Adaptive Floor Generator Using Machine Learning

Adaptive Floor Generator Using Machine Learning

User Rating: 4 / 5

Star ActiveStar ActiveStar ActiveStar ActiveStar Inactive
GANs (Generative Adversarial Neural Networks), Which depends on the statistical approach, will constitute a real breakthrough for artificial intelligence in the field of architecture. Instead of relying on machines to collect and improve a set of variables, relying on Generative Adversarial Neural Networks to extract and simulate structural features represents a qualitative leap.
The floor plan includes three distinct steps:
  1. building footprint massing
  2. program repartition
  3. furniture layout
Each step corresponds to a Pix2Pix GAN-model trained to perform one of the 3 tasks above. This is done by integrating these models, for example, in this form that will be shown about creating a model for a complete residential building, as it goes beyond the simplicity of simple single houses.

Representation, Learning, and Framework

Pix2Pix uses a conditional generative adversarial network (cGAN) to learn a mapping from an input image to an output image. The network consists of two main pieces, the Generator and the Discriminator. And the type of information that the program can learn is controlled by controlling the image format. Fast iterations & tests using an NVIDIA Tesla V100 GPU for the training process on Google Cloud Platform (GCP) were running. 

GAN learns layout of the rooms and the doors and windows positions

Stack and Models

Three-step generation stack, (1) footprint massing, (2) program repartition (3) furniture layout. each model of the stack handles a specific task of the workflow.

UI and Experience

The user interface is presented in a simple way as restrictions and limitations are introduced to improve the result, as shown below.

Model Chaining and Apartment-Building Generation

The utilization of GANs in this part of the entire apartment building design. The project uses an algorithm to chain models I, II, and III, one after the other, processing multiple units as single images at each step.

The challenge of drawing floor plates hosting multiple units marks the difference between single-family houses and apartment buildings. Strategically, the ability to control the position of windows and units’ entrances is critical to enable unit placement while ensuring each apartment’s quality. Since Model II takes doors and windows position as input, the generation stack described above can scale to entire floor plates generation.
The ability to design the right pipeline will determine AI’s success as a new architectural toolset. Breaking out this pipeline into discrete steps will ultimately permit the user to participate along the way. I believe their control over the machine is the ultimate guarantee of the design process quality and relevance.
Stanislas Chaillou-ArchiGAN: a Generative Stack for Apartment Building Design/NVIDIA / 17 JUL 2019

© D.A.S.C. 2020