Machine Learning / AIThesis & Telemetry
Visit LiveCNN Facial Recognition for Sleep Deprivation Detection
Deep learning computer vision system classifying early physiological fatigue markers using CNNs and facial landmark telemetry.
Click for Full View
1 / 2
Live facial landmark mesh and ocular aspect ratio telemetry pipeline.
Project Overview
Engineered as an applied machine learning thesis system for Lyceum of the Philippines University - Cavite (LPU-C), this platform detects and classifies early physiological markers of sleep deprivation and cognitive fatigue in students.
Key Capabilities & Features
- Real-time ocular telemetry tracking Eye Aspect Ratio (EAR) and blink frequency via MediaPipe FaceMesh.
- Multi-class CNN classification model categorizing fatigue states (Alert, Moderately Fatigued, Severely Sleep Deprived).
- Yawn detection telemetry using dynamic Mouth Aspect Ratio (MAR) thresholds.
- Secure clinic portal with longitudinal student health tracking, role-based access control, and telemetry logging.
Architecture & Engineering Highlights
- [1]94.8% test accuracy across controlled multi-angle lighting environments with OpenCV preprocessing.
- [2]Edge-optimized landmark extraction minimizing latency and preserving student video privacy.
- [3]Automated telemetry alert flags when persistent microsleeps are identified during monitoring sessions.
Built With
Computer Vision & ML
Python 3.10TensorFlow / KerasOpenCVMediaPipe
Web & Telemetry
Next.jsReactTypeScriptTailwind CSS
Backend & Security
Node.jsREST APIsRole-Based Auth (RBAC)
Deployment
Cloud Linux VMNginx Reverse ProxySSL / TLS