Dmitrii Marin, PhD candidate
David R. Cheriton School of Computer Science
We will review recent papers by a research groups from Oxford University and Berkeley University. They introduce a novel clustering objective that learns a neural network classifier from scratch, given only unlabelled data samples. The model discovers clusters that accurately match semantic classes, achieving state-of-the-art results in eight unsupervised clustering benchmarks spanning image classification and segmentation. The method is not specialized to computer vision and operates on any paired dataset samples. The trained network directly outputs cluster labels, rather than high dimensional representations that need external processing to be usable for semantic clustering.
200 University Avenue West
Waterloo, ON N2L 3G1