FaGED Pipeline

FaGED Pipeline is a research-oriented Python library for comparing spatial layouts and human behavioral patterns using function-aware graph representations.

It can also be used independently to construct graph representations of complex architectural layouts with open spaces and to extract layout prototypes from such configurations.

Key features

• Step-wise pipeline (Step0–Step5) for reproducible experiments
• Layout graph representation for complex open-space buildings
• Prototype extraction via community detection (e.g., Infomap)
• GED / nGED / FaGED computation + retrieval ranking outputs
• Designed for architectural research workflows

Installation

pip install faged-pipeline

Minimal example (layout graph → CaG)

from faged_pipeline import run_step2_then_step3, Step3Config

run_step2_then_step3(
    json_folder="DATA_ROOT/json",
    step2_output_folder="OUTPUT_ROOT/step2_basegraphs",
    step3_output_root="OUTPUT_ROOT/step3_transform/variants",
    step3_selected_folder="OUTPUT_ROOT/step3_transform/selected",
    step3_cfg=Step3Config(),
    save_png_step2=True,
    save_selection_csv_step3=True,
)

Full pipeline example (Step0–Step5)

The complete step-by-step notebook and documentation are hosted on GitHub:

• GitHub repository: https://github.com/songguoarchitect/faged
• Example notebook: quick_start.ipynb

Note: PyPI does not reliably render local images (relative paths).
We keep the PyPI page concise and place full visuals/examples on GitHub.

Data availability

A small example dataset (5 samples) is included for demonstration.
The complete dataset (150 images) is not publicly released due to data usage agreements.

Citation

If you use FaGED Pipeline in academic work, please cite:

@article{Guo2025FaGED,
  title   = {Function-Aware Graph-Based Layout Retrieval for Public Buildings},
  author  = {Guo, Song and Kee, Tris and Zhuang, Weimin},
  journal = {Automation in Construction},
  year    = {2025}
}

License

MIT License.