Title: Research on the extraction method of painting style features based on convolutional neural network

Authors: Hua Jiang; Ting Yang

Addresses: College of Art, Hunan City University, Yiyang 413000, China ' Art Design Institute Porcelain College, Hunan Vocational College of Science and Technology, Changsha 410004, China

Abstract: In order to overcome the problem of low accuracy in the traditional method for style feature extraction of painting works, this paper proposes a method for style feature extraction of painting works based on convolution neural network. Firstly, the parameters in the digital image of painting works are quantised, and then the feature parameters are fused by fusion technology and used as input information. Then the wind fusion features of painting works are extracted by using the deep hash coding of triple recombination structure in convolutional neural network. The experimental results show that the precision value of this method always stays at a high level with the change of the recall value, which can be kept above 0.7, and the AP value is always above 0.9. It shows that this method has strong adaptability and high precision of feature extraction.

Keywords: paintings; style features; convolution neural network; CNN; deep hash coding; feature extraction.

DOI: 10.1504/IJART.2022.122448

International Journal of Arts and Technology, 2022 Vol.14 No.1, pp.40 - 55

Received: 19 Oct 2020
Accepted: 10 Oct 2021
Published online: 25 Apr 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article