Insights from a Large-Scale Database of Material Depictions in Paintings

Deep learning has paved the way for strong recognition systems which are often both trained on and applied to natural images. In this paper, we examine give-and-take relationship between such visual rich information available in fine arts. First, find that designed images can work surprisingly well paintings. particular, interactive segmentation tools be used cleanly annotate polygonal segments within paintings, a task is time consuming undertake by hand. We also FasterRCNN, model been object scenes, quickly repurposed detection of materials Second, show from paintings beneficial neural networks intended training instead improve quality learned features further large number valuable source test data evaluating domain adaptation algorithms. Our experiments based novel large-scale annotated database material depictions detail separate manuscript.