Design of a MobilNetV2-Based Retrieval System for Traditional Cultural Artworks

Abstract

Aiming at the problem that it is difficult for art teachers to take into account each student in the art appreciation education in colleges and universities, this paper proposes a retrieval system for traditional cultural works of art. Dense connections are used to replace residual connections between bottlenecks in MobileNetV2 network and gradient transmission in the network. The dilution factor is used to control the size of the network to solve the problem of the rapid increase in the number of network channels. In addition, the non-local attention mechanism is effectively combined with the improved MobileNetV2 network structure, which effectively improves the classification accuracy of the network. Compared with VGG16, ResNet18, and ResNet34, the classification accuracy is increased by 21.3%, 9.2%, and 3%, respectively. The method in this paper has achieved good results in the classification of art works. According to the images of art works to be appreciated, it helps students understand the relevant cultural knowledge independently and reduce the burden of teachers.