Graph learning
Train models on nodes and edges that capture BIM relationships such as contains, bounded_by, hosted_by and supported_by.
Large-scale synthetic IFC building graphs for machine learning
Buildata: A foundation dataset for BIM AI
Buildata is a large-scale synthetic dataset of IFC building graphs for machine learning. Instead of focusing on 3D geometry as the primary unit, Buildata organizes each sample as a building graph with BIM elements, relationships, attributes and metadata.
building graph dataset
01
Nodes
Each building contains BIM elements represented as IFC entities such as walls, doors, windows, slabs, spaces and structural components.
02
Edges
Relationships describe how building elements connect, contain, host, support or bound one another inside the graph.
03
Features
Attributes, property sets and metadata make the graph useful for classification, reasoning, prediction and benchmarking workflows.
Unit1 sample = 1 building
GraphNodes, edges, hierarchy and metadata
ScopeIFC semantic structures for machine learning
what buildata contains
A building graph dataset, not a 3D model library
Buildata is designed for AI systems that need to learn BIM semantic structures. Each sample represents a complete building organized as a graph, with entities, relationships, hierarchy and structured attributes aligned with IFC logic.
Semantic classification
Learn to identify IFC entities, predefined types and semantic patterns across complete buildings.
Attribute prediction
Use features and property sets to test metadata completion, validation and BIM reasoning workflows.
featured datasets
Explore IFC building graph datasets
technology
Structured for reproducible BIM AI experiments
Buildata organizes each building sample into graph-ready files that separate metadata, nodes, edges, hierarchy and schema information for reproducibility.
01. metadata.json
Describes the building sample, typology, schema, generator profile, counts and reproducibility settings.
02. nodes.json
Represents BIM elements as IFC nodes with attributes, features and selected property sets.
03. edges.json
Captures semantic relationships that convert the building into a BIM graph for machine learning.
04. hierarchy.json
Preserves the IFC hierarchy from project and site down to building, storey and space levels.
explore buildata
Start with sample building graphs for BIM AI
Download the first sample, inspect its graph structure and evaluate how IFC semantic data can support machine learning workflows.