AR Hand Canvas

Project Architecture & Overview

A touchless augmented reality application allowing users to draw in mid-air using their webcam. Powered by Google MediaPipe hands machine learning models to track 21 3D landmarks at 60 FPS, with intelligent pinch-to-draw gestures, color selection, and smooth Bézier stroke smoothing.

Key Engineering Milestones

• ✦Sub-millisecond 21-joint 3D hand tracking directly in standard web browsers
• ✦Gesture state machine: Index finger draws, two fingers select, open palm navigates
• ✦Quadratic Bézier curve interpolation eliminating stroke jitter and hand tremor
• ✦Instant PNG image export and multi-layer drawing palette