IFCNetCorePro：High-quality Multimodal Dataset for IFC Object Classification

IFCNetCorePro is a high-quality multimodal dataset for the task of IFC object classification. Based on the original IFCNetCore dataset, we adopt three different view sampling strategies: the original IFCNetCore strategy, the ArcShapeNet strategy, and our novel bounding box-based approach. As a result, three sets of mutil-view images are generated for each object. In addition, a heterogeneous graph IFG is constructed based on each IFC-SPF file in IFCNetCore and provides the IFC schema graph (ISG) embedding information used for IFG node feature initialization.