Driver Drowsiness Detection System
Real-time computer vision system that monitors driver fatigue using dlib's facial landmark estimation, eye-aspect-ratio (EAR), and CNN classifiers.
Project Overview
A software system using OpenCV, dlib, and a custom CNN model to track facial landmarks in real time, detecting micro-sleeps and alerting drivers instantly.
The Problem
Driver fatigue is a leading cause of highway accidents globally. Active, low-latency detection systems are needed to warn drivers before accidents happen.
The Solution
Built a pipeline capturing frames from an in-cabin camera, locating the face and eye regions using facial landmark estimation, calculating EAR, and classifying frames with a light CNN when eye closure exceeds threshold limits.
The Impact
Maintained 30 FPS on low-power edge computing devices (Raspberry Pi 4) with a 94% validation accuracy on the National Tsing Hua University Driver Drowsiness Dataset (NTHU-DDD).
Date / Year
2025