Metacosmic three-dimensional building model stylized migration method

Abstract

The invention provides a stylized migration method for a meta-universe three-dimensional building model. The method comprises the following steps: a1, preprocessing a building component model file; a2, carrying out stylized migration on the chartlet of the building component; and a3, learning features of different styles by using the deep learning model, performing style transformation on the input original two-dimensional or three-dimensional building component by using the features, and performing texture and concave-convex mapping replacement on the transformed two-dimensional or three-dimensional building component. According to the application of the style migration based on deep learning on the three-dimensional model, the requirement for style migration of the three-dimensional model in the virtual digital world is met, the efficiency and the real-time performance of style migration are improved, and meanwhile, the capacity of recombining and reusing migrated results is achieved.

Description

technical field [0001] The invention relates to the technical field of metaverse, in particular to a method for stylized transfer of three-dimensional architectural models of metaverse. Background technique [0002] The core idea of ​​Metaverse is to provide people with a virtual world that can interact in real time. In such a virtual world, people can walk around digitally and chat and interact with other users. Architecture is one of the basic elements in the metaverse virtual world. The three-dimensional properties and design methods of the building itself are naturally connected with 3D visualization technology. On this basis, how to design and transform the building facades in the virtual world and provide The expression of diverse artistic effects is a hot spot in the academic and industrial circles. [0003] The stylized transfer technology that has emerged in recent years is a very creative work in the field of deep learning. It can subtly integrate the painting sty...

AI Technical Summary

Problems solved by technology

[0005] 1. The existing style transfer technology is limited to the final output image, and cannot perform style transfer on the 3D digital model itself, and it is difficult to meet the model interaction requirements in the virtual world

[0006] 2. The existing style transfer technology requires separate model training based on the original image and the target image for each style transfer, which greatly affects the efficiency and real-time performance of style transfer

[0007] 3. The existing style transfer technology usually transfers the image as a whole rather than the part of the natural understanding of semantics, and does not have the ability to recombine and reuse the transferred results

Images