# Archilyse Research

> Archilyse is a research project that transforms floor plan archives into simulation-dense context graphs. It provides verified building data, open datasets, and an AGPL-licensed pipeline for spatial analytics and AI applications in the built environment.

Key concepts:

- **Dense Context Graphs**: Rooms are embedded in full spatial hierarchy (room → unit → floor → building → surroundings) with 367 simulation columns per area
- **Simulation Densification**: Physically grounded signals including viewshed analysis, daylight simulation, traffic noise, and spatial connectivity
- **Verified Data**: Median area deviation <1.2% against portfolio indices, with mandatory QA gates

## Datasets

- [Swiss Dwellings](https://archilyse.standfest.science/swiss-dwellings): Open dataset of 45,000+ Swiss apartments with verified geometry and 367 simulation columns per area. CC BY 4.0 licensed. DOI: 10.5281/zenodo.7788422
- [Modified Swiss Dwellings (MSD)](https://archilyse.standfest.science/modified-swiss-dwellings): Graph benchmark for floor plan auto-completion with 5,372 plans and 18.9K units. Published at ECCV 2024.

## Pipeline

- [Data Pipeline](https://archilyse.standfest.science/pipeline): 6-stage open-source pipeline (AGPL) for processing floor plans into simulation-ready data. Throughput: 500-800 m²/h.
- [GitHub Repository](https://github.com/Archilyse/Archilyse): Source code for the Archilyse pipeline

## Documentation

- [Research Background](https://archilyse.standfest.science/background): Why existing floor plan datasets fail and how this project addresses data quality gaps
- [Industry Applications](https://archilyse.standfest.science/industry): Use cases for property valuation (AVMs), ESG reporting, and facility management

## Downloads

- [Swiss Dwellings on Zenodo](https://zenodo.org/records/7788422): Direct download (CC BY 4.0)
- [MSD on GitHub](https://github.com/caspervanengelenburg/msd): Benchmark dataset and code
- [Quickstart Notebook](https://colab.research.google.com/drive/1ln7y04SJ3xawyOK_ayeY3IQ3PGecLWnt): Google Colab tutorial

## Publications

- **Swiss Dwellings** (2022): Scientific Data, Nature Publishing Group. DOI: 10.5281/zenodo.7788422
- **Modified Swiss Dwellings** (2024): ECCV Workshop on Computational Aspects of Deep Learning
- **Bias Reduction** (2022): "Reducing Bias for Evidence-Based Decision Making in Design" - Springer, Design Computation Input/Output 2022. Foundational paper on the data refinery paradigm.
- **PhD Thesis** (2017): "Applying Emergent Features of Architectural Geometry" - ETH Zurich

## Contact

- [Contact & Licensing](https://archilyse.standfest.science/contact): Research collaborations and commercial licensing inquiries