Modeling urbanization patterns with generative adversarial networks

TL;DR

This paper introduces a novel GAN-based approach to simulate realistic urban patterns, capturing complex spatial organization and key metrics of global urban land-use data.

Contribution

It presents a new method leveraging GANs to generate hyper-realistic urban patterns that replicate real-world spatial structures.

Findings

Successfully reproduces complex urban spatial organization

Quantitatively recovers key urban spatial metrics

Generates a synthetic urban universe

Abstract

In this study we propose a new method to simulate hyper-realistic urban patterns using Generative Adversarial Networks trained with a global urban land-use inventory. We generated a synthetic urban "universe" that qualitatively reproduces the complex spatial organization observed in global urban patterns, while being able to quantitatively recover certain key high-level urban spatial metrics.