The microscopic visual forms in architectural art design following deep learning

作者：Guo, Yi^*

来源：Journal of Supercomputing, 2022, 78(1): 559-577.

DOI：10.1007/s11227-021-03888-0

• Deep learning
• Micro-morphology
• Generative adversarial networks
• Architectural art
• Convolutional neural network

The purpose is to analyze the architectural art design of microscopic visual forms based on deep learning. First, GAN (Generative Adversarial Networks) of deep learning is applied to the field of architecture, and a multi-adversarial information sharing GAN is proposed through the improvement of GAN, and an architecture generation model of microscopic visual forms based on deep learning is constructed. The model is stimulated and its accuracy, distortion, and stability are analyzed. The results show that the accuracy of the model can reach more than 80% on different datasets. Compared with the other models in the related field, the model built in this study can show the features of the building images with the minimum distortion. Meanwhile, the curve hovers are around 0 in the process of model training, which is balanced. Therefore, the research can significantly improve the accuracy and the effect of feature extraction, and provide an experimental basis for the later architectural design of microscopic visual forms.

• 单位
  y