AI News: Using GANs For Architecture
Subscribe:  iCal  |  Google Calendar
Cologne DE   20, Aug — 25, Aug
Vienna AT   23, Aug — 25, Aug
Anaheim US   27, Aug — 30, Aug
SEATTLE US   30, Aug — 3, Sep
Vancouver CA   4, Sep — 7, Sep
Latest comments

Oh shit!

by S.K.O
1 days ago

VUE without competition

Can you please give us a walkthrough how to implement this into Maya? would be super helpful. Thanks a lot.

AI News: Using GANs For Architecture
24 June, 2019

Let’s discuss another story from the AI world. Stanislas Chaillou, Harvard Graduate School of Design, has recently presented a paper which discusses the importance of AI for the architectural practice and shows how they built an entire generation methodology, using Generative Adversarial Neural Networks (GANs) for archviz and similar projects.

The work basically shows how you can use GANs for floorplan design with the help of visual plans as a data format for both GAN-models’ inputs and outputs. It’s all about saving time again — the project could allow architects and designers to save time on repetitive tasks, but we think that it could also be pretty great for level design tasks.The system here consists of two key models, the Generator and the Discriminator, that share feedback and generate relevant images. The Discriminator helps the system recognize images from a set of data, being able to distinguish between a real example and a “fake” image. The Generator, on the other hand, acts as a creator of new images that have qualities of the same dataset. So, it creates new images, and the Discriminator gives some feedback about the quality of its output — that’s how Generator is trained to produce even more realistic images.
Why do they need GANs? It’s all about setting up layouts for architectural designs — a task that involves several steps. “Each step is in fact captured by a thoroughly trained GAN model,” noted Chaillou. “The careful study of the organization learned by each model reveals the existence of a deeper bias, know to our discipline as an architectural style. Unpacking our “generation stack” will help us coin down the different levels of stylistic, and isolate the underpinnings of style’s influence.”The result is a pipeline that lets designers generate a room layout, and furnish it to reassemble all apartment units into a tentative floorplan.

The system comes with a simple interface that lets you set up all kinds of layouts (you can actually try the tool here). You just need to a set of constraints and boundaries to generate the resulting plan. Then, you can modify the input to refine the result.

Seems a bit complicated? It’s actually pretty easy once you look into it. You can find the paper with the explanations here or check out the final tool there. Don’t forget to discuss the story in the comments.

The goal of the ClearCut courses is to teach you a solid workflow that is used in the AAA game industry. The first episode covers the process of creating an AAA fire hydrant from start to finish.

Any future updates are included and will be available for download in case they are released. Next episodes are not included.

Check the full description

Contact Emiel Sleegers

Leave a Reply

Related articles