Multimodal Clustering Datasets

The ORL and PIE facial datasets consist of 400 and 680 facial images respectively, both with three feature modalities of grayscale, LBP, and Gabor. The Football dataset includes 248 Premier League-related samples with 20 club categories, and the Politics dataset includes 348 Irish political-related samples divided into 7 groups.

Both datasets adopt nine social and textual modalities. All datasets are obtained from public academic resources and stored in standard .mat format, which contains multimodal feature matrices and clustering labels. The original data is complete without missing values or annotation errors.