RealTag
A visual tagging system where users annotate real-world objects and regions with custom context -- names, notes, care instructions -- and the app tracks them as they move.
Client
RealTag
Scope
Region annotation · custom context · object re-identification
Stack
Mobile app · segmentation overlays · vision backend
A camera app with on-screen annotation, a Python backend for detection and re-identification, and a data model that stores user-defined tags separately from model output.
Problem → delivery
Standard object detectors only recognize fixed classes. Users needed to tag arbitrary real-world things and attach their own meaning to them, not just detect what the model already knows.
A camera app with user-defined region annotation: mark objects, attach custom context, and let the backend handle re-identification as objects move in and out of frame. Segmentation and bounding flows handle spatial accuracy.
Annotation data and re-identification logic live on the backend, so the model can be improved without a client update. The phone handles camera capture and overlay rendering; the server handles the heavy spatial matching.
Built for real-world use
Fixed detector classes are limiting when the real need is user-defined semantics -- tagging a specific piece of equipment, not just detecting a generic object. The architecture separates what the model sees from what the user means.
Technologies
Tools grouped by role so you can scan without drowning in abbreviations.
Mobile app
Vision backend
Got a project in mind?
Tell us what you need to ship. If it’s a fit, we’ll talk scope and timeline.
Get Free Consultation