AR Smart Mirror
Real-time interactive AR virtual try-on system designed for retail fashion.
The Problem
In-store clothing fitting is slow, creates long queues in fitting rooms, and increases garment wear. Retailers need an instant, hygienic, and engaging digital mirror that lets shoppers visualize outfits without undressing.
My Role
Built solo with AI-assisted development — computer vision pipeline design, MediaPipe pose detection integration, Canvas AR overlay engine, FastAPI backend services, Docker multi-container topology, and Cloudflare Zero Trust ingress.
Highlights
- Sub-30ms real-time body tracking and landmark extraction using MediaPipe Pose
- Dynamic Canvas AR garment warping and affine transformation for realistic clothing drape
- FastAPI backend with PostgreSQL catalog and Redis session pub/sub synchronization
- Mobile controller pairing allowing customers to browse catalogs and trigger try-ons from their smartphones
- Production-grade Docker Compose architecture with Nginx reverse proxy and Cloudflare Tunnel
Stack
Constraints
- Hard real-time latency target (<33ms per frame / 30 FPS) to avoid motion sickness and visual lag.
- Zero customer biometric persistence — video feeds must process in memory without saving raw footage to disk.
- Compatibility with standard consumer hardware (e.g. RTX 3060) without requiring specialized depth cameras.
System Architecture
Key Trade-offs
The decisions worth defending — what I chose, what I turned down, and why.
Pose tracking execution model
Chose
Client-side MediaPipe Pose on Canvas API with GPU acceleration
Rejected
Server-side video streaming and frame-by-frame inference
Server inference introduces network jitter, high bandwidth overhead, and privacy concerns streaming raw camera feeds over the network.
Garment deformation approach
Chose
Real-time 2D affine mesh deformation keyed to skeletal anchor landmarks
Rejected
Full 3D physics-based volumetric cloth simulation
3D cloth simulation cannot sustain 30+ FPS on edge kiosk hardware; 2D mesh triangulation provides fluid response with low computational cost.
Remote store interconnect
Chose
Cloudflare Tunnel over outbound HTTPS
Rejected
Direct port forwarding and dynamic DNS on store routers
Retail store Wi-Fi and ISP firewalls prevent inbound port forwarding; Cloudflare Tunnels punch through NAT securely without public IPs.
What I'd Do Differently
An honest retrospective — the stuff I'd change with more time, more users, or a second pass.
- 1Implement WebGPU shaders for cloth lighting and shadow blending to improve realism under dynamic store lighting.
- 2Add an offline fallback queue for store analytics so telemetry buffers locally during internet interruptions.
- 3Optimize landmark smoothing filters (OneEuro filter) to eliminate micro-jitter when shoppers stand at edge distances.
Computer Vision & Pose Tracking Shell
Benchmark the real-time AR try-on pipeline (status, camera init, pose detect, mesh align -shirt, latency bench).
Technical Deep-Dive
Architecture, specifications, and implementation details.
06 — Cloudflare Tunnel Setup
Access the AR Mirror from any network — no port forwarding, no static IP.
The tunnel ships as the ar-tunnel service in docker-compose.yml.
#Prerequisites
- Cloudflare account (free)
ice4legends.comalready added to your Cloudflare account (it is)
#Recommended path — token-based, fully managed in the dashboard
This is what the repo is wired for. No config.yml, no credentials file on
disk — the routes live in the Cloudflare Zero Trust dashboard.
##1. Create the tunnel in the dashboard
- Open https://one.dash.cloudflare.com
- Networks → Tunnels → Create a tunnel
- Connector type: Cloudflared
- Name:
ar-mirror→ Save - On the "Install and run a connector" page select the Docker tab.
You'll see a command like:
Copy only the token — the long string after
docker run cloudflare/cloudflared:latest tunnel --no-autoupdate run --token eyJh…(long string)--token.
##2. Paste the token into .env
TUNNEL_TOKEN=eyJh...the long string from step 1...
Keep .env out of git — it's already gitignored.
##3. Add public hostnames in the dashboard
Still in the tunnel's setup page → Public Hostnames tab → add two:
| Subdomain | Domain | Service type | URL |
|---|---|---|---|
mirror | ice4legends.com | HTTP | ar-frontend:80 |
controller | ice4legends.com | HTTP | ar-frontend:80 |
Both point at the same nginx container — nginx routes by Host: header
(nginx/nginx.conf).
DNS records for mirror.ice4legends.com and controller.ice4legends.com
are created automatically by the dashboard when you save the hostnames.
##4. Start the tunnel alongside the stack
The tunnel is behind a Compose profile so the stack still starts cleanly when no token is set. Bring everything up with:
docker compose --profile tunnel up -d
docker compose ps
You should see all seven services Up, including ar-tunnel.
To stop just the tunnel and leave the rest running:
docker compose stop ar-tunnel
#Alternative — legacy config-file mode
Useful only if you already have ~/.cloudflared/<uuid>.json from running
cloudflared tunnel create on the host. To use it from a container, replace
the ar-tunnel service body with:
ar-tunnel:
image: cloudflare/cloudflared:latest
command: tunnel --no-autoupdate run ar-mirror
volumes:
- C:/Users/Matei/.cloudflared:/etc/cloudflared:ro
depends_on:
- ar-frontend
restart: unless-stopped
…and provide a config.yml next to the credentials file that uses
service: http://ar-frontend:80 (not localhost:80 — localhost inside
the container points at the container itself, not the host).
#Verify
Local: http://localhost
LAN: http://192.168.0.104
Public: https://mirror.ice4legends.com
Controller: https://controller.ice4legends.com
#Troubleshooting
| Issue | Fix |
|---|---|
| Tunnel offline | Run: cloudflared tunnel run ar-mirror |
| 502 Bad Gateway | Check: docker compose ps — all containers Up? |
| DNS not resolving | Wait 2 minutes after setup for propagation |
| Can't login | Check internet, try: cloudflared tunnel login --force |