Autoencoders are neural nets with a an encoder-decoder architecture. The encoder goes to a bottleneck; then the decoder decodes it back into an image.
We have implemented 3 autoencoders here:
- Without a fully connected layer
- with fully connected layer
- Variational Autoencoder
It has not been bundled but will download with the code. Else,
create data
folder and put MNIST
and FashionDataMNIST
into the folder.