Doctoral research conducted at Academy of Fine Arts and Design in Bratislava, focusing on font generation using transformer-based machine learning models. The study involved three comprehensive surveys examining applications, datasets, and model representations in type design. I developed a custom dataset and trained multiple transformer models to validate whether specific dataset regularization techniques improve font generation quality.
The research introduced four novel heuristics for visual evaluation of generated fonts, enabling systematic assessment by type designers and font engineers. Results demonstrated that dataset regularization can enhance the quality of machine-generated typefaces, providing valuable insights for the type design community.