Furniture manufacturing · Case study · 2 min read
A model that reads blueprints
We trained a model to read architects' blueprints and pull out what a furniture maker needs, which takes a slow, careful job off the manufacturer's desk.
- Type
- Machine learning
- Built with
- A custom-trained model
- Our part
- Built inside the client's team
The problem
A furniture maker who builds for architects receives drawings: floor plans, elevations, details. Before anything can be quoted or cut, someone has to read through them and work out what is actually being asked for: which pieces, how big, how many. It is careful, repetitive work, and it sits in front of every job.
What we built
A model trained on architects' drawings that finds the furniture in them and returns what the maker needs as a structured list, so nobody has to read it off the page line by line. The list is a starting point for a person to check, not a replacement for one.
The hard part
Drawings are not made for machines. Every office draws differently, the same cabinet appears in a plan, an elevation and a detail, and the information that matters is split between lines, symbols and notes. The work was in teaching the model to tell furniture from everything else on the sheet, and to bring the different views of one piece together.
What we can show
This was built for one manufacturer and the details are theirs. On a call we can walk you through how it works.