Automatic Classification of Woodcuts and Copperplate Engravings
Michael J. Cormier, Sarah Park, Lauren Beck · 2020
In this paper we introduce two related Bayesian approaches to distinguishing woodcut and copperplate engraving illustrations. Our approach uses simple methods grounded in the characteristics of the techniques to achieve high performance and excellent interpretability. Specifically, we use the distributions of image gradient magnitudes in illustrations to capture differences in the use of lines in each artistic technique, and the output of a probability for each class aides in interpretability. Our experiments demonstrate high accuracy and robustness with respect to parameters for both algorithms.