۱٫ Introduction

۲٫ Material and Methods

۳٫ Experimental Results

۴٫ Conclusions

۵٫ References

Painters can be affected during their life simultaneously from different movements, the same movements or a few different movements. This situation makes the problem of identifying, and classifying painters more difficult. In this paper, we have tested the latest deep neural networks on this problem. There are 17 painters who lived in different term and influenced by different art movements, an average of 46 paintings per painters in our data set to test this problem. GoogleNet, DenseNet, ResNet50, ResNet101 and Inceptionv3 networks are applied to this data set. Although DenseNet gives the highest result, considering the cost parameters such as training time and file size, the Inceptionv3 and ResNet50 which provide near to DenseNet results is the optimum networks.

Introduction

In recent years, the transfer of the art work to digital environment has accelerated, so large libraries have begun to be created on the internet. These libraries brought together paintings from museums, churches or collectors in different place of the world to a wide range of people, such as educators, curious painting viewers and art students. However, museums, churches and collectors have hundreds of thousands of paintings, and digitization is not an easy task so a painter classifier will become more purposive and practical, for museum curators, curious painting viewers and art students. We hope that creating an painter classifier allows curators, educators, curious painting viewers and art students to automatically tag objects and allow visitors to browse paintings more freely.