Evoboard: Geoboard-Inspired Evolved Typefonts

João Eduardo Batista, Fraser Garrow, Carlo Huesca-Spairani, Tiago Martins

Abstract: Type design is a field that deals with the creation of visually appealing designs for the written language. The work of the designer is time-consuming and requires many iterations until the final solution is achieved. Although a human expert is required to validate the final results, this task can be aided by automatic design software. We propose Evoboard, an automatic algorithm that evolves a typefont using a geoboard-inspired representation where each character is a self-intersecting polygon. Evoboard uses a genetic algorithm to optimize the number of vertices of the polygon and their positions in a grid. The evolution of the population is guided by an Optical Character Recognition (OCR) model that aims to maximize the recognition of the polygon as the target character. Thanks to this simple pipeline, both the OCR model and the representation can be easily modified by the user to their needs. We evolve a set of 36 alphanumeric characters that are both highly legible and aesthetically appealing, two important aspects of type design.