This was for an assignment as part of my computer vision module at university.
A convolutional neural network to classify images from the Fashion-MNIST dataset. On average, the network is able to achieve an accuracy of ~92, up from ~89% on the original.
This project is best viewed on
Google Colab.
If this is being run on Colab, you will likely need to
change the runtime type. This can be done by going to Runtime
> Change runtime type
> GPU
. Otherwise, the
training will take hours!
- Clone the repository
- Install required libraries (if not using Colab) - requirements
- Train the network
- Look at the results!
- tensorflow
- numpy
- matplotlib
- scikitlearn
- jupyter
To install these locally, run the command
pip install tensorflow numpy matplotlib scikitlearn jupyter
in terminal. This can also be done using conda.
This is dual-licensed under MIT and APACHE licenses.
Credits to @radenjezic153 for building the basic framework which I was able to improve upon.