Every piece of furniture — an almirah, a work desk, or even a stool — we buy is either based on our seeing it somewhere or liking it because we once owned something similar. This is especially true for vintage furniture that we don’t find at stores very often, for they are full of the latest designs and products. Besides, you might also like a piece at a friend or relative’s house, but may not be comfortable asking where they got it from. Some of us might not want the same item but a slightly tweaked version of it. However, we may not know where to find it.

To tackle all this, a team of researchers from the University of Washington and Shandong University have devised a method called “Fabrication-Aware Reverse Engineering for Carpentry”. The team’s method proposes to generate fabrication blueprints from images of carpentered items. 

How do they do it?

Researchers said that an individual may like a piece of furniture but may not want the same product in terms of its dimensions. You might want to make it slightly bigger and broader, they said, and for that, take a few pictures from different angles of the object that you want. This reverse engineering involves taking a set of images of a carpentered object as inputs, followed by generating a CAD model, ready to build a replica of the object or an edited version of it.

“Our method makes use of domain-specific constraints to recover not just valid geometry, but a semantically valid assembly of parts, using a combination of image-based and geometric optimisation techniques,” read the abstract of the paper. 

They added that the team has demonstrated the method on an array of wooden objects and furnishings. The team said that they can automatically obtain designs that are not just easy to edit but also accurate recreations of the ground truth. They further show how reverse engineering for carpentry can be used to fabricate a physical replica of the captured object. Besides, a customised version of the product can also be obtained by directly editing the reconstructed model in a CAD software.

There are some challenges too. “Arriving at a fabricable solution requires identifying the parts, and optimising for their precise shapes and the part-to-part connections constraining those shapes” are only some of the challenges researchers have highlighted in their paper. 

The paper will be published in Eurographics Symposium on Geometry Processing.