Providing Automated Feedback on the Gestalt Principle of Symmetry by Means of Machine Learning

Abstract

The Gestalt Principles are sometimes used within the field of design to help students understand the underlying structure of images. Assessment activities focused on determining the Gestalt Principles contained in an image is a very subjective exercise, with students and teachers not necessarily agreeing on the content of an image. This subjectivity becomes more problematic when dealing with large class groups or when multiple markers are employed. This study proposes a multi-layer perceptron-based model for classifying images as exhibiting the principle of symmetry. The model exhibits a high level of accuracy when classifying images as containing aspects of symmetry, whether across the entire image or only partially, or not. The model forms part of a larger process which generates images, based on original input images, and highlights areas of symmetry. These annotated images may be used to help provide feedback to students.