BackIndependent Doors
Door AI
AIComputer VisionPythonReact

Independent Doors · 2024–2025

Door AI

An AI system that reads architectural floor plans and automatically detects, classifies, and counts every door — outputting a complete quote in under 30 seconds.

The Challenge

Independent Doors' estimators were spending 1–2 full days manually reviewing floor plans for every large project — counting doors, identifying types (hinged, bi-fold, sliding, cavity), and pricing each one. With dozens of quotes per month, this was a serious bottleneck and a source of human error.

The Solution

We built a full-stack AI pipeline: a YOLOv8 computer vision model trained on 1,400+ annotated architectural drawings, capable of detecting 12 distinct door types. A web portal lets estimators upload a PDF floor plan, see every door highlighted and classified on-screen, and receive an auto-generated itemised quote. The backend runs on Hugging Face, frontend on Vercel.

Results

  • Automated a 1–2 day manual process down to ~30 seconds
  • Detects 12 door types with ~65% accuracy (targeting 85%+ with ongoing training)
  • Web portal built and deployed for production use
  • Roadmap: direct integration with Infusion accounting software for one-click quoting

Details

ClientIndependent Doors
CategoryAI / Computer Vision
LocationChristchurch, New Zealand
Year2024–2025

Stack

PythonYOLOv8 / UltralyticsReactNext.jsVercelHugging FaceNode.js

Work together

Have a project in mind? Let's talk.

Get in touch →