States are image and action are continuous values (steering, gas, brake). We will use the expert data from the human player.
Skip this step if you want to use the provided expert data. Run the following command to collect your own human demonstration.
python collect_human_carracing.py --n 5
Data will be saved in expert_data folder.