The Use of Deep Learning Algorithms to Realize the Automatic Evaluation of Paintings and the Analysis of the Quality of Artwork

Abstract

The public now has easier access to photos of items in museums and galleries because to the digitization of fine art collections. This produced a need for software tools that can quickly retrieve and categorize art. In this manuscript, the use of deep learning algorithms to realize the automatic evaluation of paintings and the analysis of the quality of artwork (AE-PQA-HDMHNN)is proposed. Initially, the images collected from the WikiArt dataset are given as input. Afterward, the collected image is fed to pre-processing. In pre-processing, Multimodal Hierarchical Graph Collaborative Filtering (MHMHNN) is used for remove the background noise and enhances the quality of image. Then the pre-processed output is fed to Feature Extraction Using Multi-Objective Matched Synchrosqueezing Chirplet Transform (MOMSCT) is constructed for the extracting the histograms features. After extraction the output is fed to High-Dimensional Memristive Hopfield Neural Network(HDMHNN)for the Classification of Fine-Art Paintings. In general Fine-Art Paintings was given for classification using Tyrannosaurus optimization algorithm (TOA)to optimize the High-Dimensional Memristive Hopfield Neural Network(HDMHNN)for classifying Australian Aboriginal Art, Expressionism, Impressionism, Post Impressionism, Realism and Romanticism. The proposed AE-PQA-HDMHNN approach is implemented in python. The performance of the proposed AE-PQA-HDMHNN approach contains 14.09%, 22%, and 14.4% high accuracy,28.51%, 18.21% and 22.98% higher precision and 0.12%, 0.41%, and 1.44% high F1−scorewhen analysed to the existing methods like Two-Stage Deep Learning Approach to the Classification of Fine-Art Paintings (TS-FAP-SVM),image classification approach for painting using improved convolutional neural algorithm (ICP-CNN),End-to-End Artworks Generation Via Deep Convolutional Based Generative Adversarial Networks (EEAG-GAN)methods respectively.