About the Project
Research Overview
The integration of generative AI into architecture is transforming early-stage design processes, particularly by enabling rapid and automated floorplan generation. These tools increase efficiency by exploring large design spaces and producing multiple layout options in a short time. However, current workflows struggle to integrate critical performative analyses such as daylight availability, thermal comfort, and energy demand into the earliest phases of design.
The absence of integrated performance feedback at the stage where design decisions have the greatest impact risks overlooking important optimization opportunities. Most existing datasets for automated architectural design are limited in scope - many are synthetic rather than based on real-world buildings, which introduces a "garbage in, garbage out" risk when models are trained only on generated floorplans.
Furthermore, no publicly available datasets combine architectural layouts with geocoded information, climate data, and energy simulations, even though building performance is deeply tied to location and environmental context. This project addresses these limitations by developing a real-world architectural dataset enriched with geospatial information, environmental data, and energy simulation outputs.
Learn more: Read the full research blog post on IAAC's website
Problem Statement
AI-assisted floorplan tools prioritize speed and quantity, often neglecting performance factors like daylight and energy use. This thesis proposes a scalable workflow to enrich floorplan datasets with context-aware performance attributes, enabling more informed early-stage design and future ML applications.
Research Question
How can we create a workflow to enrich existing floorplan databases with performative data?
Credits & Acknowledgments
Research & Development
Joaquín Broquedis, Marco Durand, Matea Pinjusic
Master's Thesis Students
Institute for Advanced Architecture of Catalonia (IAAC)
Master in Advanced Computational Design for Architecture
Academic Supervision
Thesis Advisor: Angelos Chronis
Institute for Advanced Architecture of Catalonia (IAAC)
Data Sources
Building performance data and IFC models sourced from:
- Open building performance databases
- Collaborative research partnerships
- Simulated building models for research purposes
Base Dataset:
Matthias Standfest, Michael Franzen, Yvonne Schröder, Luis Gonzalez Medina, Yarilo Villanueva Hernandez, Jan Hendrik Buck, Yen-Ling Tan, Milena Niedzwiecka, & Rachele Colmegna. (2022). Swiss Dwellings: A large dataset of apartment models including aggregated geolocation-based simulation results covering viewshed, natural light, traffic noise, centrality and geometric analysis (3.0.0) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.7788422
Complete data attribution available in the Sources section.
Technical Implementation
This platform utilizes open-source technologies including:
- Three.js: 3D model visualization
- D3.js: Interactive graph visualizations
- IFC.js: BIM model processing
- Python: Data processing and analysis
Special Thanks
Gratitude to the IAAC community, fellow researchers, and the open-source software community whose tools and methodologies made this research possible.
Alexander Groth, Co-Founder and CEO of CityWeft
Jon Mirtschin, Director at Geometry Gym
How to Cite
Recommended Citation
Broquedis, J., Durand, M., & Pinjusic, M., Chronis, A. (2025). Performance Data Enriched Floorplan Dataset. Institute for Advanced Architecture of Catalonia, Master in Advanced Computational Design for Architecture, 2025
BibTeX
@dataset{iaac_performance_2025,
title={Performance Data Enriched Floorplan Dataset},
author={Joaquín Broquedis and Marco Durand and Matea Pinjusic},
supervisor={Angelos Chronis},
year={2025},
institution={Institute for Advanced Architecture of Catalonia}
}
License
This dataset is freely available under Creative Commons Attribution 4.0 International License. You may use, share, and adapt the data for any purpose, including commercial use, provided you give appropriate credit to the original creators.
How to cite: Please use the citation format provided above when using this dataset in your research or projects.
Sources & References
External resources, references, and data sources used in compiling this architectural dataset.
From Building Information Modeling to Building Energy Modeling
Buildings, 14(8), 2444
Optimization study for efficient transformation from BIM to energy modeling.
View Publication →Relational Inductive Biases, Deep Learning, and Graph Networks
arXiv preprint arXiv:1806.01261
Fundamental work on graph neural networks and relational reasoning.
View Publication →Early Stage Design Decisions for Sustainable Buildings
The Scientific World Journal, 2014, 365364
Achieving sustainable buildings at lower costs through early design decisions.
View Publication →ReCo: A Dataset for Residential Community Layout Planning
Proceedings of the 31st ACM International Conference on Multimedia
Large-scale dataset for residential community layout planning research.
View Publication →CubiCasa5K: Annotated Floorplans Dataset
5,000 annotated floorplans with semantic labels.
Access Dataset →Ten Questions on Urban Building Energy Modeling
Building and Environment, 168, 106508
Comprehensive review of urban building energy modeling challenges and opportunities.
View Publication →Graph2Plan: Learning Floorplan Generation from Layout Graphs
ACM Transactions on Graphics, 39(4), 118
Novel approach to floorplan generation using graph neural networks.
View Publication →CubiCasa5K: Dataset and Multi-task Model for Floorplan Analysis
arXiv preprint arXiv:1904.01920
Comprehensive dataset and methodology for automated floorplan analysis.
View Publication →Swiss Dwellings: Large Dataset of Apartment Models
Zenodo Dataset v1.0.0
Large dataset including geolocation-based simulation results for Swiss apartment models.
Access Dataset →AI Methodologies in Automated Floorplan Generation
npj Clean Energy, 1, 2
Review of artificial intelligence approaches for high-performance floorplan generation.
View Publication →OpenStudio – Cross-platform Tools for EnergyPlus
Open-source building energy modeling tools and software platform.
Visit Website →Building Simulations Supporting Decision Making in Early Design
Renewable and Sustainable Energy Reviews, 61, 187–201
Review of building simulation tools for early-stage architectural design.
View Publication →Urban Building Energy Modeling—A Review
Building and Environment, 97, 196–202
Comprehensive review of urban building energy modeling as a nascent field.
View Publication →Building Performance Database (BPD)
Comprehensive database of building performance metrics and energy data.
Access Database →MSD: Benchmark Dataset for Floor Plan Generation
Computer Vision – ECCV 2024
Benchmark dataset for floor plan generation of building complexes.
View Publication →Doors Detection in Floor Plans
Machine learning model for door detection in architectural floor plans.
View Dataset →Building Types Online
Comprehensive database of building types and architectural information.
Visit Database →IDA ICE – Simulation Software
Building energy simulation software for performance analysis.
Visit Website →Energy Efficiency in Buildings
EU focus on energy efficiency in building design and construction.
View Report →WAFFLE: Multimodal Floorplan Understanding in the Wild
Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision
Advanced multimodal approach to floorplan analysis and understanding.
View Publication →Hugging Face Datasets: Floorplan Masks
Repository of floorplan-related machine learning datasets.
Browse Datasets →Evaluation of Deep Learning-based Automatic Floor Plan Analysis
Applied Sciences, 11(11), 4727
AHP-based assessment of deep learning technologies for floor plan analysis.
View Publication →Raster-to-Vector: Revisiting Floorplan Transformation
Proceedings of the IEEE International Conference on Computer Vision
Advanced methods for converting raster floorplans to vector representations.
View Publication →House-GAN++: Generative Adversarial Layout Refinement
IEEE/CVF Conference on Computer Vision and Pattern Recognition
Advanced generative model for professional architectural layout design.
View Publication →InvSR: Image Enhancement and Super-resolution
Web application for image enhancement and super-resolution processing.
Try Application →Switzerland Energy Efficiency Profile
Comprehensive analysis of energy efficiency trends and policies in Switzerland.
View Report →HouseDiffusion: Vector Floorplan Generation via Diffusion Model
IEEE/CVF Conference on Computer Vision and Pattern Recognition
Novel diffusion model approach for vector floorplan generation.
View Publication →EnergyPlus Essentials
Version 24.1.0
Comprehensive guide to EnergyPlus building energy simulation software.
View Documentation →Machine Learning as Surrogate Model for Heating-Demand Optimization
Proceedings of SimBuild 2024 (IBPSA-USA)
ML approaches for early-stage heating demand optimization in buildings.
View Publication →Automated Floorplan Generation in Architectural Design
Automation in Construction, 140, 104385
Comprehensive review of methods and applications for automated floorplan generation.
View Publication →Vectorizer AI: Raster to SVG
Web application for converting raster images to vector SVG format.
Try Application →Surrogate Modelling for Sustainable Building Design
Energy and Buildings, 198, 170–186
Comprehensive review of surrogate modeling approaches in sustainable building design.
View Publication →Köppen Climate Classification
Comprehensive information on Köppen climate classification system.
View Article →Bringing Embodied Carbon Upfront
Report on embodied carbon in building construction and design.
View Report →Pseudo Floor Plan 12k Dataset
Large dataset of pseudo floor plan images for machine learning research.
Access Dataset →