All projects
Computer Vision Tool·2021

Facedapter

Real-time facial landmark detection and adaptation.

A computer vision tool for detecting facial landmarks in real time and applying adaptive overlays for AR-style experiences. Built as a web app using TensorFlow.js and WebGL rendering, running fully in-browser with no server-side processing.

Fa

Tech Stack

TensorFlow.jsWebGLTypeScriptReactCanvas API

Key Highlights

Ran 68-point facial landmark detection at 30 FPS in-browser using TensorFlow.js and WebGL acceleration.

Zero server round-trips — all inference happens client-side, preserving user privacy.

Built a custom WebGL renderer for adaptive overlay compositing on live webcam feed.

Screenshots

Main dashboard

Screenshot coming soon

Detail view

Screenshot coming soon

Mobile responsive

Screenshot coming soon

Screenshots will be added soon. Drop images into /public/projects/facedapter/