Drowsiness Detection System
A real-time edge AI prototype combining computer vision and rapid inference to track driver fatigue and trigger preventative alerts.
Real-Time CV
Edge Deployment
OpenCV
Safety Alerts
Telemetry Overlay Simulation
Active Monitoring
Eye Aspect Ratio (EAR)
0.32
Fatigue Risk Score
12%
Frames Per Second
28.4
Logic & Signal Features
- Eye Closure Duration: Measuring the Eye Aspect Ratio (EAR) across sequential frames.
- Blink Frequency Trend: Analyzing the baseline blink rate against current rate to detect micro-sleeps.
- Decision Strategy: Sliding-window smoothing to minimize noise and threshold-based risk escalation.
- Cooldown Rules: Integrated logic to avoid alert spamming directly post-incident.
Edge Deployment
- Capture: Reads frames continuously at the edge device level (Raspberry Pi).
- Vision: Detects facial landmarks to extract fatigue-relevant bounding areas.
- Inference: Computes instantaneous drowsiness scores based on the EAR threshold.
- Alert: Hardware triggers audio/visual warning physically connected to GPIO pins.
Why Building at the Edge Matters
Cloud-based CV systems introduce network latency which is unacceptable for active safety systems. By porting lightweight inference models to the Edge, this system acts independently and instantaneously in low-connectivity areas.
Low Latency
High Reliability
Offline First