Skip to main content

Getting Started with Project Aria Data Utilities

In this guide, we introduce how to install the projectaria_tools Python package and provide tutorials to go through the APIs to access and visualize Aria data.

You can run this tutorial via Google Colab or install projectaria_tools locally and use Jupyter notebooks.

This page covers Project Aria Tools core tutorials:

  • Dataprovider quickstart tutorial: using the Aria Data Provider to access sensor data from a VRS file. Content covered includes:
    • Retrieve image data
    • Summarize a VRS file using thumbnails
    • Obtaining sensor calibrations
    • Access project/unproject functionalities
    • Undistorting an image
  • Sophus Pybind Tutorial: access Sophus Library SO3, SE3, interpolate and iterativeMean features
  • Machine Perception Services (MPS) quickstart tutorial: a guide to visualizing MPS (derived) data. Content covered includes visualizing:

You can also explore the open datasets through tutorials:

Running Jupyter Notebooks on Google Colab

Use the following links to run the Python notebooks in an installation free playground:

Running the Jupyter Notebooks locally

Step 0 : Check system requirements and download codebase

Go to the Download Codebase page to:

  • Check your system is supported
  • Download projectaria_tools codebase from the github

Step 1 : Install/Update Python 3

To use the Jupyter notebooks in this tutorial you'll need Python 3.9 or above. To ensure all utilities work effectively, we recommend keeping Python 3 up to date.

Step 2 : Create a virtual environment

rm -rf $HOME/projectaria_tools_python_env

python3 -m venv $HOME/projectaria_tools_python_env

source $HOME/projectaria_tools_python_env/bin/activate

Step 3 : Install projectaria_tools from pypi

pip3 install --upgrade pip

pip3 install projectaria-tools'[all]'

Step 4: Run Dataprovider quickstart tutorial

cd $HOME/Documents/projectaria_sandbox

jupyter notebook projectaria_tools/core/examples/dataprovider_quickstart_tutorial.ipynb
Jupyter notebook error

If you get a Jupyter notebook error please upgrade Python 3 to the latest version and recreate your virtual environment.

Step 5: Run Sophus Pybind Tutorial

cd $HOME/Documents/projectaria_sandbox

jupyter notebook projectaria_tools/core/examples/sophus_quickstart_tutorial.ipynb

Step 6: Run Machine Perception Services (MPS) quickstart tutorial

In the MPS tutorial, the notebook walks through how to visualize gaze, trajectory, and point cloud from MPS data.

cd $HOME/Documents/projectaria_sandbox

jupyter notebook projectaria_tools/core/examples/mps_quickstart_tutorial.ipynb


Check the Troubleshooting Guide if you encounter issues using this tutorial.