the data files (download link: https://www.kaggle.com/c/11461/download-all)
5. In the command-line navigate to this directory, and run jupyter notebook .

How to contribute

0. (Optional) make issue for task that you're creating
1. Branch off of develop
2. Write your code there.
3. When you are ready to push to develop, open a pull request
4. Go through PR process!

Some descriptions of files

Classical Machine Learning

Classical_Models.ipynb is a Jupyter Notebook containing various classical models for image classification. The model explored include:

• Linear SVM
• Random Forest
• TODO add: K-nearest neighbours

Numpy Neural Net

This is a fully connected, feed-forward neural net, implemented from scratch using only Numpy.

• homecooked_NN.py contains the Neural net class.
• handmade_NN.ipynb contains the Jupyter Notebok used in the training.

Convolutional Autoencoder

conv_autoencoder.ipynb contains a convolutional autoencoder, which was tested as a potential de-noising option.

Categories

• sink
• pear
• moustache
• nose
• skateboard
• penguin
• peanut
• skull
• panda
• paintbrush
• nail
• apple
• rifle
• mug
• sailboat
• pineapple
• spoon
• rabbit
• shovel
• rollerskates
• screwdriver
• scorpion
• rhinoceros
• pool
• octagon
• pillow
• parrot
• squiggle
• mouth
• empty
• pencil