Article: Application of deep learning algorithms in the design of urban subway public art space Journal: International Journal of Environment and Sustainable Development (IJESD) 2026 Vol.25 No.5 pp.44 - 72 Abstract: This paper aims to address the problem of insufficient integration of user visual attention behaviour modelling and spatial practicality in the design of subway public art spaces. This paper first constructs a cultural semantic labelling system and spatial attribute structure for subway stations based on deep neural networks. Second, it achieves multimodal deep alignment between semantic content and visual images through a contrastive language-image pre-training (CLIP) model. Then, it designs a multi-objective optimisation generation framework. Finally, it introduces a spatial structure adaptive analysis mechanism to achieve deep integration of generated content with the real subway environment. Experimental results show that each cultural category achieved high average scores of 0.9025 and 0.88 in terms of semantic consistency and accuracy of cultural background expression, respectively, indicating that the method performs well in terms of visual guidance effect and actual spatial foothold. Inderscience Publishers - linking academia, business and industry through research

Title: Application of deep learning algorithms in the design of urban subway public art space

Authors: Qian Wang

Addresses: Faculty of Humanities and Arts, Macau University of Science and Technology, Macau, China

Abstract: This paper aims to address the problem of insufficient integration of user visual attention behaviour modelling and spatial practicality in the design of subway public art spaces. This paper first constructs a cultural semantic labelling system and spatial attribute structure for subway stations based on deep neural networks. Second, it achieves multimodal deep alignment between semantic content and visual images through a contrastive language-image pre-training (CLIP) model. Then, it designs a multi-objective optimisation generation framework. Finally, it introduces a spatial structure adaptive analysis mechanism to achieve deep integration of generated content with the real subway environment. Experimental results show that each cultural category achieved high average scores of 0.9025 and 0.88 in terms of semantic consistency and accuracy of cultural background expression, respectively, indicating that the method performs well in terms of visual guidance effect and actual spatial foothold.

Keywords: visual attention modelling; public art space; deep learning; cultural semantic framework; spatial adaptability analysis.

DOI: 10.1504/IJESD.2026.151850

International Journal of Environment and Sustainable Development, 2026 Vol.25 No.5, pp.44 - 72

Received: 23 May 2025
Accepted: 20 Oct 2025
Published online: 23 Feb 2026 *

Title: Application of deep learning algorithms in the design of urban subway public art space

Keep up-to-date