A VALUED CUSTOMER OF THIRDEYE DATA
Biloba Architecture: generating house floor plans with AI
Biloba Architecture is a full-service architecture, interiors, and planning practice. Early-stage residential design absorbs a great deal of a practice’s time in a way that is hard to bill properly. A client describes roughly what they want, someone produces options, the client reacts, and the cycle repeats before the work that actually requires an architect begins.
LOCATION
Charlotte, North Carolina, United States
COMPANY SIZE
Small-scale Enterprise
INDUSTRY
Real Estate & Civil Engineering › Explore solutions for this industry
AI-generated floor plans
Biloba wanted to automate the generation of house floor plans. ThirdEye Data built the end-to-end system that does it, in three stages that map onto how the conversation with a client actually goes.
First the system recommends. Given what a client says they want, it proposes floor plans that fit those preferences, which replaces the blank page at the start of the process. Then it allows customisation, so the client can adjust what they have been shown rather than describing changes in words and waiting to see them. Finally it generates, producing floor plans from the full set of inputs.
The generation itself uses deep learning rather than a rules engine assembling rooms from a library. That distinction matters for the output. A rules-based generator produces plans that are correct and recognisably machine-made. A model trained on real plans produces something that reads as designed, which is what makes it usable as a starting point with a client rather than an illustration of a concept.
A working system covering recommendation, customisation and generation, in use as the front end of the residential design process.
Related solution: generative computer vision for design
Generative work in a design context has a harder test than generative work elsewhere. The output is judged by professionals who will immediately see what is wrong with it, and an implausible result does more damage than no result, because it costs the practice credibility with a client.
ThirdEye Data builds generative systems for work where the output has to survive that inspection. For architecture that means plans that respect how spaces actually connect, not just plans with the right number of rooms. We also build the surrounding path, because generation on its own is rarely the product: the system has to understand what was asked for, let a person adjust what it produced, and generate again from the amended brief. We have built this for residential design, and adjacent work extracting structured information from existing architectural drawings, which is the same problem in reverse.
Explore our generative AI solutions ›
DO YOU HAVE A SIMILAR USE CASE?
Talk to our team to learn more about what we did for Biloba Architecture and how that experience applies to your own environment.
Explore What We Did for Other Customers
