NVlabs/physical_ai_av

Devkit and documentation for the NVIDIA Physical AI Autonomous Vehicles Dataset

What it solves

It provides a standardized way for researchers to access and work with the NVIDIA Physical AI Autonomous Vehicles Dataset, which is a massive, geographically diverse collection of multi-sensor data designed to help build end-to-end driving systems.

How it works

The project provides a Python developer kit (SDK) and documentation, including a wiki and interactive notebooks, that allows users to download and process the multi-sensor data hosted on Hugging Face after agreeing to the NVIDIA license agreement.

Who it’s for

Autonomous vehicle (AV) researchers focusing on Physical AI and end-to-end driving systems.

Highlights

  • Python developer kit for streamlined data access.
  • Integration with Hugging Face for dataset hosting.
  • Includes interactive notebooks and a wiki for guidance.
  • Supports one of the largest and most geographically diverse multi-sensor AV datasets.