roomgraph

A vector floor-plan PDF goes in. Rooms, areas, doors, windows and an adjacency graph come out.

Status: Alpha
License: MIT
Language: Python
Tests: 135

Install

pip install git+https://github.com/aec-platform/roomgraph.git

Not on PyPI yet — this installs from source.

Run

roomgraph extract plan.pdf -o out/

Most plans that reach a reviewer are PDFs, not models, and the geometry is right there in the file — it just has no rooms in it. roomgraph reconstructs them: closed regions, areas, openings, and the graph of what connects to what, as JSON, GeoJSON and a rough IFC. The PDF reader, the geometry, the IFC writer and even the animated GIF in its README are all in the package, in the standard library.

What makes it different

Room reconstruction from vector geometry — no OCR, no raster tracing, no model required
JSON, GeoJSON and a rough IFC from one extraction
Zero dependencies, down to the GIF encoder

Source →