Classification of Abstract Images using Machine Learning

Abstract painting uses a visual language of form, color, and line to create a composition that may exist with a degree of independence from visual references in the world. Sometimes, it isn't even about giving the impression of real life without all the little details. This makes the task of classification of the paintings into genres altogether more difficult. In this paper, we describe a systematic method for a machine learning based approach to classifying digital images of abstract art into their most apt artistic styles. To increase the effectiveness of classification, we stack the two classifiers namely Convolutional Neural Network and Deep Neural Network. The hybrid model thus formed outperforms the separate singular models. Furthermore, the task of analysis of color emotions of the artistic image helps to gain further insights of the said classes.