Title: New media interaction in art design based on deep learning binocular stereo vision

Authors: Yongchao Liu; Ziping Zhao

Addresses: School of Art, Qingdao Agricultural University, Qingdao, 266109, China ' School of Art, Qingdao Agricultural University, Qingdao, 266109, China

Abstract: Advances in computer vision technology and the widespread promotion of artworks have led to a profound connection between the two. Highly imitated fakes are often found in art works, which damage consumers as well as creators' own interests, so the study improves the accuracy of art works authenticity identification through the development of computer vision technology. The research is based on machine binocular stereo vision technology, the convolution neural network structure of single shot multi-box detector (SSD) is fused and trained to establish a high-precision object recognition model, which recognises objects by matching the feature points of binocular images. In the experiment, the SSD model has a stable loss value of 0.9 in the loss function performance test, and the overlap rate of the model is around 0.85, which indicates that the model has a high accuracy of object recognition. In the feature point matching algorithm, the parallax value of the multi-feature point fusion matching algorithm is stable in the range of 67 to 75 after filtering. The model proposed in the study has high accuracy in object recognition, which can play an important role in artwork authenticity identification.

Keywords: single shot multi-box detector; SDD; convolutional neural network; object recognition; art design; multi-feature point fusion matching algorithm.

DOI: 10.1504/IJCISTUDIES.2023.137845

International Journal of Computational Intelligence Studies, 2023 Vol.12 No.3/4, pp.238 - 254

Received: 01 Sep 2022
Accepted: 22 Sep 2023
Published online: 05 Apr 2024 *

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