An Architectural Landscape Pattern GANs

In the study of the characteristics of architectural landscape patterns,the traditional research methods mainly focus on the acquisition of information such as the contours of buildings,the lack of extraction of hidden features and the creation of architectural landscape patterns with acquired features,which are essentially Supervised classification can only use the extracted features to determine the type of building and cannot use these features for integration and redrawing.Even the research on building reconstruction in the feature extraction algorithm still needs to artificially provide building parameters,and it cannot adapt to a large number of different buildings.Due to the lack of methods for extracting and reconstructing common features of buildings,traditional architectural feature extraction can only be limited to limited obvious features in some studies focusing on the spatial distribution of landscapes such as the influence of architectural patterns on thermal environment.Research on the need to mine the hidden features of buildings,such as the relationship between architectural landscape configuration and heat island effect,is inconclusive,and there is a lack of research conclusions on how to optimize to achieve the most scientific and reasonable architectural landscape pattern.As an unsupervised learning generation,the anti-network is expected to use AI to extract a large number of hidden features in the architectural landscape pattern,and to realize the construction landscape pattern that cannot be realized by traditional methods.Based on the generation of confrontation network,this paper proposes a new generationconfrontation network model to better realize the generation of architectural landscape pattern.The research work is as follows:1.Analyze the problems and causes of the confrontation network in the process of building landscape pattern generation.The innovation proposes a new network model to improve the problem and verify the superiority of the new model.2.In view of the shortcomings in the extraction of building features,the new model is further innovated to propose an improved method of zoning and segmentation training of building pattern,which makes the model training more rapid and generates a more creative architectural landscape pattern.3.Experiments on new models on various data sets were performed to evaluate the versatility of this model.Through the training of the classical dataset in the machine learning field,excellent results are obtained,which proves that the model has good versatility.4.Through practical application,through further extraction and reconstruction of more complex architectural landscape features,we have obtained an excellent architectural landscape pattern that has not yet been classified.It further proves that the model has superiority to the extraction and reconstruction of building features,and makes it possible to optimize the influence of architectural landscape pattern on the environment and other aspects,which is worthy of further study.