The core task of any autonomous driving system is to transform sensory inputs
into driving commands. In end-to-end driving, this is achieved via a neural
network, with one or multiple cameras as the most commonly used input and
low-level driving command, e.g. steering angle, as output.