Evolutionary Reproduction of Dutch Masters: The Mondriaan and Escher Evolvers

Abstract

Creative evolutionary systems are often concerned with producing imagesof high artistic quality. A key challenge to such a system is tobe human-competitive by producing the same quality. Then, mimickingexisting human artists could be seen as a canonical benchmark, notunlike the Turing test for intelligence. This chapter discusses twoapplications aimed at evolving images in the styles of two well-knownDutch painters: Mondriaan and Escher. For both cases we have an evaluationcriterion based on “style-fidelity” as perceived and judged by theusers. In other words, here we have a target style, which makes the(subjective) selection less free than in applications solely aimingat nice images. Technically, the Mondriaan evolver is less difficult,given that his most popular style “simply” uses horizontal and verticallines, and primary colours to fill the resulting rectangles. TheEscher evolver project is more challenging. First, because Escher'sstyle is less simple to capture. Second, the system is tested invivo, in a real museum, posing requirements on the interface. Wedescribe how to meet the style challenge based on the mathematicalsystem behind Escher's tiling. Designing a suitable representationand the corresponding variation operators based on this system specifiesan appropriate search space guaranteeing the Escher style to someextent and leaving enough freedom for the selection. As for the secondobjective, we describe two versions that differ in the way the imagesare presented to, respectively evaluated by the visitors. The experiencesgained during a six-month exhibition period in the City Museum inThe Hague, The Netherlands, are discussed from the visitors' perspectiveas well as from the algorithmic point of view and are illustratedwith some “evolved Escher's”.