Neural network fusing text and image data and design method of building structure of neural network

Abstract

The invention belongs to the technical field of civil structure engineering and computer deep learning application, and particularly relates to a neural network fusing text and image data and a design method of a building structure of the neural network. A generator in the neural network for fusing the text and the image data generates an image, a discriminator judges the authenticity of the image, the generator comprises a text code and a feature extraction network to obtain a text feature matrix, the text feature matrix and the image feature matrix are fused, and an image network is generated based on fused features. And according to the language description text of the to-be-designed building structure and the pixelated image of the building feature element, the neural network fusing the text and the image data is adopted to complete the design of the building structure. According to the method, the corresponding structural design is quickly output according to the building design image and the structural basic design attribute text, and the automatic building structural design under the guidance of image and text multi-modal data fusion is realized.

Description

technical field [0001] The invention belongs to the technical field of civil structural engineering and computer deep learning application, and in particular relates to a neural network fused with text and image data and a design method for a building structure thereof. Background technique [0002] In the preliminary design stage of building structures, engineers need to consider the requirements of building layout, the constraints of basic design specification texts for structural design, and the guidance of commonly used structural design experience texts. At present, manual structural design is often difficult to comprehensively consider the constraints of architectural layout, design specifications and experience, and the degree of automation and intelligence in structural design is insufficient. An automated preliminary structural design method that comprehensively considers image and text data constraints is urgently needed. [0003] However, the format of architectur...

AI Technical Summary

Problems solved by technology

The disadvantage of this invention is that it is impossible to consider the influence of text constraints of different structural design attributes on the structural design results in the process of building structure design, and the method needs to group data when training the generative confrontation network, which leads to the reduction of training data and reduces the impact of generative confrontation. network performance

Images