Room Detection & Segmentation
Automating room detection & segmentation from architectural floor plans using custom computer vision and spatial reasoning.
1. The Problem
Extracting enclosed room geometry, room names, and precise net area measurements requires tedious CAD tracing or manual polyline creation.
2. Input Drawing
Architectural floor plan PDF containing wall lines, room text labels (e.g. Conference Room 204), and dimension strings.
3. AI Pipeline Process
Geometric wall graph construction closes bounding wall polygons while 360° OCR associates room text labels with their enclosing polygon area.
4. Annotated Visual Output
Color-coded room polygon overlay highlighting every enclosed space with calculated area callouts.
5. Structured Data Export
JSON geometry payload containing room names, room numbers, net floor areas (sq ft / m²), perimeter lengths, and polygon vertex coordinates.
6. Business Impact
Powers automated indoor mapping, real estate area calculations, flooring takeoffs, and PropTech spatial APIs.
Frequently Asked Questions
How does the system handle open-concept floor plans without physical walls?
We apply virtual boundary heuristics based on column lines, floor transition symbols, and ceiling plan divisions.