ChauffeurNet was trained with 30 million expert driving examples using imitation supervised learning. The model inputs are the top-down images we defined in the Input and output representations section, as illustrated in the following flattened (aggregated) input image:
The image is best viewed in color. Source: https://arxiv.org/abs/1812.03079
Next, let's look at the components of the ChauffeurNet training process:
ChauffeurNet training components: (a) the model itself, (b) the additional networks, and (c) the losses. Source: https://arxiv.org/abs/1812.03079 ...
