Painter Prediction from Artworks with Transfer Learning

Abstract

This paper provides information about prediction of our national painters from their paintings. Since most of the painters whose paintings are used in this research lived in the past, many of their artwork couldn't reach the present and couldn't be found on one web site on the internet. Moreover, some of them have paintings less than a hundred. As result, an original dataset is created from these painters paintings, which are collected at different sources. Since the art classification is a difficult task and takes many years to expertise, using machine learning techniques can provide efficiency. Purpose of this paper is to identify and classify national artists from their artwork using machine learning techniques. In order to solve art classification problem, transfer learning with four different architectures is used and the results are compared. © 2021 IEEE.