NeuroDrive: Real-Time Driver Drowsiness & Cognitive Load Monitoring
🚗 NeuroDrive: Stay Awake. Stay Alive.
The problem: Drowsy drivers often close their eyes for 2–3 seconds without realising it. At highway speed, that’s enough to cause a crash, and most cars have nothing to warn them.
Our solution: NeuroDrive watches the driver’s face through a normal laptop webcam and sounds a loud alarm before drowsiness becomes dangerous. No extra hardware, no installation, and no video ever leaves the device. The AI runs fully inside the browser.
How it works
- Webcam streams live video in the browser
- MediaPipe Face Landmarker maps 468 points on the driver’s face every frame
- We calculate six real signals from those points:
- Eye closure (EAR): eyes counted as closed when the Eye Aspect Ratio drops below 0.21
- Blink rate: blinks per minute
- Yawns: mouth stays wide open for over 1 second
- PERCLOS: % of the last 60 seconds with eyes closed
- Head nods: sudden forward drop of the head
- Looking away: face turned away from the road
- These combine into a Drowsiness Score (0–100) and a Cognitive Load level
- If eyes stay closed for 2 seconds or the score passes 70, a full-screen red alert appears and an alarm keeps ringing until the driver clicks “I’m Awake”
✅ How to test it (30 seconds)
- Open the app in Chrome and click Allow for the camera
- Sit facing the camera in good light; you’ll see your live EAR value on screen
- Close your eyes for 2 seconds and the alarm goes off
- Try a big yawn or nodding your head down and watch the score rise
Tech stack
React · TypeScript · Vite · MediaPipe Tasks Vision · Recharts · Web Audio API · Tailwind CSS
What’s next
Mobile app, night-time support with infrared cameras, and a live alert dashboard for fleet owners.
Built with
geminiopenai
AI tools
Coded withGoogle AI StudioClaude Opus 5.5
Prize Categories Entered
- Submitted

Best Open-Source AI Project