Skip to main content

Welcome to Visage

Visage is a prototype assistive system designed to help users identify people in real-time using computer vision and face embeddings. This project demonstrates how identity assistance technology might work in future AR glasses for dementia support.

Visage System Overview

What is Visage?​

Visage combines modern computer vision techniques with a thoughtful user experience to provide real-time face recognition assistance. The system captures video from a webcam, detects faces, generates embeddings, and matches them against a local memory bank of known individuals.

Key Features​

Real-time Identification 👁️​

Continuous face detection and identification through webcam streams with sub-second latency

Local Memory Bank 💾​

All face data stored locally in SQLite — no cloud dependencies, ensuring privacy

Multiple Models 🧠​

Support for DeepFace models including Facenet, VGG-Face, and ArcFace

Modern Web UI 🪟​

React-based interface with real-time WebSocket updates and smooth animations

Use Case​

Visage is built as a technical demonstration for assistive technology that could benefit individuals with:

  • Prosopagnosia (face blindness)
  • Early-stage dementia or memory impairment
  • Social anxiety related to forgetting names and relationships
Medical Disclaimer

Visage is an exploratory technical prototype, not a medical device. It is not FDA-approved and should not be used as a substitute for professional medical advice, diagnosis, or treatment.

Technology Stack​

Visage is built with modern, production-ready technologies:

  • Backend: FastAPI with async/await support
  • Embeddings: DeepFace with TensorFlow backend
  • Storage: SQLite with vector similarity search
  • Frontend: React with Vite, TailwindCSS, and Framer Motion
  • Real-time: WebSocket connections for live video streaming
  • Deployment: Docker Compose for easy containerization

Architecture at a Glance​

Project Goals​

  1. Privacy-First: All data processing happens locally
  2. Real-time Performance: Sub-second identification latency
  3. Extensible: Easy to swap models, add features, or integrate with other systems
  4. Educational: Clear, documented code for learning about face recognition systems
  5. Ethical: Transparent about limitations and appropriate use cases

Next Steps​

  • Quickstart - Get Visage running locally in 5 minutes
  • Architecture - Dive deep into system design and data flow
  • API Reference - Explore REST and WebSocket endpoints
  • Ethics - Understand privacy, consent, and limitations
  • Future-Proofing - Learn about opt-in key rotation, metrics, and experimental privacy processors