This autoencoder projects uses the handwritten number dataset to create a model from. In essesence, it's like ML compression. There's two parts to the model, the Encoder, which reduces an input down to a vector. And a Decoder, which takes in input and brings it back to an image. In this case, we can join both of them and use the images as both the input and output. Then, we can try modifying the inputs to the decoder to see what happens!