GitHunt

Vision Pilot - ADAS and Self Driving

VisionPilot

License
Discord
GitHub commit activity
GitHub Repo stars

PyTorch
OpenCV
ROS
LinkedIn
YouTube
Website

⭐ Star us on GitHub — your support motivates us a lot!

Free and fully open-source stack for ADAS and Self-Driving

This codebase contains productionizable and safety certifiable code for an open-source ADAS and Full Self Driving system that can power safe autonomous driving around the world. It is designed to be integrated with automotive OEMs and Tier-1 suppliers in series production passenger cars. We offer the complete codebase as free and fully open-source, including the release of all of our model architectures, data parsing scripts, training pipelines and model weights to help democratize access to this vital technology. All of our technology is available under the persmissive Apache 2.0 licence and can freely be used for both commercial and research purposes.

To learn more about how to participate in this project, please read the onboarding guide

End-to-End AI Architecture

We utilize an End-to-End AI Architecture in which each component technology of the Vision Pilot system is powered through neural-networks. Over the evolution of the project, we are introducing component-based/modular End-to-End AI, full monolithic End-to-End AI, as well as a new Hybrid End-to-End AI framework for optimal system explainability, introspection, verification and safety validation.

No Reliance on High Definition Maps

We do not require 3D high definition maps. Our stack can operate in a 'mapless' mode or use existing 2D navigational (sat-nav style) maps, also called ADAS maps. Human beings can drive without prior knowledge of the detailed 3D geometry of roadways by relying on real-time scene perception, and our system mimics this process.

Technology Roadmap

Rather than going for a single moonshot, we are sequentially building the core software stack of the Vision Pilot system in a modular manner, and releasing versions of the Vision Pilot system that serve a core customer need, where each new release expands upon the operational-design-domain, technology prowess and safety envelope of the prior release.

As part of our technology roadmap, we will sequentially develop three versions of the Autoware Autonomous Highway Pilot System, these are:

Vision Pilot

Models

The Models folder contains all code related to AI model development

Modules

Specific code modules for features such as Software-Defined vehicle implementaiton, Simulation, Control etc. can be found in the Modules folder.

Languages

C++46.4%Python42.9%Jupyter Notebook5.5%CMake2.7%Shell1.9%Just0.3%Cuda0.2%Dockerfile0.1%
Apache License 2.0
Created September 24, 2024
Updated March 11, 2026
autowarefoundation/autoware_vision_pilot | GitHunt