Skip to content

ssocolow/autoencoder-walks

Repository files navigation

autoencoder-walks

A visualization of walking around the latent space of an Autoencoder trained on MNIST.

Contributors:

  • Milo Chang
  • Jamie Tang
  • Simon Socolow

For example, you can see what the model thinks is in between an 8 and a 9: demo pic

Quickstart

python3 UI.py will start the demo, which loads the pre-trained model from newmodel.pt and ten digits from firstTenDigits.pt.
The code for the model can be found in model.py, and uses an architecture inspired by this notebook.

Notes:

  • If you start at a 1 and go towards a 3, it won't look very good. But if you take a few steps towards a 4, then go to a 3, you can get a much better 3.

Colab Notebook

Check out our notebook to learn more about Autoencoders and a hands-on denoising application! colab notebook pic

About

Deep Learning Final Project

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors