Roof-GAN: Learning to Generate Roof Geometry and Relations for Residential Houses

TL;DR

Roof-GAN is a new generative adversarial network that creates detailed, structured roof geometries for residential houses, capturing both primitive shapes and their spatial relationships in a graph-based model.

Contribution

It introduces a novel GAN architecture that generates structured roof models with explicit geometric primitives and relationships, including a differentiable vectorizer and a new evaluation metric.

Findings

Outperforms existing methods in generating diverse, realistic roof structures

Successfully models complex relationships between roof primitives

Provides a new metric for evaluating structured geometry generation

Abstract

This paper presents Roof-GAN, a novel generative adversarial network that generates structured geometry of residential roof structures as a set of roof primitives and their relationships. Given the number of primitives, the generator produces a structured roof model as a graph, which consists of 1) primitive geometry as raster images at each node, encoding facet segmentation and angles; 2) inter-primitive colinear/coplanar relationships at each edge; and 3) primitive geometry in a vector format at each node, generated by a novel differentiable vectorizer while enforcing the relationships. The discriminator is trained to assess the primitive raster geometry, the primitive relationships, and the primitive vector geometry in a fully end-to-end architecture. Qualitative and quantitative evaluations demonstrate the effectiveness of our approach in generating diverse and realistic roof models over the competing methods with a novel metric proposed in this paper for the task of structured geometry generation. Code and data are available at https://github.com/yi-ming-qian/roofgan .