Lane understanding is a foundational capability for safe and reliable automated driving — and it gets especially challenging when conditions are not perfect.

🌙 Nighttime driving
🌧️ Heavy rain and wiper occlusions
🛣️ Complex lane changes and unclear road edges

That is precisely where Ego Lanes comes in.

🚀 Ego Lanes is an AI-based, open source lane detection system that processes raw camera images and identifies:
🟣 The current right ego lane
🔵 The current left ego lane
🟢 All other lane markings and road edges

What makes it stand out?

🧠 Neural-network–based perception directly from raw images
🔄 Dynamically updates lane labels during lane change maneuvers
🌙 Robust performance in low-light and nighttime scenarios
🌧️ Reliable detection under heavy rain, occlusions, and visual noise
📷 Requires only a single front-facing camera
⚡ Runs in real time on low-power embedded hardware

These capabilities unlock core ADAS and autonomy features such as:

Lane Departure Warning
Lane Keep Assist
Hands-Free Driving

Ego Lanes is designed as a production-oriented, open source building block for automotive OEMs and Tier-1 suppliers — developed openly, validated transparently, and improved collaboratively.

🙏 A big thank-you to Tran Huu Nhat Huy for his open source contributions in developing and training the Ego Lanes neural network.

🧩 Ego Lanes is developed within the Autoware Privately Owned Vehicle (PoV) Working Group, where we build and validate open source AI models for real-world autonomous driving — together with the global community.

💬 Interested in testing, contributing, or joining the PoV WG?
Like, comment, or reach out — we’ll get you onboard.

🔜 More open source AI models are coming soon. Stay tuned.