House-GAN: Relational Generative Adversarial Networks for Graph-constrained House Layout Ge...

Content provided by Nelson Nauata, the first author of the paper House-GAN: Relational Generative Adversarial Networks for Graph-constrained House Layout Generation.

This paper proposes a novel graph-constrained generative adversarial network, whose generator and discriminator are built upon relational architecture. The main idea is to encode the constraint into the graph structure of its relational networks. We have demonstrated the proposed architecture for a new house layout generation problem, whose task is to take an architectural constraint as a graph (i.e., the number and types of rooms with their spatial adjacency) and produce a set of axis-aligned bounding boxes of rooms. We measure the quality of generated house layouts with the three metrics: the realism, the diversity, and the compatibility with the input graph constraint. Our qualitative and quantitative evaluations over 117,000 real floorplan images demonstrate that the proposed approach outperforms existing methods and baselines.

What’s New:This paper proposes a novel house layout generation problem, whose task is to take a bubble diagram as an input, and generates a diverse set of realistic and compatible house layouts. The house layout generation poses a new challenge: The graph is enforced as a constraint. We present a novel generative model called House-GAN that employs relational generator and discriminator, where the constraint is encoded into the graph structure of their relational neural networks.

How It Works:In the house layout generation problem, a bubble diagram is represented as a graph where 1) nodes encode rooms with their room types and 2) edges encode their spatial adjacency. A house layout is represented as a set of axis-aligned bounding boxes of rooms. In House-GAN architecture, we employ convolutional message passing neural networks (Conv-MPN), which differ from graph convolutional networks (GCNs) in that 1) a node represents a room as a feature volume in the design space (as opposed to a 1D latent vector), and 2) convolutions update features in the design space (as opposed to multilayer perceptron). The architecture enables more effective higher-order reasoning for composing layouts and validating adjacency constraints.

Key Insights:This paper proposes a house layout generation problem and a graph-constrained relational generative adversarial network as an effective solution. We demonstrate the benefits of exploiting spatial information in our house layout generation problem, via convolutional message passing, as opposed to state-of-the-art GCN-based methods in the literature. We believe that this paper makes an important step towards computer aided design of house layouts.

Behind The Scenes:Our Conv-MPN paper served as inspiration for House-GAN and it has been shown to be a simple and effective technique for geometry related tasks we have tested so far. It is accepted to CVPR 2020.

Anything else?

We believe some challenges are 1) finding effective and memory efficient solutions for preserving spatial information during message passing and 2) tackling potential limitations in the current state of adversarial networks. The current state of our method is still in early stages in enabling automated house layout generation. We believe that being able to include more realistic architectural constraints and outputting cad-level models will be essential for next steps.

The paper House-GAN: Relational Generative Adversarial Networks for Graph-constrained House Layout Generation is on arXiv . Click here to visit the project website.

Meet the authors: Nelson Nauata , Kai-Hung Chang , Chin-Yi Cheng, Greg Mori , and Yasutaka Furukawa from Simon Fraser University and Autodesk Research.

Share Your Research With Synced Review

Share My Research is Synced’s new column that welcomes scholars to share their own research breakthroughs with over 1.5M global AI enthusiasts. Beyond technological advances, Share My Research also calls for interesting stories behind the research and exciting research ideas. Share your research with us by clicking here .