For people with limited mobility, the keyboard itself can be the barrier. This project removes it: a real-time computer-vision system that lets someone type and control a computer using nothing but their eyes.
It streams webcam frames through OpenCV, isolates each eye and locates the pupil using dlib facial landmarks, and classifies gaze direction (left, right, center) and blink state frame by frame, translating those signals into live keystrokes. A per-user calibration routine adapts the detection thresholds to each face and lighting setup so it stays accurate in real conditions.
The result is a working hands-free input system with direct accessibility applications, built entirely from classical computer-vision techniques and a standard webcam – no specialized hardware required.