Simulation of Human Opinions about Calligraphy Aesthetic
Ana Veleiro Pérez, Eduardo Cermeño, Juan Alberto Siguenza · 2014
This paper proposes a method for simulating human opinions about graphical artistic expressions like calligraphy using computers. Scanned images of handwriting texts from a large database are labeled as "beautiful writing" or "ugly writing" by two persons based on their own likes. Our objective is to replicate these opinions using machine learning techniques. Shape features are extracted from the images in order to encode aesthetic principles. A classifier based on k-nearest-neighbors algorithm is trained to automatically label images. The results are promising since most of the different configurations of the system present good performance. Both, method and feature selection results could be of use for future work on aesthetic classification by computers.