This document proposes using a deep convolutional neural network to embed digitized paintings into a latent space, where k-means clustering is applied to group the paintings based on visual features. The method is evaluated on a database of paintings from 50 artists from 9 stylistic periods. Quantitative results show the deep clustering approach outperforms other methods based on silhouette coefficient and Calinski-Harabasz index. Qualitative results demonstrate the model groups paintings by style with coarse clusters and uses both style and content features with finer clusters.