Autoencoder Demo for Fresh Machine Learning #5
This code demos a simple autoencoder that is able to generate handwritten digit images after training against the MNIST dataset. This the code for 'Build an Autoencoder in 5 Min' on Youtube
- Python 2.7+ (https://www.python.org/download/releases/2.7/)
- Tensorflow (https://www.tensorflow.org/versions/r0.10/get_started/os_setup.html#pip-installation)
- numpy (http://www.numpy.org/)
Use pip to install any missing dependencies
Once you've downloaded the repository simply type the following into terminal
python main.py
The autoencoder will automatically start downloading the MNIST dataset, then training on it. When finished, it'll be able to generate it's own images. See this site for a live autoencoder demo in the browser
Credit for the vast majority of code here goes to the Tensorflow team. I've merely created a wrapper around all of the important functions to get people started.