Abstract
Clustering is a class of unsupervised learning methods that has beenextensively applied and studied in computer vision. Little work has been doneto adapt it to the end-to-end training of visual features on large scaledatasets. In this work, we present DeepCluster, a clustering method thatjointly learns the parameters of a neural network and the cluster assignmentsof the resulting features. DeepCluster iteratively groups the features with astandard clustering algorithm, k-means, and uses the subsequent assignments assupervision to update the weights of the network. We apply DeepCluster to theunsupervised training of convolutional neural networks on large datasets likeImageNet and YFCC100M. The resulting model outperforms the current state of theart by a significant margin on all the standard benchmarks.