Evolving Stencils for Typefaces: Combining Machine Learning, User’s Preferences and Novelty

• Autor(a) principal: Martins, Tiago
• Data de Publicação: 2019
• Outros Autores: Correia, João, Costa, Ernesto, Machado, Penousal
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10316/106866; https://doi.org/10.1155/2019/3509263

Typefaces have become an essential resource used by graphic designs to communicate. Some designers opt to create their own typefaces or custom lettering that better suits each design project. This increases the demand for novelty in type design, and consequently the need for good technologicalmeans to explore newthinking and approaches in the design of typefaces. In thiswork, we continue our research on the automatic evolution of glyphs (letterforms or designs of characters).We present an evolutionary framework for the automatic generation of type stencils based on fitness functions designed by the user.The proposed framework comprises two modules: the evolutionary system, and the fitness function design interface. The first module, the evolutionary system, operates a GeneticAlgorithm,with a novelty searchmechanism, and the fitness assignment scheme.The secondmodule, the fitness function design interface, enables the users to create fitness functions through a responsive graphical interface, by indicating the desired values and weights of a set of behavioural features, based onmachine learning approaches, andmorphological features. The experimental results reveal the wide variety of type stencils and glyphs that can be evolved with the presented framework and show how the design of fitness functions influences the outcomes, which are able to convey the preferences expressed by the user. Thecreative possibilities createdwith the outcomes of the presented framework are explored by using one evolved stencil in a design project. This research demonstrates how Evolutionary Computation andMachine Learning may address challenges in type design and expand the tools for the creation of typefaces.