Deep learning and computer vision-based intelligent optimization system for environmental art design spatial layout
The design for environmental art can be an ideal combination of aesthetics and functionality. The traditional approaches to spatial layout design are limited by human experience, lack adaptability, and are heuristic and time-consuming, without explicitly constructing complex spatio-visual relationships. Developing an intelligent system that integrates deep learning (DL) and computer vision (CV) to boost spatial layout design in environmental art is explored, and an evolutionary DL method is proposed, which is based on an evolutionary algorithm and a Saplings Growing Up-tuned Spatio-temporal Graph Neural Network (SGU-STGNN). A dataset of 4550 environmental art images is used for research, and the images are comprehensively preprocessed with robust and generalizable operations like normalization, resizing, and augmentation. Intelligent spatial layout optimization relies on convolutional neural networks (CNNs) to detect key visual and spatial information. A multi-criteria decision-making process method (the Analytic Hierarchy Process – AHP) is used to evaluate extracted features, integrating expert judgment with scientific criteria to develop layouts that balance beauty, ecology, and usability. The SGU mechanism facilitates adaptive evolutionary tuning for optimized model convergence and learning stability, whereas the STGNN mechanism captures dynamic spatial–temporal relations for better contextual design understanding. The system is implemented in Python 3.10 + and allows for flexible development and integration of models. The experimental results show that SGU-STGNN achieves better Feature detection accuracy (96.53%), carbon footprint (11.03), and FID (11.01) than baseline methods. Research highlights the potential of Artificial Intelligence (AI)-driven approaches to advance intelligent, efficient, and sustainable spatial layout optimization in environmental art design.
Authors
- Zipeng Zhang (ORCID: https://orcid.org/0009-0005-2047-5278)
Institutions
- Gansu Institute of Political Science and Law (CN)
Publication Details
- Journal
- Discover Applied Sciences
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1007/s42452-026-09480-6
- Primary Topic
- Advanced Neural Network Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00