CN114912175B - Method for automatically generating vectorized indoor layout plan - Google Patents
Method for automatically generating vectorized indoor layout plan Download PDFInfo
- Publication number
- CN114912175B CN114912175B CN202210493144.XA CN202210493144A CN114912175B CN 114912175 B CN114912175 B CN 114912175B CN 202210493144 A CN202210493144 A CN 202210493144A CN 114912175 B CN114912175 B CN 114912175B
- Authority
- CN
- China
- Prior art keywords
- graph
- layout plan
- window
- boundary
- candidate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 26
- 238000010586 diagram Methods 0.000 claims abstract description 13
- 230000008878 coupling Effects 0.000 claims description 5
- 238000010168 coupling process Methods 0.000 claims description 5
- 238000005859 coupling reaction Methods 0.000 claims description 5
- 238000013461 design Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 2
- 238000012938 design process Methods 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000002910 structure generation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/13—Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Civil Engineering (AREA)
- Architecture (AREA)
- Medical Informatics (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Structural Engineering (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for automatically generating a vectorized indoor layout plan, which comprises the following steps: step 1, determining the boundary of an indoor layout plan; step 2, determining the type and the position of the window from the boundary; step 3, taking the top left corner vertex of the boundary as a generating starting point of the graph, namely the current generated graph; step 4, taking the boundary, the window and the current generation diagram as input, and generating layout semantics by a semantic prediction model based on a deep neural network; step 5, taking boundaries, windows, layout semantics and a current generation diagram as input, predicting new vertexes and edges by using a diagram generation model based on a deep neural network, and updating the current generation diagram; step 6, repeating the steps 4 and 5 until the new candidate vertexes and candidate edges are not predicted in the step 5, and obtaining a connected graph and layout semantics at the moment; and 7, obtaining a vectorized indoor layout plan based on the connected graph and the layout semantics.
Description
Technical Field
The invention relates to the field of a vectorization indoor layout plan generating method, in particular to a method for automatically generating vectorization indoor layout plan.
Background
The vectorized indoor layout plan is the basis of indoor design and scene modeling, and is widely applied to the fields of animation movies, computer games, virtual reality and the like. The traditional layout plan design is designed manually by a designer or is designed with the aid of interactive modeling software, so that the design process is tedious and tedious, and a great deal of time and effort are consumed. Therefore, the automatic generation of the vectorized indoor layout plan has important application value.
Deep learning is the ability of a computer to learn itself and improve. The deep neural network is the basis of deep learning, and the deep neural network has stronger characterization capability and stronger learning capability. The real world has accumulated a large number of high quality indoor layout plans, one straightforward idea is to learn design criteria from existing designs and apply them to the automatic generation of layout plans, which can be learned and modeled by deep neural network-based generation methods.
The layout plan may be represented by the structure of a graph, wherein vertices of the graph represent nodes in the layout plan and edges of the graph represent wall segments in the layout plan. Semantic information of rooms associated with vertices serves as attributes of the corresponding vertices. In addition to the special cases of the isolated wall, the layout plan view can be expressed as a connected graph. The problem of generating the layout plan can thus be converted into the problem of generating the connected graph. Breadth-first traversal may be used for graph generation, which is one traversal method of a graph: taking traversal of the connected graph as an example, starting from a certain vertex v in the graph, sequentially accessing non-accessed adjacent points of v, then sequentially accessing adjacent points of the points from the adjacent points respectively, and repeating the process until all vertexes in the graph are accessed.
And (3) giving the boundary of the indoor layout plan, sequentially predicting the vertexes and the edges of the graph in an iterative mode through a depth neural network according to the breadth-first traversal order, generating the vertexes and the edges of the graph, and finally obtaining the connected graph. Once the connected graph is generated, the vectorized layout plan can be directly obtained, so that automatic generation of the vectorized indoor layout plan can be realized. Therefore, the method for generating the vectorized indoor layout plan can be constructed based on the deep neural network so as to solve the problem of manually generating the vectorized indoor layout plan.
Disclosure of Invention
The invention aims to provide a method for automatically generating a vectorized indoor layout plan, which aims to solve the problems of tedious and lengthy design process, low automation and intelligent degree of the indoor layout plan in the prior art.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a method of automatically generating a vectorized indoor layout plan, comprising the steps of:
step 1, determining a boundary of an indoor layout plan, wherein the boundary is of a polygonal structure and comprises a front door;
step 2, determining the type and the position of the window from the boundary obtained in the step 1:
predicting the type and the position of a window based on a window prediction model of a depth neural network, wherein the predicted window is divided into two types, and the type and the position of the window provide priori guidance for generating a layout plan according to the floor window of a living room and the common window of other rooms in which the window is positioned;
step 3, taking the top left corner vertex of the boundary obtained in the step 1 as a generation starting point of the graph, wherein the starting point is the current generation graph, and the current generation graph is continuously updated along with the iterative generation of the subsequent steps;
step 4, taking the boundary of the indoor layout plan determined in the step 1, the type and the position of the window determined in the step 2 and the current generated graph obtained in the step 3 as inputs, and predicting layout semantics by a semantic prediction model based on a deep neural network;
step 5, taking the boundary of the indoor layout plan, the current generated graph obtained in the step 3 and the layout semantics obtained in the step 4 as inputs, and generating model prediction candidate vertexes and candidate edges by using the graph based on the deep neural network; if a candidate edge exists between the vertex of the current generation diagram and the candidate vertex, adding the candidate vertex and the candidate edge into the current generation diagram as new vertex and edge; if a candidate edge also exists between the two newly added vertexes, the candidate edge is also added into the current generated graph as a new edge;
step 6, repeating the steps 4 and 5, replacing the current generated graph obtained in the step 3 with an updated current generated graph, generating continuously updated layout semantics and the current generated graph through coupling of a semantic prediction model and a graph generation model until new candidate vertexes and candidate edges are not predicted in the step 5, obtaining the connected graph and semantic information of the indoor layout plan at this time, wherein the current generated graph finally obtained through the graph generation model is the connected graph of the indoor layout plan, and the layout semantics finally obtained through the semantic prediction model is the semantic information of the indoor layout plan;
and 7, obtaining a vectorized indoor layout plan based on the connected graph and the semantic information obtained in the step 6.
In a further step 2, the position and type of the window on the boundary are predicted by a window prediction model based on a deep neural network.
In a further step 5, a depth neural network based graph generation model predicts new vertices and edges in order of breadth-first traversal.
In a further step 6, continuously updated layout semantics and the current generated map are generated by coupling the semantic prediction model and the map generation model based on the deep neural network.
In a further step 7, through the minimum ring operation of the traversal diagram, finding all the minimum rings from the communication diagram obtained in the step 6, namely, a room; for each room, semanteme of a pixel corresponding to the center of the room is selected as a room label in layout semanteme, the room label is distributed to all connected graph vertexes forming the corresponding room to serve as vertex attributes, a complete connected graph with semantic attributes is obtained, and a vectorized indoor layout plan can be obtained according to the complete connected graph with semantic attributes.
The invention has the advantages that: according to the boundary of the given indoor layout plan, the invention can automatically generate a high-quality plan to assist a designer in creation and design. The method is simple, easy to deploy in a computer, high in automation degree, capable of reducing complicated manual design, and capable of fully and reliably theoretically supporting the model based on the deep neural network.
Drawings
Fig. 1 is a flow chart of the operation of the present invention.
Fig. 2 is a schematic view of window prediction.
Fig. 3 is a schematic diagram of a layout plan.
Detailed Description
The invention will be further described with reference to the drawings and examples.
As shown in fig. 1, the method for automatically generating the vectorized indoor layout plan according to the present invention comprises the following steps:
step 1, giving the boundary of the indoor layout plan as input.
Step 2, as shown in fig. 2, the type and position of the window are determined from the boundary obtained in step 1. The type and position of the window are predicted by a window prediction model based on a deep neural network. Predicted windows are classified into two types, a floor window of a living room and a normal window of other rooms according to a room where the window is located. The type and the position of the window provide priori guidance for generating the layout plan, and the generation quality of the layout plan can be improved.
In the invention, the window prediction model based on the deep neural network adopts an adjusted D-LinkNet model, and the architecture of the window prediction model mainly comprises three modules, namely an encoder module, a cavity convolution layer module and a decoder module. To train the model, a public indoor layout plan dataset RPLAN is employed. The data in the dataset does not contain windows, which are generated by a rule-based method for model training.
And 3, taking the vertex of the upper left corner of the boundary obtained in the step 1 as a generating starting point of the graph, wherein the starting point is the current generated graph. The current generation map is updated continuously with the iterative generation of subsequent steps.
And 4, as shown in a layout semantic generation module in fig. 3, taking the boundary of the indoor layout plan determined in the step 1, the type and the position of the window determined in the step 2 (the boundary in the subsequent step contains the window by default) and the current generated graph obtained in the step 3 as inputs, and predicting the layout semantic by a semantic prediction model based on a deep neural network.
In the invention, a semantic prediction model based on a deep neural network adopts an adjusted D-LinkNet model, and the architecture of the semantic prediction model mainly comprises three modules, namely an encoder module, a cavity convolution layer module and a decoder module. To train the model, a public indoor layout plan dataset RPLAN is employed. The data in the dataset contains layout semantics that can be used directly for model training.
Step 5, as shown in a layout structure generation module in fig. 3, taking the boundary of the indoor layout plan, the current generated graph obtained in the step 3 and the layout semantics obtained in the step 4 as inputs, and generating a model prediction candidate vertex and candidate edge by using the graph based on the deep neural network; if a candidate edge exists between the vertex of the current generation diagram and the candidate vertex, adding the candidate vertex and the candidate edge into the current generation diagram as new vertex and edge; if there is also a candidate edge between the two newly added vertices, the candidate edge is also added to the current generated graph as a new edge.
In the invention, a graph generation model based on a deep neural network adopts an adjusted D-LinkNet model, and the architecture of the graph generation model mainly comprises three modules, namely an encoder module, a cavity convolution layer module and a decoder module. To train the model, a public indoor layout plan dataset RPLAN is employed. The data in the dataset does not contain a graph structure, which is directly extracted from the indoor layout plan in the dataset for model training.
Step 6, as shown in fig. 3, repeating steps 4 and 5 (in step 4 and 5, the current generated graph obtained in step 3 needs to be replaced by an updated current generated graph), and generating continuously updated layout semantics and the current generated graph through coupling of the semantic prediction model and the graph generation model until new candidate vertexes and candidate edges are not predicted in step 5, and at this time, obtaining a connected graph and semantic information of the indoor layout plan: the current generated graph finally obtained through the graph generation model is the connected graph of the indoor layout plan, and the layout semantics finally obtained through the semantic prediction model is the semantic information of the indoor layout plan.
And 7, finding all the minimum rings in the connected graph obtained in the step 6 through the minimum ring operation of the traversal graph, namely, a room. For each room, the semantics of the pixel corresponding to the center of the room is selected as the room label in the layout semantics. And assigning the room labels to all connected graph vertices forming the room as vertex attributes to obtain a complete connected graph with semantic attributes. Based on the complete connected graph with semantic attributes, a vectorized layout plan can be directly obtained.
The embodiments of the present invention are merely described in terms of preferred embodiments of the present invention, and are not intended to limit the spirit and scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope of the present invention, and the technical content of the present invention as claimed is fully described in the claims.
Claims (5)
1. A method for automatically generating a vectorized indoor layout plan, comprising the steps of:
step 1, determining a boundary of an indoor layout plan, wherein the boundary is of a polygonal structure and comprises a front door;
step 2, determining the type and the position of the window from the boundary obtained in the step 1:
predicting the type and the position of a window based on a window prediction model of a depth neural network, wherein the predicted window is divided into two types, and the type and the position of the window provide priori guidance for generating a layout plan according to the floor window of a living room and the common window of other rooms in which the window is positioned;
step 3, taking the top left corner vertex of the boundary obtained in the step 1 as a generation starting point of the graph, wherein the starting point is the current generation graph, and the current generation graph is continuously updated along with the iterative generation of the subsequent steps;
step 4, taking the boundary of the indoor layout plan determined in the step 1, the type and the position of the window determined in the step 2 and the current generated graph obtained in the step 3 as inputs, and predicting layout semantics by a semantic prediction model based on a deep neural network;
step 5, taking the boundary of the indoor layout plan, the current generated graph obtained in the step 3 and the layout semantics obtained in the step 4 as inputs, and generating model prediction candidate vertexes and candidate edges by using the graph based on the deep neural network; if a candidate edge exists between the vertex of the current generation diagram and the candidate vertex, adding the candidate vertex and the candidate edge into the current generation diagram as new vertex and edge; if a candidate edge also exists between the two newly added vertexes, the candidate edge is also added into the current generated graph as a new edge;
step 6, repeating the steps 4 and 5, replacing the current generated graph obtained in the step 3 with an updated current generated graph, generating continuously updated layout semantics and the current generated graph through coupling of a semantic prediction model and a graph generation model until new candidate vertexes and candidate edges are not predicted in the step 5, obtaining the connected graph and semantic information of the indoor layout plan at this time, wherein the current generated graph finally obtained through the graph generation model is the connected graph of the indoor layout plan, and the layout semantics finally obtained through the semantic prediction model is the semantic information of the indoor layout plan;
and 7, obtaining a vectorized indoor layout plan based on the connected graph and the semantic information obtained in the step 6.
2. The method of automatically generating a vectorized indoor layout plan of claim 1 wherein in step 2, the location and type of windows on the boundary are predicted by a window prediction model based on a deep neural network.
3. The method of claim 1, wherein in step 5, the depth neural network-based graph generation model predicts new vertices and edges in order of breadth-first traversal.
4. The method of automatically generating a vectorized indoor layout plan of claim 1 wherein in step 6, constantly updated layout semantics and the current generated map are generated by coupling a semantic prediction model based on a deep neural network and a map generation model.
5. The method for automatically generating a vectorized indoor layout plan according to claim 1, wherein in step 7, by traversing the minimum ring operation of the graph, all the minimum rings are found from the connected graph obtained in step 6, namely, the room; for each room, semanteme of a pixel corresponding to the center of the room is selected as a room label in layout semanteme, the room label is distributed to all connected graph vertexes forming the corresponding room to serve as vertex attributes, a complete connected graph with semantic attributes is obtained, and a vectorized indoor layout plan can be obtained according to the complete connected graph with semantic attributes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210493144.XA CN114912175B (en) | 2022-05-07 | 2022-05-07 | Method for automatically generating vectorized indoor layout plan |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210493144.XA CN114912175B (en) | 2022-05-07 | 2022-05-07 | Method for automatically generating vectorized indoor layout plan |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114912175A CN114912175A (en) | 2022-08-16 |
| CN114912175B true CN114912175B (en) | 2024-03-08 |
Family
ID=82767750
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210493144.XA Active CN114912175B (en) | 2022-05-07 | 2022-05-07 | Method for automatically generating vectorized indoor layout plan |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114912175B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110033513A (en) * | 2017-12-28 | 2019-07-19 | 达索***公司 | Generate the 3D model for indicating building |
| CN110060255A (en) * | 2017-12-28 | 2019-07-26 | 达索***公司 | Semantic segmentation is carried out to 2D plan view using classifier pixel-by-pixel |
| CN112417539A (en) * | 2020-11-16 | 2021-02-26 | 杭州群核信息技术有限公司 | Method, device and system for designing house type based on language description |
| CN112907750A (en) * | 2021-03-05 | 2021-06-04 | 齐鲁工业大学 | Indoor scene layout estimation method and system based on convolutional neural network |
-
2022
- 2022-05-07 CN CN202210493144.XA patent/CN114912175B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110033513A (en) * | 2017-12-28 | 2019-07-19 | 达索***公司 | Generate the 3D model for indicating building |
| CN110060255A (en) * | 2017-12-28 | 2019-07-26 | 达索***公司 | Semantic segmentation is carried out to 2D plan view using classifier pixel-by-pixel |
| CN112417539A (en) * | 2020-11-16 | 2021-02-26 | 杭州群核信息技术有限公司 | Method, device and system for designing house type based on language description |
| CN112907750A (en) * | 2021-03-05 | 2021-06-04 | 齐鲁工业大学 | Indoor scene layout estimation method and system based on convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| 基于CNN的住宅平面图元素识别与布局语义分析;张蔚;王腾;卢政达;杨思宇;孟维亮;郭建伟;张晓鹏;;中国体视学与图像分析;20200625(第02期);全文 * |
| 融合信息化边界和多模态特征的室内空间布局估计;刘天亮;陆泮宇;戴修斌;刘峰;罗杰波;;南京信息工程大学学报(自然科学版);20191128(第06期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114912175A (en) | 2022-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109849576B (en) | Method for assisting drawing by referring to gray level diagram | |
| CN109816772B (en) | Processing method for rapidly generating virtual reality scene from CAD house type file | |
| US20220148327A1 (en) | Machine learning techniques for extracting floorplan elements from architectural drawings | |
| CN112445876A (en) | Entity alignment method and system fusing structure, attribute and relationship information | |
| CN109697451A (en) | Similar image clustering method and device, storage medium, electronic equipment | |
| CN112417539A (en) | Method, device and system for designing house type based on language description | |
| Hou et al. | Drawing-based procedural modeling of Chinese architectures | |
| CN115470707A (en) | City scene simulation system | |
| CN111744199A (en) | Image processing method and device, computer readable storage medium and electronic device | |
| Sun et al. | Wallplan: synthesizing floorplans by learning to generate wall graphs | |
| CN114782530A (en) | Three-dimensional semantic map construction method, device, equipment and medium under indoor scene | |
| Hu et al. | Extended interactive and procedural modeling method for ancient Chinese architecture | |
| van Aanholt et al. | Declarative procedural generation of architecture with semantic architectural profiles | |
| She et al. | 3D building model simplification method considering both model mesh and building structure | |
| Yang et al. | Automated semantics and topology representation of residential-building space using floor-plan raster maps | |
| CN114912175B (en) | Method for automatically generating vectorized indoor layout plan | |
| Parente et al. | Integration of convolutional and adversarial networks into building design: A review | |
| Besuievsky et al. | Customizable lod for procedural architecture | |
| CN117710534A (en) | Animation collaborative making method based on improved teaching and learning optimization algorithm | |
| Fang et al. | A framework for human-computer interactive street network design based on a multi-stage deep learning approach | |
| Besuievsky et al. | A Configurable LoD for Procedural Urban Models intended for Daylight Simulation. | |
| CN116363446A (en) | Zero sample image classification method, device, terminal and medium | |
| CN116152451A (en) | Multidimensional parameterized city information model construction method, system and computer equipment | |
| Wang et al. | [Retracted] An Intelligent Animation Interaction Design Algorithm Based on Example and Parameterization | |
| Adão | Ontology-based procedural modelling of traversable buildings composed by arbitrary shapes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |