Machine learning-based environmental art automated design method

Abstract

In the field of environmental art design, there are significant differences in the level of designers. Some designers lack innovation capabilities and technical support, making it difficult to meet complex design needs, affecting the overall design quality and efficiency. There is also a lack of effective interaction and system between some automated design systems and designers. Problems such as poor interpretability lead to low overall design efficiency and lack of innovation. To address these issues, this article studies an environmental art automated design method based on machine learning (ML) in the field of commercial spaces. This study aims to use machine learning technology to build an automated method for environmental art design, improve design efficiency and optimize creative generation through intelligent models, promote the transformation of environmental art design from a traditional model to an innovative model driven by data and supported by algorithms, and achieve efficient and personalized design goals. Firstly, the image data of commercial space design is normalized and combined with labels such as functional attributes and design styles. Generative adversarial networks (GANs) are used to generate design schemes, and the diversity of design schemes is optimized through feature fusion technology. The interpretability of the design is enhanced through ensemble learning and local interpretable model-agnostic explanations (LIMEs) model, enabling designers to understand and optimize the decision-making process of the model. The experimental results show that compared with the traditional manual design, the longest time of the automated design method in design efficiency testing is 18.1 hours, far shorter than that of the manual design method. In design diversity testing, the automated design method is more diverse in terms of style and number of elements. Meanwhile, the automated design scheme overall outperforms manual design in terms of space utilization, and the average score for layout decisions in interpretability testing reaches a maximum of 9.5 points. These data clearly demonstrate the advantages of the automated design method in improving design efficiency and innovation, and it is easy for designers to understand. Methods based on machine learning have significantly improved the efficiency, quality, and innovation of environmental art design, providing intelligent solutions for the design industry.