Read Data from Dataset

Setting up Argoverse dataset

To run the example, you need to install the argoverse-api and download the map files.

Install argoverse-api repo

Clone the official argoverse API repo to some place then install it:

git clone
cd argoverse-api
pip install -e .

Download and extract argoverse map files

cd argoverse-api
tar -xzvf hd_maps.tar.gz

After unzip the map files, your directory structure should look like this:

└── map_files
└── license
└── argoverse
    └── data_loading
    └── evaluation
    └── map_representation
    └── utils
    └── visualization

Enjoy Driving in Real Town!

You can launch a script to drive manually in a town in Pittsburgh with replayed traffic flow recorded in Argoverse. Note: Press T to launch auto-driving! Enjoy!

# Make sure current folder does not have a sub-folder named metadrive
python -m metadrive.examples.drive_in_argoverse

Specify the Data to Replay

MetaDrive currently supports replaying the map and traffic flow in the sample dataset of argoverse-tracking. We will add more scenarios for customized selection in the near future. As shown in ArgoverseEnv Class, we use a few parameters to specify the data to be loaded. Here is the detailed explanation of those parameters:

  • argoverse_city: The shortcut of the specified city.

  • argoverse_map_center/radius, radius: Only the roads and traffic within the circle centering in argoverse_map_center with radius argoverse_map_radius will be loaded.

  • argoverse_spawn_lane_index: Node index indicating where the ego agent is initialized.

  • argoverse_destination: Node index indicating the destination of the ego agent.

  • argoverse_log_id: We select one sample of argoverse-tracking data with this ID as the folder name.