Painting Style Classification Using Deep Neural Networks

Abstract

In this paper we describe the problem of painting style classification into five classes: impressionism, realism, expressionism, post-impressionism and romanticism. While most previous approaches relied on image processing and manual feature extraction from painting images, our model based on the ResNet architecture and pre-trained on the ImageNet dataset operates on the raw pixel level. The training has been performed on a large dataset (about 43k images for five class style classification problem). To increase the quality of final model a large number of various augmentations were used: random Affine transform, crop, flip, color jitter (i.e. contrast, hue, saturation), normalization, a scheduler for the optimizer. Finally model weights were pruned which allowed increasing accuracy up to 51.5% and decreasing computation time as well.