Retrieving Visually Linked Digitized Paintings

Visual arts are of paramount importance for the cultural, historic and economic growth of our societies. One of the building blocks of most analysis in visual arts is to find similarity relationships among paintings of different artists and painting schools. To help art historians better understand visual arts, this chapter presents a framework for visual link retrieval in digital painting datasets. The proposed framework is based, on one hand, on a deep convolutional neural network aimed at performing feature extraction, and, on the other hand, on a fully unsupervised nearest neighbour searching mechanism to retrieve visual links among digitized paintings. The fully unsupervised strategy makes the proposed method particularly desirable especially in those cases where metadata are scarce, unavailable or difficult to collect.