DriveSafe: Edge-AI Driver Monitoring System | Final Year Project Demonstration

DriveSafe: Final Year Project Video Demonstration
Submitted for the BEng (Hons) Software Engineering - University of Westminster.

Project Overview:
DriveSafe is a Proof-of-Concept, cross-platform mobile application designed to detect driver fatigue and prevent accidents. Unlike standard camera apps, DriveSafe utilizes a Multi-Model Edge-AI architecture, seamlessly switching between Google ML Kit (2D computer vision) and Apple ARKit (3D TrueDepth infrared sensing) to monitor the driver's biometric state in real-time.

All inference is executed 100% on-device to ensure strict data privacy.

Key Technical Features Demonstrated:
• Personalized Calibration: Dynamically establishes resting baselines for Eye Aspect Ratio (EAR), Mouth Aspect Ratio (MAR), and Head Pitch.
• Temporal Fusion Engine: Tracks fatigue over a 60-second rolling PERCLOS buffer rather than relying on error-prone instantaneous frame checking.
• Sentinel Occlusion Handling: Concurrently runs an Image Labeling neural network to detect dark sunglasses, successfully overriding 'contour hallucination' and dynamically shifting the scoring weight entirely to head posture and yawning.
• Multi-Sensory Alerting: Triggers maximum-intensity haptic and audio feedback to jolt the driver awake during a microsleep.

Video Chapters:
0:00 - Introduction & Personalized Calibration
0:45 - Normal Monitoring & Yawning Detection (Warning Level)
1:15 - Microsleep & Head Droop Detection (Critical Level)
1:45 - Edge-Case: Sunglasses Occlusion Handling (Multi-Model AI)
2:30 - ARKit TrueDepth Toggle & Conclusion

Student: Chaniru Mannapperuma
Student ID: W2051988
University: University of Westminster
Course: 6COSC023W - Computer Science Final Project

(Note: This video is submitted in partial fulfillment of the requirements for the BSc/BEng Computer Science/Software Engineering degree).

Видео DriveSafe: Edge-AI Driver Monitoring System | Final Year Project Demonstration канала Chaniru Mannapperuma