2025 — System design, vision, Edge AI, IoT and monitoring backend
IA-Powered Driver Fatigue Detection & Intervention
Real-time embedded system on Raspberry Pi 5: computer vision, Edge AI, alerting and automatic intervention.
Context
Embedded engineering project to detect driver fatigue from a camera feed, then trigger alerts and actions (sound, light, relay, GSM, GPS) without relying on the cloud for inference.
Build
- MediaPipe Face Mesh pipeline for eye closure, yawning and head orientation.
- Quantized INT8 TensorFlow Lite CNN models for eye-state and mouth-state classification.
- Fatigue scoring, temporal smoothing, LEDs, buzzer, relay, RFID, GSM/SIM800L and GPS.
- Monitoring stack: Node.js, MySQL, Firebase/Firestore, REST services and React.
Outcome
End-to-end chain: camera capture, Face Mesh, INT8 TFLite CNNs, smoothed fatigue scoring, actuators, cloud sync and a real-time React monitor.