A Web-Based Generative Modeling Platform for Building Layout Optimization

Date

2026-1-20

Author

Yasavul, Halil

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

986
views

0
downloads

Cite This

In Architecture, Engineering and Construction (AEC) industry, adoption of web platforms for optimization and computational design tools is a major challenge. This study introduces Crafus, a web-based platform that runs generative design workflows directly in the browser using Python scripts. The system has a Python - FastAPI server backend, that runs optimization and geometry generation and a frontend based on React and Three.js. The backend streams optimization process to frontend in real-time, keeping user informed of the progress. This approach enables advanced design explorations without requiring sophisticated local hardware. The report introduces the platform with a two-stage hybrid generative framework for apartment layout. The first stage uses a Variational Autoencoder that learns spatial organization and room area patterns from 4128 real apartment layouts. The second stage uses the outputs of the first stage in a Genetic Algorithm with Slicing Tree approach to create rectangular floor plans. The method is further extended to support irregular building footprints using latent space search and Voronoi-based constraints. The results have shown that the proposed method is a scalable and interactive approach to web-based generative design with real-time feedback.

Subject Keywords

Computational Design, Genetic Algorithms, Variational Autoencoders, Slicing Tree, Optimization

URI

https://hdl.handle.net/11511/118342

Collections

Graduate School of Informatics, Term Project