Research On Generative Design Of Rural Residential Building Layout Based On Deep Learning

For a long time,the development of urban and rural areas in China has been uneven,and rural development lags behind urban development.The architectural design of rural residential buildings is unreasonable.In the context of the rural revitalization strategy,a reasonable layout design of rural residential buildings can help promote rural revitalization.The contradiction between the short design cycle and the long thinking time of rural residential buildings is a major difficulty in the current rural residential design process.Recently,the advent of chat GPT has set off a wave of artificial intelligence following AlphaGo,and people have seen the application potential of artificial intelligence in other industries.This article applies the deep learning technology in artificial intelligence technology to the planar layout design of rural residential buildings,attempting to let machines learn and master the laws of architectural planar layout,and be able to output architectural planar layout plans in a short time based on input land constraints,assisting architects in diverse design thinking.This study proposes a method for generating and designing the layout of rural residential buildings based on the Pixtopix algorithm.In order to achieve the objectives of this study,the main work of this thesis is divided into the following five stages: Firstly,based on domestic and foreign literature,the research on the application of deep learning technology in building generation and design is reviewed,and the research content,research methods,and key and difficult points that may be encountered in the research process are determined;In the second stage,the relevant theoretical knowledge of deep learning was elaborated.Through comparing and selecting common neural networks,the Pixtopix model was determined as the algorithm basis for this experiment;In the third stage,an experimental design method based on deep learning was proposed and discussed in detail from four aspects: design preparation,database establishment,model building,and evaluation of model generation results.In addition,this stage also introduced the basic functional space of rural residential buildings,providing a basis for later experimental data processing;In the fourth stage,based on the previous experimental design method,three typical experiments during the entire experimental research process were selected and elaborated in detail from four aspects: data processing,model training,result analysis,experimental summary and improvement.On the one hand,the experimental improvement part included using LS loss function and Wasserstein distance loss function to replace the loss function of the Pixtopix model framework to conduct comparative experiments,On the other hand,it includes improving the way data is annotated and adding more constraints.Finally,the expected goals of this experimental study have been achieved,through adjusting the model structure and data processing methods.The fifth stage is the discussion and summary of the entire experimental research process,including the qualitative evaluation of the final experimental results using a scoring evaluation method,the discussion of optimization methods in the experimental process,and the summary of building layout generation strategies based on small sample data sets.Finally,the shortcomings and limitations are summarized from both experimental and model aspects.The final experimental results show that the deep learning model can be trained using the rural residential building layout data set to enable it to learn and master the layout characteristics of the building,achieving the experimental purpose of automatically generating the rural residential building layout map.Although the experimental research in this article is at the initial stage of application,we can still see its potential application value in the future.