A computer vision-based system that detects driver/user fatigue in real-time using facial landmark analysis. The system monitors eye closure, yawning, and head posture to determine fatigue levels without requiring ML model training.
- Real-time face mesh detection using MediaPipe
- Side-by-side visualization: Normal view + Analyzed mesh view
- Multi-metric fatigue analysis:
- EAR (Eye Aspect Ratio): Detects eye closure
- MAR (Mouth Aspect Ratio): Detects yawning
- Head Pose: Monitors head tilting/drooping
- Color-coded feedback: Green = Normal, Red = Fatigue indicators
- Fatigue level meter: 0-100% visual indicator
- No training required: Rules-based detection
| NORMAL STATE | FATIGUE STATE |
|---|---|
 |
 |
| EAR: 0.32 (Open) | EAR: 0.18 (Closed) |
| MAR: 0.41 (Normal) | MAR: 0.72 (Yawning) |
| Head Tilt: 5° | Head Tilt: 22° |
| Fatigue: 15% | Fatigue: 85% |
| Status: AWAKE | Status: DROWSY |
