YaMiSo1995/

BIMCompNet_100-samplesubset

	BIMCompNet - 100-Sample Balanced Subset and Pretrained Model Checkpoints




	
	📄 Description

This release includes a 100-sample balanced subset of the BIMCompNet dataset, designed for fast benchmarking and reproducible evaluation of geometric deep learning models on BIM component classification. The dataset contains 100 randomly sampled instances per IFC class, selected from the full 87-class dataset, ensuring class balance for fair training and evaluation. In addition to the… See the full description on the dataset page: https://huggingface.co/datasets/YaMiSo1995/BIMCompNet_100-samplesubset.