A collection of different autoencoder types in Keras. It is inspired by this blog post.

Python is easiest to use with a virtual environment. All packages are sandboxed in a local folder so that they do not interfere nor pollute the global installation:

virtualenv --system-site-packages venv

Whenever you now want to use this package, type

source venv/bin/activate

in every terminal that wants to make use of it. Theano needs a newer pip version, so we upgrade it first:

pip install --upgrade pip

To install the dependencies, use pip:

pip install -r requirements.txt

If you want to use tensorflow as the backend, you have to install it as described in the tensorflow install guide. Then, change the backend for Keras like described here. Now everything is ready for use!

One can change the type of autoencoder in main.py.

python main.py

I currently use it for an university project relating robots, that is why this dataset is in there. Feel free to use your own!