Blink to Pause, Pinch to Click: A Browser-Based Gesture Remote

What it does The repo is the starter code for a JavaScript course project that layers computer vision over a fake streaming platform. It uses the webcam to detect eye blinks—translated to play/pause—and hand poses like open palms, closed fists, and pinches to scroll, hover, and click through a video catalog. All the machine-learning heavy lifting is intended to run inside a Web Worker so the main thread stays responsive.

The interesting bit Instead of treating the camera as a passive sensor, the project draws your hands directly onto the UI canvas while keeping underlying DOM elements clickable, essentially turning your body into a translucent input device. The checklist also hints at distinguishing left-eye from right-eye blinks, which suggests the ambition is to pack a surprising amount of input bandwidth into a single webcam stream.

Key highlights

• Built on MediaPipe Face Detection and Hands, plus TensorFlow.js, running in the browser.
• Maps specific gestures to UI actions: open palm scrolls up, closed fist scrolls down, pinch clicks, and a blink toggles playback.
• Designed to offload ML inference to a Web Worker to avoid janking the page.
• Includes daily class implementations and a live demo of the final state.
• Layout is a realistic streaming-service UI, borrowed from an existing CodePen design.

Caveats

• The feature checklist is almost entirely unchecked; this is explicitly a starter template for a week-long course, not a finished library.
• Several FAQ entries warn about WebGL and Windows compatibility issues, so expect some environment wrestling.
• The README is in Portuguese, which may slow down navigation for English-only readers.

Verdict Worth a look if you want a guided, hands-on introduction to browser-based gesture recognition. Skip it if you need a drop-in, production-ready input library.