Unsupervised Image Classification Approach Outperforms SOTA Methods by ‘Huge Margins’ | Synced

Researchers from Katholieke Universiteit Leuven in Belgium and ETH Zürich in a recent paper propose a two-step approach for unsupervised classification.