Automatic Design And Optimization Of Indoor Spatial Layout

There are a lot of layout problems in modern engineering and real life,and indoor spatial layout design is one of the most important problems.The indoor spatial lay-out focuses on the indoor scenes where people rely on production and lving.Besides,models of indoor scenes are gradually used in more and more fields,such as CG an-imated movies,computer games,and virtual reality,etc.The urgent need for indoor scene modeling has motivated related research on indoor spatial layout.Due to the wide application prospects and huge commercial value of indoor spatial layout,it has become an important research topic in the fields of computer graphics and computer vision.As computer-aided design is widely used in the field of intelligent design,the indoor spatial layout has also become an important research field.In the problem of indoor spatial layout,it is usually required to create a layout or optimize a layout on a given area or space.This paper focuses on the automatic design and optimization of indoor spatial layout and three efficient automatic design and optimization methods are proposed based on the given layout boundary or layout space in this paper.We first propose a optimization method based on design constraints for the auto-matic generation of floor plans.Given the boundary of the building and the requirements of the user as design constraints,our method can automatically generate an indoor spa-tial layout.The main idea of the algorithm is to represent the room unit in the indoor spatial layout with polygons composed of multiple rectangles,and transform the con-straints and rules in the layout design into constraints on the relevant rectangles.We select some important design constraints,such as the size of the room,the location of the room,the adjacency of the room,etc.,and formulate these constraints into the con-straint form of mixed integer quadratic programming in a way that is compatible with our model.To improve the applicability of the method,we propose a hierarchical algo-rithm framework based on the above method.A large number of experimental results show that our algorithm can be used in the indoor spatial layout design of residential buildings,and can be extended to the spatial layout of large-scale scenes,such as of-fice buildings,shopping malls and supermarket.Through comparison,we find that our method is faster than previous methods.The above optimization method has strict requirements on the design constraints.A straightforward idea is to learn design guidelines from existing design solutions and use them in the automatic design of indoor spatial layout.Therefore,we propose a deep learning method for automatically and efficiently designing floor plans.Our algorithm automatically generates floor plans using only layout boundaries as constraints.The main idea of the algorithm is to imitate the design process of the designer,and propose a two-stage approach by locating rooms first and then walls to generate a floor plan that can adapt to the given boundary.Based on this,we develop a two-stage deep learn-ing networks for interior spatial layout design.To effectively train our networks,we also construct a dataset full of floor plans,RPLAN,with sufficient data annotations,containing more than 80,000 indoor floor plans from real residential buildings.A large number of generative results show the feasibility and effectiveness of our method.By comparing the plausibility of interior floor plans generated by different algorithms,it is shown that our method substantially outperforms existing methods,and in many cases,floor plans generated by our method are comparable to ones designed by the designer.The above methods are proposed for the automatic design of indoor spatial layout.In the optimization of indoor spatial layout,we propose a single-sample-driven indoor spatial layout optimization technology.Given an indoor scene and a smaller indoor space,we propose a scene retargeting technique which can reduce the scale of the indoor scene to the scale that the indoor space can cover.The main idea of the algorithm is to rearrange the objects in the indoor scene to obtain a new indoor scene and to achieve the consistency in spatial perception between the new scene and the original scene by maintaining the relevant metrics.Our scene retargeting method preserves the layout and structure of the indoor scene while reducing the scale of the scene while maintaining the spatial perception of the scene.During the optimization process,in addition to maintaining the layout and structure of the indoor scene,we also need to reduce the distortion and artifacts of objects in the scene as much as possible.The experimental results show that our scene retargeting method is effective,and comparing with the result of uniform scaling,it further shows the effectiveness of our retargeting algorithm.