Show HN: Pipeline and datasets for data-centric AI on real-world floor plans (archilyse.standfest.science)

A new project has been announced that revolutionizes how floor plans are processed for data-centric AI in the built environment by converting inconsistent architectural drawings into usable, simulation-verified data. This pipeline addresses a critical issue in current AI models, which often struggle to generalize due to poor input data derived from incomplete drawings. The initiative comprises a 6-stage data refinery that achieves rapid throughput and accuracy, generating dense context graphs and analysis-ready tables that quantify spatial performance factors like daylight exposure, noise, and connectivity. This advancement is significant for the AI/ML community as it lays the groundwork for more reliable models in real estate applications, facilitating measurable business outcomes and improving automation. The project includes valuable datasets like Swiss Dwellings, focused on full simulation features, and a benchmark for AI model evaluation known as MSD. By moving beyond merely geometric representation to contextual understanding, the initiative promises to enhance applications in real estate valuation, facility management, and even ESG reporting. The overall aim is to provide robust, simulation-driven insights that lead to better decision-making in architectural planning and management.