AI-assisted Building Information Modeling for intelligent analysis, prediction, and design
AI4BIM turns building data into practical analysis and early design support. One web platform combines IFC exploration and BIM questions with room classification, house-price prediction, and generative layout tools.
IFC Explorer · model upload, 3D room exploration, and BIM question answering
Demonstrators in action
Three short walkthroughs show how building-data extraction, computer vision, and predictive machine learning support different stages of a building workflow.
Building intelligence
Upload an IFC model to extract storeys, rooms, walls, floor areas, metadata, and geometry with IfcOpenShell. The browser renders room volumes as 3D boxes, while a model summary powers plain-language questions about the building.
Computer vision
A vision model classifies an uploaded room photo and returns confidence scores. Fine-tuned variants reached roughly 98% on their training dataset but generalised poorly, while the base model remained steadier at 86–89% across both datasets.
Machine learning
A demonstrator connecting structured building attributes to a trained model: an XGBoost regressor estimates a price from living area, bedrooms, bathrooms, construction year, and kitchen count, using a model trained on a public US housing dataset.
Source repository
View the implementation, project structure, and setup documentation for the AI4BIM platform.
