Project info
- Developer Feliks Jakimow
- Showcase year 2021
- Programme Computer Game Applications Development
Synopsis
Artificial neural networks are commonplace in modern applications due to their adaptability in solving new and old problems - especially ones that are traditionally tedious.
By using a gated recurrent unit (GRU) encoder-decoder network, countless design features in a city's design can be accurately captured. Used in conjunction with procedural generation, a full city layout can be achieved with little effort from the user, only requiring them to choose a city located anywhere in the world.
Generating City Layouts Using Artificial Neural Networks
Manually creating a city layout is tedious; artificial neural networks can be used to capture a city's layout accurately, with minimal input from the user.
Motivation
I've had an interest in using procedural generation to generate unique, sometimes surreal, places for people to explore - and to me cities have been a missing link in this field. You could get very good looking environments with conventional methods but they never had much life to them as they lacked man-made structures. I hope to plug the gap with this project!
Credits
“Generating City Layouts Using Artificial Neural Networks” is a 2021 Digital Graduate Show project by Feliks Jakimow, a Computer Game Applications Development student at Abertay University.
Find out more
