Contact person: Finish AR Nail Try-On Webapp

Phone:Show

Email:Show

Location: Johannesburg, South Africa

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I have a browser-based AR Nail Try-On prototype where MediaPipe Hands already tracks finger landmarks reliably. Now I need the missing piece: real-time nail segmentation.

Here’s the current state
• MediaPipe Hands is integrated and stable.
• Camera feed and canvas rendering pipeline are in place.
• UI scaffolding is done.

What I still need
1. Plug the Roboflow “seg_nail_test v1” semantic-segmentation model into the existing pipeline. The app must fetch masks frame-by-frame, merge them with the landmark data, and draw polished overlays that follow each nail naturally. High mask fidelity takes priority; a moderate frame rate is acceptable.
2. Make it work consistently on both mobile and desktop, with special attention to Chrome where most traffic comes from.
3. Remove all CORS headaches. If that means adding a lightweight Node/Express proxy or tweaking headers on the Roboflow calls, do it.
4. Hide the Roboflow API key securely—no hard-coding. An .env file, serverless function, or another secure strategy is fine as long as the key never leaks in client code.
5. Ship clean, well-commented code plus a short README so I can redeploy or extend the project easily.

Acceptance criteria
• Nail masks align correctly on at least 90 % of frames in good lighting.
• End-to-end latency ≤ 250 ms on a mid-range phone (Pixel 6 or similar).
• No console CORS errors or exposed keys when audited with browser dev tools.
• Works out of the box after npm install && npm start.

Tech already in use: vanilla JS, MediaPipe Hands, WebGL canvas. Feel free to add [login to view URL], [login to view URL], or other helper libs if they lighten the lift." (client-provided description)