Train a convolutional autoencoder 10 points
Create a convolutional autoencoder that compresses the racing game screenshots to a
small number of bytes , and transforms them back to original .
Steps
1. Create and train an undercomplete convolutional autoencoder and train it using
the training data set from the first task.
2. You can choose the architecture of the network and size of the representation
h = f (x). The goal is to learn a representation that is smaller than the original,
and still leads to recognisable reconstructions of the original.
3. Explain the difference between an undercomplete and a de-
noising autoencoder.
4. The input images are 16×16 = 256 pixels. What is the size of
your hidden representation h = f (x) .
Include your calculation in your report.
What to submit:
• Submit the python code of your undercomplete autoencoder .
• For your report, write a brief description of your steps to create the model and your
prediction. Include the description undercomplete vs. denoising autoencoder, and
your calculations. How do you measure the quality of your model?
• Include screenshots of 1-2 output images next to the original inputs .
