Artistic Genre Classification for Digitized Painting Collections

This paper presents an automatic digital image classification system for the recognition of the artistic genre of paintings represented in consumer-quality digital images. The system is developed as a tool that helps a better understanding of visual arts by untrained users and is a first step into an automatic art painting guide. The developed system is based on the classical feature space paradigm; for each painting image a set of 12 relevant, descriptive features are extracted and feed to a classifier. The paper presents the possible use of SVM, AdaBoost and Neural Networks. The experiments are performed onto an image database containing almost 5000 digital painting images from six different genres (Baroque, Renaissance, Rococo, Romanticism, Impressionism and Cubism). We claim that the proposed approach outperforms the reported state of the art, in terms of classification performance, speed and size of the tested database.