Geometric Extraction from Architectural PDFs #1478
Description
Hi everyone 👋,
I’m developing a prototype for automated quantity takeoff from architectural floor plan PDFs / BIM outputs (IFC). The goal is a system that extracts accurate, auditable geometry suitable for Quantity Surveying workflows.
Core deliverable
The system should programmatically extract:
• Walls: accurate linear lengths and thicknesses (mm)
• Openings: detect doors/windows and derive true W × H dimensions (mm)
• Auditability: coordinate-based “proof” (SVG/GeoJSON overlay or similar) mapping every measurement back to the source vector primitives / geometry
Technical challenge
The priority is a vector-first measurement engine, working primarily from vector PDFs (and/or IFC geometry), by parsing:
• paths / strokes / line segments / arcs
• transforms, scaling, and page coordinate systems
Key hurdles I’m looking to solve:
1. Scale calibration: reliably convert PDF units → millimetres using reference dimensions and/or scale bars
2. Geometry reasoning: distinguish true wall/opening geometry from noisy overlapping layers, dimension strings, and non-standard symbols
3. Inconsistent exports: handle the messiness of drawings exported from different versions of Revit/AutoCAD (including occasional “Not to Scale (NTS)” notes)
Relevant Tooling
I am currently exploring stacks involving PyMuPDF/pdfminer for extraction and Shapely/IfcOpenShell for spatial reasoning.
What I’m looking for
I’m looking for an engineer who has successfully tackled vector PDF / IFC geometry extraction + measurement calibration for takeoff-type workflows.
If you have a preferred approach for:
• coordinate transforms and geometry normalization, or
• scale calibration logic, or
• robust wall/opening extraction from vector primitives / IFC
I would really appreciate your thoughts and I'm also open to a deeper technical discussion and possible paid collaboration.
Thanks