Strezoski Gjorgji Worring Marcel

Abstract

Vast amounts of artistic data is scattered on-line from both museums and artapplications. Collecting, processing and studying it with respect to allaccompanying attributes is an expensive process. With a motivation to speed upand improve the quality of categorical analysis in the artistic domain, in thispaper we propose an efficient and accurate method for multi-task learning witha shared representation applied in the artistic domain. We continue to show howdifferent multi-task configurations of our method behave on artistic data andoutperform handcrafted feature approaches as well as convolutional neuralnetworks. In addition to the method and analysis, we propose a challenge likenature to the new aggregated data set with almost half a million samples andstructured meta-data to encourage further research and societal engagement.