A method and device for generating a three-dimensional model of architectural space design

Abstract

The present application provides a method and device for generating a three-dimensional model of architectural space design, the method comprising: obtaining a three-dimensional model of the spatial layout of the target building based on a plane structure diagram and text description information of the target building; using a plurality of three-dimensional models of facilities to be screened to generate an object sequence corresponding to the scene, and determining a plurality of target facility three-dimensional models from the plurality of three-dimensional models of facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model; placing each target facility model in the model area corresponding to the scene in the three-dimensional model of the spatial layout based on the spatial attributes of each target facility three-dimensional model, so as to generate a three-dimensional model of spatial design that meets the requirements of the interior design. Through the method and device, the reconstructed interior layout is organically combined with the generated facility model to complete the overall spatial design for a specific scene.

Description

Technical Field

[0001] The present application relates to the field of space design technology, and in particular to a method and device for generating a three-dimensional model of architectural space design. Background Art

[0002] Architectural space layout and design are often crucial links in construction projects. They directly affect the functionality, aesthetics and comfort of the building. Reasonable design of space layout and color matching of facilities is the key to creating a comfortable, practical and beautiful building environment.

[0003] Traditional architectural space planning and design often relies on inherent indoor floor plans. Indoor floor plans have basic structural information, but cannot flexibly respond to users' needs for indoor layouts, resulting in the 3D model generated based on the original floor plan being difficult to meet the real needs of space design. In addition, users do not have a variety of facility models to choose from in the generated 3D model, and cannot intuitively understand the overall design style of the space through the 3D model. Therefore, how to generate a 3D space design model with a comprehensive layout that meets the user's functional and design style requirements has become a technical issue that cannot be underestimated. Summary of the invention

[0004] In view of this, the purpose of this application is to provide a method and device for generating a three-dimensional model of architectural space design, organically combining the reconstructed interior layout with the generated facility model, completing the overall space design for a specific scene, and generating an overall space solution that meets the user's functional requirements and design style requirements.

[0005] In a first aspect, an embodiment of the present application provides a method for generating a three-dimensional model of a building space design, the generating method comprising:

[0006] Acquire a plan structure diagram of a target building and text description information for interior design requirements of the target building, and generate a three-dimensional model of the spatial layout of the target building based on the plan structure diagram and the text description information;

[0007] For each scene in the spatial layout three-dimensional model, a plurality of to-be-screened three-dimensional facility models matching the scene style are screened out from a pre-built facility model library based on the text description information;

[0008] Generate an object sequence corresponding to the scene using the plurality of the three-dimensional models of the facilities to be screened, and determine a plurality of target facility three-dimensional models from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene;

[0009] Each target facility model is placed in a model area corresponding to the scene in the spatial layout three-dimensional model based on the spatial attributes of each target facility three-dimensional model to generate a spatial design three-dimensional model that meets the interior design requirements.

[0010] Furthermore, generating the spatial layout three-dimensional model of the target building based on the planar structure diagram and the text description information includes:

[0011] Generate a layout fusion feature vector based on the plan layout features of the plan structure diagram and the text layout features of the text description information;

[0012] Inputting the layout fusion feature vector into a pre-trained spatial layout generation model to obtain the spatial layout of the target building, and generating a scene function matrix based on the spatial layout, the layout fusion feature vector and the text layout feature;

[0013] The scene function matrix is ​​input into a pre-trained three-dimensional generative adversarial network to generate a three-dimensional model of the spatial layout of the target building.

[0014] Furthermore, before obtaining the plan structure diagram of the target building and the text description information for the interior design requirements of the target building, the generating method further includes:

[0015] For each facility type, images of the original model corresponding to the facility type under multiple viewing angles and information text data corresponding to the facility type are obtained;

[0016] Extracting visual features from the plurality of images to obtain a visual style feature vector, and extracting text features from the information text data to obtain a text feature vector;

[0017] Inputting the visual style feature vector and the text feature vector into a pre-trained cross-modal feature fusion model for cross-modal feature fusion to obtain a cross-modal feature vector;

[0018] Inputting the cross-modal feature vector into a pre-trained diffusion model to obtain an initial three-dimensional model corresponding to the facility type;

[0019] The initial three-dimensional model and the cross-modal feature vector are input into a pre-trained texture generative adversarial network to obtain a basic three-dimensional model of the facility type to generate the facility model library; wherein the surface of the basic three-dimensional model is provided with a facility texture map that conforms to the information text data.

[0020] Furthermore, the step of selecting a plurality of to-be-selected three-dimensional facility models that match the scene style of the scene from a pre-built facility model library based on the text description information includes:

[0021] Determine a scene facility model corresponding to the text description information from a plurality of basic three-dimensional models in the facility model library;

[0022] Determine the first component type of each scene facility model and the second component type of each basic three-dimensional model in the facility model library, determine the style influence weight between the scene facility model and the basic three-dimensional model from a pre-built facility knowledge graph, and query the first facility style vector of each scene facility model in the facility style vector library;

[0023] Calculate a scene style vector of the scene based on the first component type, the second component type, the style influence weight, and the first facility style vector;

[0024] Querying the second facility style vector of each basic three-dimensional model in the facility style vector library, and calculating the style similarity between the scene style vector and the second facility style vector of each basic three-dimensional model;

[0025] The multiple basic three-dimensional models are sorted using the style similarity corresponding to each basic three-dimensional model, and multiple three-dimensional facility models to be screened are determined from the sorting.

[0026] Furthermore, the facility knowledge graph is constructed through the following steps:

[0027] Obtain multiple scene sample images, and use the target recognition algorithm to identify the facility matching combinations between multiple facility types in each scene sample image;

[0028] The facility collocation combinations of multiple facility types are analyzed through an association rule mining algorithm to determine the correlation between the multiple facility types to generate the facility knowledge graph.

[0029] Furthermore, the object sequence includes a category sequence, a position sequence, a direction sequence and a size sequence; the object sequence corresponding to the scene is generated by using the plurality of the three-dimensional models of the facilities to be screened, and a plurality of target facility three-dimensional models are determined from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model, including:

[0030] Inputting each three-dimensional model of the facility to be screened into the category model to determine the category sequence, and inputting the category sequence into the direction model to determine the direction sequence;

[0031] Inputting the category sequence and the position sequence into a position model to determine the position sequence, and inputting the category sequence, the direction sequence, and the position sequence into a size model to determine the size sequence;

[0032] Determine a plurality of target facility three-dimensional models from the plurality of facility three-dimensional models to be screened based on the category sequence;

[0033] The placement position of each target facility three-dimensional model is determined according to the position sequence, the placement direction of each target facility three-dimensional model is determined according to the direction sequence, and the size of each target facility three-dimensional model is determined according to the size sequence.

[0034] Furthermore, after generating the three-dimensional spatial design model that conforms to the text description, the generating method further includes:

[0035] Based on the facility knowledge graph, each facility node in the spatial design three-dimensional model is encoded and added to the neural network graph to obtain a scene generation design neural network graph of the target building;

[0036] Each facility node is input into a multi-layer perceptron to obtain a design scheme score of the spatial design three-dimensional model.

[0037] In a second aspect, an embodiment of the present application further provides a device for generating a three-dimensional model of a building space design, the generating device comprising:

[0038] A spatial layout three-dimensional model generation module, used to obtain a plane structure diagram of a target building and text description information for interior design requirements of the target building, and generate a spatial layout three-dimensional model of the target building based on the plane structure diagram and the text description information;

[0039] A module for determining the three-dimensional model of facilities to be screened is used to screen out a plurality of three-dimensional models of facilities to be screened that match the scene style of the scene from a pre-built facility model library based on the text description information for each scene in the three-dimensional model of the spatial layout;

[0040] A target facility three-dimensional model determination module is used to generate an object sequence corresponding to the scene using the plurality of the three-dimensional models of the facilities to be screened, and determine a plurality of target facility three-dimensional models from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene;

[0041] The space design three-dimensional model generation module is used to place each target facility model in the model area corresponding to the scene in the space layout three-dimensional model based on the spatial attributes of each target facility three-dimensional model, so as to generate a space design three-dimensional model that meets the interior design requirements.

[0042] In a third aspect, an embodiment of the present application further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method for generating a three-dimensional model of architectural space design are performed as described above.

[0043] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for generating a three-dimensional model of architectural space design as described above are executed.

[0044] A method for generating a three-dimensional model of architectural space design provided in an embodiment of the present application comprises the steps of obtaining a plan structure diagram of a target building and text description information of interior design requirements for the target building, and generating a three-dimensional model of the spatial layout of the target building based on the plan structure diagram and the text description information; for each scene in the three-dimensional spatial layout model, screening out a plurality of three-dimensional facility models to be screened that match the scene style of the scene from a pre-built facility model library based on the text description information; generating an object sequence corresponding to the scene using the plurality of three-dimensional facility models to be screened, and determining a plurality of target facility three-dimensional models from the plurality of three-dimensional facility models to be screened based on the object sequence, as well as spatial attributes of each target facility three-dimensional model; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene; placing each target facility model in a model area corresponding to the scene in the three-dimensional spatial layout model based on the spatial attributes of each target facility three-dimensional model, so as to generate a three-dimensional spatial design model that meets the interior design requirements.

[0045] This application realizes the generative design of facilities in the overall indoor scene by integrating the facility knowledge graph and the facility generation model. According to the room structure and the user's text requirements, the problem of scene style customization and the selection and layout coordination of facilities in the scene is solved. Input the scene structure plan and the user's text description, select the scene infrastructure required by the user from the facility model library, and select the facility three-dimensional model that meets the scene style from it, and then use multiple facility three-dimensional models to generate the object sequence corresponding to the scene to realize the final facility layout generation. The reconstructed indoor layout is organically combined with the generated facility model to complete the overall space design for a specific scene, and generate an overall space solution that meets the user's functional requirements and design style requirements.

[0046] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 A flowchart of a method for generating a three-dimensional model of a building space design provided in an embodiment of the present application;

[0049] Figure 2 One of the structural schematic diagrams of a device for generating a three-dimensional model of a building space design provided in an embodiment of the present application;

[0050] Figure 3 A second structural schematic diagram of a device for generating a three-dimensional model of a building space design provided in an embodiment of the present application;

[0051] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, each other embodiment obtained by those skilled in the art without making creative work belongs to the scope of protection of the present application.

[0053] First, the application scenarios to which the present application is applicable are introduced. The present application can be applied in the field of space design technology.

[0054] Architectural space layout and design are often crucial links in construction projects. They directly affect the functionality, aesthetics and comfort of the building. Reasonable design of space layout and color matching of facilities is the key to creating a comfortable, practical and beautiful building environment.

[0055] Research has found that traditional architectural space planning and design often relies on inherent indoor floor plans. Indoor floor plans have basic structural information, but cannot flexibly respond to users' needs for indoor layouts, resulting in the 3D model generated based on the original floor plan being difficult to meet the real needs of space design. In addition, users do not have a variety of facility models to choose from in the generated 3D model, and cannot intuitively understand the overall design style of the space through the 3D model. Therefore, how to generate a 3D model of space design that has a comprehensive layout and meets the user's functional and design style requirements has become a technical issue that cannot be underestimated.

[0056] Based on this, the embodiment of the present application provides a method and a device for generating a three-dimensional model of architectural space design, which organically combines the reconstructed interior layout with the generated facility model, completes the overall space design for a specific scene, and generates an overall space solution that meets the user's functional requirements and design style requirements.

[0057] See also Figure 1 , Figure 1 The following is a flow chart of a method for generating a three-dimensional model of a building space design provided in an embodiment of the present application. Figure 1 As shown in, the generation method provided in the embodiment of the present application includes:

[0058] S101, obtaining a plan structure diagram of a target building and text description information for interior design requirements of the target building, and generating a three-dimensional model of the spatial layout of the target building based on the plan structure diagram and the text description information.

[0059] It should be noted that the plan structure diagram of the target building is a plan view in architectural design, which is used to show the layout and structure of the building on the horizontal plane. It is usually drawn from a bird's-eye view of the building, showing the location and size of each room, corridor, staircase, door, window, wall, column, equipment and other structural elements of the building. The text description information refers to the text description of the layout design requirements and style design requirements of the target building input by the user.

[0060] Regarding the above step S101, during the specific implementation, the plane structure diagram of the target building and text description information of the interior design requirements of the target building are obtained, and a three-dimensional model of the spatial layout of the target building is generated based on the plane structure diagram and the text description information.

[0061] As an optional embodiment, with respect to the above step S101, generating the spatial layout three-dimensional model of the target building based on the planar structure diagram and the text description information includes:

[0062] Step 1011: Generate a layout fusion feature vector based on the plan layout features of the plan structure diagram and the text layout features of the text description information.

[0063] For the above step 1011, in the specific implementation, first determine the floor plan layout features of the floor plan structure diagram and the text layout features of the text description information. Then perform multimodal feature fusion on the floor plan layout features and the text layout features to obtain a layout fusion feature vector. Here, according to the embodiment provided by the present application, a multi-stream neural network architecture is used to fuse the embedded text layout requirement features and floor plan structure features to reconstruct the indoor layout. The multi-stream neural network processes different modal information by selecting appropriate neural networks and multi-stream weights to solve the conflict problem of inconsistent requirements in multimodal fusion.

[0064] Here, as an optional embodiment, a convolutional neural network is used to extract the spatial geometric features of the plane structure diagram, and the spatial layout information of different scales in the spatial geometric features is determined by a multi-scale convolution kernel, and the plane layout features are constructed based on the spatial layout information. Then, the Transformer model is used to encode the text description information, determine the spatial layout semantic features, and construct the text layout features based on the spatial layout semantic features. Finally, the layout differential relationship features between the plane layout features and the text layout features are extracted, and multimodal feature fusion is performed on the plane layout features, text layout features, and layout differential relationship features to obtain a layout fusion feature vector.

[0065] Step 1012, input the layout fusion feature vector into a pre-trained spatial layout generation model to obtain the spatial layout of the target building, and generate a scene function matrix based on the spatial layout, the layout fusion feature vector and the text layout feature.

[0066] For the above step 1012, in the specific implementation, the layout fusion feature vector obtained in the above step 1011 is input into the pre-trained spatial layout generation model to generate the spatial layout of the target building. Then, by parsing the spatial layout, the layout fusion feature vector and the text layout features extracted from the text description information, a specific scene function matrix is ​​generated. Here, as an optional embodiment, when generating the spatial layout, the layout fusion feature vector is input into the encoder of the spatial layout generation model, and the layout fusion feature vector is encoded by the encoder to obtain the spatial layout latent variables, and then the spatial layout latent variables are input into the decoder of the spatial layout generation model to obtain the spatial layout. The spatial layout generation model here adopts a generation model based on a conditional variational autoencoder. The conditional variational autoencoder can generate diversified content that meets specific conditions by modeling the potential distribution of the data, thereby enhancing the flexibility and personalization of the generated content.

[0067] Step 1013: input the scene function matrix into a pre-trained three-dimensional generative adversarial network to generate a three-dimensional model of the spatial layout of the target building.

[0068] Here, the three-dimensional generative adversarial network improves the authenticity of the generated model through an adversarial learning mechanism to meet the layout and functional requirements. This application uses a three-dimensional generative adversarial network to generate a three-dimensional model of the spatial layout that matches the input features. For the above step 1013, in the specific implementation, the scene function matrix is ​​used as input, and the three-dimensional model of the spatial layout of the target building is generated through the generator of the three-dimensional generative adversarial network.

[0069] According to the embodiment provided by the present application, before obtaining the plan structure diagram of the target building and the text description information for the interior design requirements of the target building in the above step S101, the generation method further includes:

[0070] I: For each facility type, obtain images of the original model corresponding to the facility type under multiple viewing angles, as well as information text data corresponding to the facility type.

[0071] Here, the facility type refers to the types of various facilities set in the building space. As an example, when a home building is designed for space, the facility type can be the type of furniture in the home building, such as sofas, chairs, etc., which is not specifically limited in this application. The original model refers to a pre-built three-dimensional model of the facility type. Multiple perspectives can include a front perspective, a left perspective, a right perspective, a top perspective, a back perspective, and a bottom perspective. The information text data can include the IFC standard product information of the target facility and the description information of the facility style that the user currently wants to generate.

[0072] In the above step I, when it is specifically implemented, for each facility type, the original model corresponding to the facility type is obtained, and images of the original model under multiple viewing angles are obtained. Here, continuing the above embodiment, when the multiple viewing angles include the front viewing angle, the left viewing angle, the right viewing angle, the top viewing angle, the back viewing angle and the bottom viewing angle, the original model I There are 6 images, represented as I 1, I 2,..., I 6. At the same time, obtain the information text data entered by the user for the facility type.

[0073] II: extracting visual features from the plurality of images to obtain a visual style feature vector, and extracting text features from the information text data to obtain a text feature vector.

[0074] Regarding the above step II, in the specific implementation, feature extraction is performed on the acquired data, and visual style feature vectors and text feature vectors are extracted from the multiple images and information text data respectively.

[0075] Specifically, as an optional implementation, for each image, the image is converted into multiple image tag sequences, and the multiple image tag sequences are input into the image encoder to obtain an image embedding vector. Then the image embedding vector is input into the ViT model for encoding to obtain image features, and the image features are averaged pooled to obtain the image feature vector of the image. The image feature vectors of the multiple images are vector concatenated to obtain a target feature vector, and the target feature vector is normalized to obtain the visual style feature vector. The extraction of the text feature vector is the same as the extraction of the visual style feature vector. The information text data is converted into multiple text tag sequences, and the multiple text tag sequences are input into the text encoder for embedding to obtain a text embedding vector. The text embedding vector is input into the Transformer model for encoding to obtain text features, and the text features are averaged pooled to obtain the initial feature vector of the information text data. The initial feature vector is normalized to obtain the text feature vector of the information text data.

[0076] III: Inputting the visual style feature vector and the text feature vector into a pre-trained cross-modal feature fusion model for cross-modal feature fusion to obtain a cross-modal feature vector.

[0077] For the above step III, in the specific implementation, the visual style feature vector and the text feature vector are input into the pre-trained cross-modal feature fusion model for cross-modal feature fusion, and a unified cross-modal feature vector is obtained through the cross-modal feature fusion model. Here, the cross-modal attention mechanism is used for feature fusion.

[0078] IV: Input the cross-modal feature vector into a pre-trained diffusion model to obtain an initial three-dimensional model corresponding to the facility type.

[0079] For the above step IV, in the specific implementation, the cross-modal feature vector is input into the pre-trained diffusion model to obtain the initial three-dimensional model corresponding to the facility type. Specifically, the diffusion model introduces a multi-level resolution strategy in different stages of the reverse diffusion process. The diffusion process in the diffusion model is usually divided into three stages: low-resolution stage, medium-resolution stage and high-resolution stage. The diffusion results output from each diffusion stage are fused to obtain the above initial three-dimensional model.

[0080] V: Inputting the initial three-dimensional model and the cross-modal feature vector into a pre-trained texture generative adversarial network to obtain a basic three-dimensional model of the facility type to generate the facility model library.

[0081] Here, the surface of the basic 3D model is set with a facility texture map that conforms to the information text data. In 3D modeling, a texture map usually refers to a 2D image that is applied to the surface of a 3D model to simulate the object's color, details, material and other visual characteristics. The texture map can contain color information, bump information, reflectivity information, etc., and this information is mapped to the surface of the 3D model through mapping technology, making the model look more realistic and detailed when rendered. The facility texture map is set on the surface of the basic 3D model to achieve high-fidelity reproduction of the facility texture, and enhance the detail richness and visual realism of the generated 3D model.

[0082] Specifically, according to the embodiment provided by the present application, the texture generation adversarial network uses a texture generation network based on the Wasserstein generation adversarial network (WGAN) to add high-quality textures to the generated initial three-dimensional model of the target facility to achieve a realistic rendering effect. For the above step V, in the specific implementation, the initial three-dimensional model and the cross-modal feature vector are input into the pre-trained texture generation adversarial network to obtain a basic three-dimensional model with a facility texture map set on the surface. That is, a facility model library can be generated through the basic three-dimensional models of multiple facility types.

[0083] S102: for each scene in the spatial layout three-dimensional model, a plurality of to-be-screened three-dimensional facility models matching the scene style are screened out from a pre-built facility model library based on the text description information.

[0084] Here, the scene refers to the scene corresponding to each area divided in the spatial layout three-dimensional model. As an example, when the spatial layout three-dimensional model corresponds to the indoor residential layout, the scene may include a living room, a bedroom, a kitchen, etc., which is not specifically limited in this application.

[0085] Regarding the above step S102, in the specific implementation, for each scene in the spatial layout three-dimensional model, multiple three-dimensional facility models to be screened that match the scene style of the scene are screened from a pre-built facility model library based on the text description information.

[0086] Specifically, with respect to the above step S102, the step of selecting a plurality of to-be-selected facility 3D models that match the scene style of the scene from a pre-built facility model library based on the text description information includes:

[0087] Step 1021: Determine a scene facility model corresponding to the text description information from a plurality of basic three-dimensional models in the facility model library.

[0088] Regarding the above step 1021, in the specific implementation, based on the constructed facility model library, the scene facility model required by the user is selected according to the text description information input by the user to obtain the pre-generated scene.

[0089] Step 1022, determine the first component type of each scene facility model and the second component type of each basic three-dimensional model in the facility model library, determine the style influence weight between the scene facility model and the basic three-dimensional model from the pre-built facility knowledge graph, and query the first facility style vector of each scene facility model in the facility style vector library.

[0090] In the above step 1022, when implementing it, determine the facility model of each scene The first component type , and facility model library G Each basic 3D model The second component type , and determine the style influence weight between the scene facility model and the basic 3D model from the pre-built facility knowledge graph , and query each scene facility model in the facility style vector library The first facility style vector .

[0091] As an optional embodiment, the facility knowledge graph is constructed through the following steps:

[0092] A: Obtain multiple scene sample images, and use the target recognition algorithm to identify the facility matching combinations between multiple facility types in each scene sample image.

[0093] B: Analyze the facility collocation combinations of multiple facility types through an association rule mining algorithm to determine the correlation between the multiple facility types to generate the facility knowledge graph.

[0094] For the above steps A-B, in the specific implementation, a facility knowledge graph is constructed, multiple scene sample images are obtained, and the facility matching combinations between multiple facility types in a large number of scene sample images are identified through the target recognition algorithm. Then, the correlation between different types of facilities is analyzed through the association rule mining algorithm, and finally the facility knowledge graph is generated. KG . Specifically, the facility knowledge graph KG It is expressed by the following formula:

[0095]

[0096] in, It is a combination of facilities identified in the scene. The node set that constitutes it.E It is the edge set representing the association relationship between facility objects. Indicates the association between facility objects A and B , determined by the confidence C(A→B) and support S(A→B).

[0097] Step 1023: Calculate the scene style vector of the scene based on the first component type, the second component type, the style influence weight, and the first facility style vector.

[0098] For the above 1023, in the specific implementation, the scene style is calculated, and the scene style vector of the scene is calculated based on the first component type, the second component type, the style influence weight and the first facility style vector obtained in the above step 1022. Specifically, the scene style vector of the scene Calculated by the following formula:

[0099]

[0100] Step 1024, query the second facility style vector of each basic three-dimensional model in the facility style vector library, calculate the style similarity between the scene style vector and the second facility style vector of each basic three-dimensional model, and use the style similarity corresponding to each basic three-dimensional model to sort the multiple basic three-dimensional models, and determine multiple facility three-dimensional models to be screened from the sorting.

[0101] In the specific implementation of step 1024, first query each basic 3D model in the facility style vector library The second facility style vector , and then generate the facility selection ranking. Specifically, by calculating the scene style vector A second facility style vector for each base 3D model The style similarity between the basic 3D models is calculated, and the style similarity corresponding to each basic 3D model is used to sort the multiple basic 3D models, and multiple facility 3D models to be screened are determined from the sorting. Here, the style similarity calculation results of each basic 3D model are sorted by the function rank, and finally the facility 3D models in this scene are selected. Before Generative Design 3D models of facilities to be screened. Specifically, the scene The ordering of multiple basic three-dimensional models is expressed by the following formula:

[0102]

[0103] S103, using the plurality of three-dimensional models of facilities to be screened to generate an object sequence corresponding to the scene, and determining a plurality of target facility three-dimensional models from the plurality of three-dimensional models of facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model.

[0104] In the specific implementation of step S103, multiple three-dimensional models of facilities to be screened are used to generate an object sequence corresponding to the scene, and multiple three-dimensional models of target facilities and spatial attributes of each three-dimensional model of target facilities are determined from the multiple three-dimensional models of facilities to be screened based on the object sequence. Specifically, the spatial attributes include the placement position, placement direction and size of the three-dimensional model of the target facility in the scene.

[0105] Specifically, according to the embodiment provided by the present application, the object sequence corresponding to the scene includes a category sequence, a direction sequence, a position sequence, and a size sequence. Here, Transformer performs well in various natural language processing tasks by treating the sentence as a word sequence. The present application realizes automatic regression prediction of the category, position, direction, and size information of each facility in the scene by treating the scene as a sequence of facilities. Treating each scene as an object sequence, Frequency Sorting. For each of the four attributes, category, location, orientation, and size, a separate model is trained to predict the corresponding label for the current facility. Facility attributes are predicted in an autoregressive manner, with each facility's attributes predicted based on previously predicted facilities. In this way, each model is conditioned on the output of the previous model. The category model depends on the categories of all previous facilities. The orientation model depends on the category of the current facility and the attributes of all previous facilities, and so on.

[0106] Further, with respect to the above step S103, the object sequence corresponding to the scene is generated by using the plurality of the three-dimensional models of the facilities to be screened, and a plurality of target facility three-dimensional models are determined from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model, including:

[0107] Step 1031, input each three-dimensional model of the facility to be screened into the category model to determine the category sequence, and input the category sequence into the direction model to determine the direction sequence.

[0108] In the above step 1031, in the specific implementation, each 3D model of the facility to be screened is input into the category model to determine the category sequence. Then the category sequence is input into the direction model to determine the direction sequence. Specifically, the category sequence { The distribution of} is decomposed into the following formula:

[0109]

[0110] in, represents the probability distribution, where The form represents conditional probability. For example, Indicates the property of the given object. and category models Under the conditions, the category Probability of occurrence.

[0111] The above formula represents the category model Represents the entire category sequence { }, which is decomposed into a series of conditional probabilities, namely, the object n Category is based on the previous observation sequence of the object and category models This means that the category prediction of the current object depends on the attributes of the previous object.

[0112] The distribution of the direction sequence is decomposed into the following formula:

[0113]

[0114] Direction Model For the entire direction sequence { Here, the direction of each object is based on the class of the current and previous objects and the previous observation sequence The prediction of the direction sequence not only depends on the previous direction, but also considers the category of the current object and the previous object attributes.

[0115] Step 1032: input the category sequence and the position sequence into a position model to determine the position sequence, and input the category sequence, the direction sequence, and the position sequence into a size model to determine the size sequence.

[0116] Regarding the above step 1032, in the specific implementation, the category sequence and the position sequence are input into the position model to determine the position sequence. Then, the category sequence, the direction sequence and the position sequence are input into the size model to determine the size sequence. Specifically, the position model and the size model and Make predictions based on the generated tags, such as { } is decomposed into the following formula:

[0117]

[0118] Location Model For position sequence { Here the current position of each object Based on all the previous categories ,direction , previous object position And the observation sequence This means that position prediction needs to comprehensively consider the category, direction and attributes of the previous object.

[0119] Thus, according to the above steps 1031-1032, the present application regards each scene as a sequence of objects, and Frequency Sorting. This ordering of facilities requires generating a unique representation for each scene, up to the ordering of facilities in the same category. The locations of the facilities are then normalized according to the maximum size of the space in the building and quantized to the range of [0, 255], resulting in the new coordinates of the facilities. . Similarly, the dimensions of each facility are length, width, and height ( , , ) will also be scaled and quantized. The orientation of the facility on the ground in space θ Quantized in the range [0, 359]. Therefore, for each facility , we can get 8 sequences ({ }、{ }、{ }、{ }、{ }、{ }、{ }、{ Then, we add start and stop markers to each sequence to indicate the beginning and end, and finally pad the sequences to the maximum length of the training set.

[0120] Step 1033: Determine a plurality of target facility three-dimensional models from the plurality of facility three-dimensional models to be screened based on the category sequence.

[0121] Step 1034, determining the placement position of each target facility three-dimensional model according to the position sequence, determining the placement direction of each target facility three-dimensional model according to the direction sequence, and determining the size of each target facility three-dimensional model according to the size sequence.

[0122] In the specific implementation of the above steps 1033 to 1034, multiple target facility 3D models are determined from multiple facility 3D models to be screened according to the category sequence. The placement position of each target facility 3D model is determined according to the position sequence, the placement direction of each target facility 3D model is determined according to the direction sequence, and the size of each target facility 3D model is determined according to the size sequence, so as to determine the spatial attributes of each target facility 3D model.

[0123] S104, placing each target facility model in a model area corresponding to the scene in the spatial layout three-dimensional model based on the spatial attributes of each target facility three-dimensional model, so as to generate a spatial design three-dimensional model that meets the interior design requirements.

[0124] For the above step S104, in the specific implementation, after the spatial attributes of each target facility three-dimensional model, that is, the placement position, placement direction and size of each target facility three-dimensional model in the scene are determined, each target facility model is placed in the model area corresponding to the scene in the spatial layout three-dimensional model using the spatial attributes of each target facility three-dimensional model to generate a spatial design three-dimensional model that meets the interior design requirements.

[0125] As an optional embodiment, after the three-dimensional model of the spatial design is generated, the interpretability of the indoor generative design can also be achieved. The interpretability can enable users to better understand the design steps of the generative design, and can also facilitate the subsequent optimization of the architectural generative design. The present application constructs an interpretability framework based on rule reasoning and semantic analysis through a facility knowledge graph for indoor scenes. The framework provides explanations for each decision step and the final design solution of the generative design through the facilities described in the knowledge graph and the relationships between the facilities. As an optional embodiment, after the three-dimensional model of the spatial design that conforms to the text description is generated, the generation method also includes:

[0126] (1) Encoding each facility node in the spatial design three-dimensional model based on the facility knowledge graph and adding it to the neural network graph to obtain a scene generation design neural network graph for the target building.

[0127] For the above step (1), in the specific implementation, each facility node in the spatial design 3D model is encoded based on the facility knowledge graph and added to the neural network graph to obtain the scene generation design neural network graph of the target building. Here, specifically, a graph neural network (GNN) is used to perform reasoning by calculating the attention weights between nodes, and then a probability score is given as output. The most likely answer is then selected. The focus here is on the attention weights from the GNN, which are crucial to understanding the reasoning process of the model. The best-first search is used to track the attention weights. Nodes that are more relevant to the scene have greater weights, represented by thicker connecting lines. By following these lines, users can accurately understand how the facility model and placement are inferred, and whether its reasoning is reasonable.

[0128] Here, let the design background Represented as a collection of nodes N= , ,…, , where each node The relationship between these nodes is represented by edges. Indicates that Representation Node and Given the design context, LLM maps the input to the embedding space, expressed as ,in is the embedding dimension. Then, GNN processes the embedding through hierarchical transformation. The following formula represents the embedding update process of the facility node in the graph neural network:

[0129]

[0130] in, It is Facility nodes in the layer The updated embed, Represents a facility node The set of neighbor nodes, that is, the nodes in the layout that are connected to the facility Adjacent or directly related facilities, It is The weight matrix of the layer is used to map the embedding of the neighbor facility nodes to the representation of the current node, Represents neighbor facility nodes In the The embedded information of the layer describes the properties of the facility (type, size, location, orientation), is the bias term, It is an activation function, usually a nonlinear function, used to introduce complex feature relationships.

[0131] In the process of layout generation, the above formula will gradually aggregate the information of its neighboring facilities by updating the embedding layer by layer according to the proximity or spatial relationship between facilities. This can help each facility node form an understanding of its surrounding environment and generate an embedding representation that meets the spatial aesthetic and functional requirements.

[0132] In the above formula, the interpretability of neighbor information is aggregated layer by layer. The embedding of each facility node is updated layer by layer based on the information of its neighbor facilities. This mechanism enables the model's embedding representation to include the relationship between facilities in the space. This design helps explain how the embedding of each facility node is affected by neighbor facilities. For example, if the final embedding value of a sofa node is high, the model can trace back the embedding and weight matrix of its neighbor nodes. The specific value of is used to explain how the node is affected by the surrounding facilities (such as coffee tables, TV cabinets, etc.), thereby forming a layout suitable for the current space. By analyzing the neighbor contribution of each layer, it is possible to observe how each facility integrates the information of neighboring facilities on each layer, which to a certain extent reveals how the above steps "understand" the relationship between spatial structure and facilities when generating a three-dimensional model of spatial design. In this way, the generative design process is structured to provide a clear explanation for the choices made during the design process, thereby enhancing the interpretability and comprehensibility of the results.

[0133] (2) Input each facility node into a multi-layer perceptron to obtain a design scheme score for the spatial design three-dimensional model.

[0134] For the above step (2), in the specific implementation, each facility node is input into the multi-layer perceptron MLP to obtain the design scheme score of the spatial design three-dimensional model Specifically, the following formula represents the final output of the graph neural network, which is used to generate the design scheme score of the facility layout: :

[0135]

[0136] in, It is the output calculated by the Multi-Layer Perceptron (MLP) and represents the score of the facility layout. L Represents the final layer of the GNN. Represents the facility node embedding of the last layer of the GNN. The final embeddings of all facility nodes are concatenated into a vector, which contains the global information of the entire layout. MLP is a multi-layer perceptron that further processes these concatenated embeddings to generate a final score or indicator for evaluating the rationality of the layout.

[0137] Final Output It can be used to measure the quality of the entire facility layout. In the above formula, the interpretability of the global layout score is that the above formula concatenates the final embeddings of all facility nodes and generates an overall score through MLP. This score can be interpreted as a quality measure of the layout, such as aesthetics, space utilization, or comfort. Since the layer-by-layer update of the above formula makes the embedding of each facility contain the spatial information of its neighbors, this overall score can represent the rationality of the entire layout.

[0138] The method for generating a three-dimensional model of architectural space design provided in an embodiment of the present application comprises the following steps: first, obtaining a plan structure diagram of a target building and text description information of interior design requirements for the target building, and generating a three-dimensional model of the spatial layout of the target building based on the plan structure diagram and the text description information; then, for each scene in the three-dimensional spatial layout model, screening out a plurality of three-dimensional facility models to be screened that match the scene style of the scene from a pre-built facility model library based on the text description information; generating an object sequence corresponding to the scene using the plurality of three-dimensional facility models to be screened, and determining a plurality of target facility three-dimensional models and spatial attributes of each target facility three-dimensional model from the plurality of three-dimensional facility models to be screened based on the object sequence; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene; finally, placing each target facility model in the model area corresponding to the scene in the three-dimensional spatial layout model based on the spatial attributes of each target facility three-dimensional model, so as to generate a three-dimensional spatial design model that meets the interior design requirements.

[0139] This application realizes the generative design of facilities in the overall indoor scene by integrating the facility knowledge graph and the facility generation model. According to the room structure and the user's text requirements, the problem of scene style customization and the selection and layout coordination of facilities in the scene is solved. Input the scene structure plan and the user's text description, select the scene infrastructure required by the user from the facility model library, and select the facility three-dimensional model that meets the scene style, and then use multiple facility three-dimensional models to generate the object sequence corresponding to the scene to achieve the final facility layout generation. Finally, the interpretability of the scene generation design can be completed through the path analysis of the scene facility knowledge graph, so that users can better understand the design steps of the generative design, and facilitate the subsequent optimization of the architectural generative design. The reconstructed interior layout is organically combined with the generated facility model to complete the overall space design for a specific scene, and generate an overall space solution that meets the user's functional requirements and design style requirements.

[0140] See also Figure 2 , Figure 3 , Figure 2 This is one of the structural schematic diagrams of a device for generating a three-dimensional model of a building space design provided in an embodiment of the present application. Figure 3 This is a second structural diagram of a device for generating a three-dimensional model of a building space design provided in an embodiment of the present application. Figure 2 As shown in , the generating device 200 includes:

[0141] The spatial layout three-dimensional model generation module 201 is used to obtain a plane structure diagram of a target building and text description information for interior design requirements of the target building, and generate a spatial layout three-dimensional model of the target building based on the plane structure diagram and the text description information;

[0142] The to-be-screened facility three-dimensional model determination module 202 is used to screen out a plurality of to-be-screened facility three-dimensional models that match the scene style of the scene from a pre-built facility model library based on the text description information for each scene in the spatial layout three-dimensional model;

[0143] The target facility three-dimensional model determination module 203 is used to generate an object sequence corresponding to the scene using the plurality of the three-dimensional models of the facilities to be screened, and determine a plurality of target facility three-dimensional models from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene;

[0144] The space design 3D model generation module 204 is used to place each target facility model in the model area corresponding to the scene in the space layout 3D model based on the spatial attributes of each target facility 3D model, so as to generate a space design 3D model that meets the interior design requirements.

[0145] Furthermore, when the spatial layout three-dimensional model generating module 201 is used to generate the spatial layout three-dimensional model of the target building based on the planar structure diagram and the text description information, the spatial layout three-dimensional model generating module 201 is also used to:

[0146] Generate a layout fusion feature vector based on the plan layout features of the plan structure diagram and the text layout features of the text description information;

[0147] Inputting the layout fusion feature vector into a pre-trained spatial layout generation model to obtain the spatial layout of the target building, and generating a scene function matrix based on the spatial layout, the layout fusion feature vector and the text layout feature;

[0148] The scene function matrix is ​​input into a pre-trained three-dimensional generative adversarial network to generate a three-dimensional model of the spatial layout of the target building.

[0149] See also Figure 3 The generating device 200 further includes a facility model library building module 205. Before obtaining the plan structure diagram of the target building and the text description information of the interior design requirements of the target building, the facility model library building module 205 is used to:

[0150] For each facility type, images of the original model corresponding to the facility type under multiple viewing angles and information text data corresponding to the facility type are obtained;

[0151] Extracting visual features from the plurality of images to obtain a visual style feature vector, and extracting text features from the information text data to obtain a text feature vector;

[0152] Inputting the visual style feature vector and the text feature vector into a pre-trained cross-modal feature fusion model for cross-modal feature fusion to obtain a cross-modal feature vector;

[0153] Inputting the cross-modal feature vector into a pre-trained diffusion model to obtain an initial three-dimensional model corresponding to the facility type;

[0154] The initial three-dimensional model and the cross-modal feature vector are input into a pre-trained texture generative adversarial network to obtain a basic three-dimensional model of the facility type to generate the facility model library; wherein the surface of the basic three-dimensional model is provided with a facility texture map that conforms to the information text data.

[0155] Furthermore, when the to-be-screened facility 3D model determination module 202 is used to screen out a plurality of to-be-screened facility 3D models that match the scene style of the scene from a pre-built facility model library based on the text description information, the to-be-screened facility 3D model determination module 202 is also used to:

[0156] Determine a scene facility model corresponding to the text description information from a plurality of basic three-dimensional models in the facility model library;

[0157] Determine the first component type of each scene facility model and the second component type of each basic three-dimensional model in the facility model library, determine the style influence weight between the scene facility model and the basic three-dimensional model from a pre-built facility knowledge graph, and query the first facility style vector of each scene facility model in the facility style vector library;

[0158] Calculate a scene style vector of the scene based on the first component type, the second component type, the style influence weight, and the first facility style vector;

[0159] Querying the second facility style vector of each basic three-dimensional model in the facility style vector library, and calculating the style similarity between the scene style vector and the second facility style vector of each basic three-dimensional model;

[0160] The multiple basic three-dimensional models are sorted using the style similarity corresponding to each basic three-dimensional model, and multiple three-dimensional facility models to be screened are determined from the sorting.

[0161] See also Figure 3 The generating device 200 further includes a facility knowledge graph construction module 206, and the facility knowledge graph construction module 206 is used to construct the facility knowledge graph through the following steps:

[0162] Obtain multiple scene sample images, and use the target recognition algorithm to identify the facility matching combinations between multiple facility types in each scene sample image;

[0163] The facility collocation combinations of multiple facility types are analyzed through an association rule mining algorithm to determine the correlation between the multiple facility types to generate the facility knowledge graph.

[0164] Further, the object sequence includes a category sequence, a position sequence, a direction sequence and a size sequence; when the target facility three-dimensional model determination module 203 is used to generate an object sequence corresponding to the scene using the plurality of the three-dimensional models of the facilities to be screened, and to determine a plurality of target facility three-dimensional models from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, and the spatial attributes of each target facility three-dimensional model, the target facility three-dimensional model determination module 203 is also used to:

[0165] Inputting each three-dimensional model of the facility to be screened into the category model to determine the category sequence, and inputting the category sequence into the direction model to determine the direction sequence;

[0166] Inputting the category sequence and the position sequence into a position model to determine the position sequence, and inputting the category sequence, the direction sequence, and the position sequence into a size model to determine the size sequence;

[0167] Determine a plurality of target facility three-dimensional models from the plurality of facility three-dimensional models to be screened based on the category sequence;

[0168] The placement position of each target facility three-dimensional model is determined according to the position sequence, the placement direction of each target facility three-dimensional model is determined according to the direction sequence, and the size of each target facility three-dimensional model is determined according to the size sequence.

[0169] See also Figure 3 The generating device 200 further includes a scoring determination module 207. After generating the three-dimensional spatial design model that conforms to the text description, the scoring determination module 207 is used to:

[0170] Based on the facility knowledge graph, each facility node in the spatial design three-dimensional model is encoded and added to the neural network graph to obtain a scene generation design neural network graph of the target building;

[0171] Each facility node is input into a multi-layer perceptron to obtain a design scheme score of the spatial design three-dimensional model.

[0172] See also Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 4 As shown in , the electronic device 400 includes a processor 410 , a memory 420 and a bus 430 .

[0173] The memory 420 stores machine-readable instructions executable by the processor 410. When the electronic device 400 is running, the processor 410 communicates with the memory 420 via the bus 430. When the machine-readable instructions are executed by the processor 410, the above-mentioned Figure 1 The steps of the method for generating the three-dimensional model of the architectural space design in the method embodiment shown, the specific implementation method can be found in the method embodiment, and will not be repeated here.

[0174] The present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the computer program can execute the above-mentioned Figure 1 The steps of the method for generating the three-dimensional model of the architectural space design in the method embodiment shown, the specific implementation method can be found in the method embodiment, and will not be repeated here.

[0175] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0176] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0177] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0178] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0179] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0180] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for generating a three-dimensional model of architectural space design, characterized in that: The generation method comprises: Acquire a plan structure diagram of a target building and text description information for interior design requirements of the target building, and generate a three-dimensional model of the spatial layout of the target building based on the plan structure diagram and the text description information; For each scene in the spatial layout three-dimensional model, a plurality of to-be-screened three-dimensional facility models matching the scene style are screened out from a pre-built facility model library based on the text description information; Generate an object sequence corresponding to the scene using the plurality of the three-dimensional models of the facilities to be screened, and determine a plurality of target facility three-dimensional models from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene; Placing each target facility model in a model area corresponding to the scene in the spatial layout three-dimensional model based on the spatial attributes of each target facility three-dimensional model, so as to generate a spatial design three-dimensional model that meets the interior design requirements; The generating of the spatial layout three-dimensional model of the target building based on the plane structure diagram and the text description information comprises: Generate a layout fusion feature vector based on the plan layout features of the plan structure diagram and the text layout features of the text description information; Inputting the layout fusion feature vector into a pre-trained spatial layout generation model to obtain the spatial layout of the target building, and generating a scene function matrix based on the spatial layout, the layout fusion feature vector and the text layout feature; The scene function matrix is ​​input into a pre-trained three-dimensional generative adversarial network to generate a three-dimensional model of the spatial layout of the target building.

2. The generation method according to claim 1, characterized in that: Before obtaining the plan structure diagram of the target building and the text description information for the interior design requirements of the target building, the generation method further includes: For each facility type, images of the original model corresponding to the facility type under multiple viewing angles and information text data corresponding to the facility type are obtained; Extracting visual features from the plurality of images to obtain a visual style feature vector, and extracting text features from the information text data to obtain a text feature vector; Inputting the visual style feature vector and the text feature vector into a pre-trained cross-modal feature fusion model for cross-modal feature fusion to obtain a cross-modal feature vector; Inputting the cross-modal feature vector into a pre-trained diffusion model to obtain an initial three-dimensional model corresponding to the facility type; The initial three-dimensional model and the cross-modal feature vector are input into a pre-trained texture generative adversarial network to obtain a basic three-dimensional model of the facility type to generate the facility model library; wherein the surface of the basic three-dimensional model is provided with a facility texture map that conforms to the information text data.

3. The generation method according to claim 2, characterized in that: The step of selecting a plurality of to-be-selected facility three-dimensional models that match the scene style of the scene from a pre-built facility model library based on the text description information includes: Determine a scene facility model corresponding to the text description information from a plurality of basic three-dimensional models in the facility model library; Determine the first component type of each scene facility model and the second component type of each basic three-dimensional model in the facility model library, determine the style influence weight between the scene facility model and the basic three-dimensional model from a pre-built facility knowledge graph, and query the first facility style vector of each scene facility model in the facility style vector library; Calculate a scene style vector of the scene based on the first component type, the second component type, the style influence weight, and the first facility style vector; Querying the second facility style vector of each basic three-dimensional model in the facility style vector library, and calculating the style similarity between the scene style vector and the second facility style vector of each basic three-dimensional model; The multiple basic three-dimensional models are sorted using the style similarity corresponding to each basic three-dimensional model, and multiple three-dimensional facility models to be screened are determined from the sorting.

4. The generation method according to claim 3, characterized in that: The facility knowledge graph is constructed through the following steps: Obtain multiple scene sample images, and use the target recognition algorithm to identify the facility matching combinations between multiple facility types in each scene sample image; The facility collocation combinations of multiple facility types are analyzed through an association rule mining algorithm to determine the correlation between the multiple facility types to generate the facility knowledge graph.

5. The generation method according to claim 1, characterized in that: The object sequence includes a category sequence, a position sequence, a direction sequence and a size sequence; the object sequence corresponding to the scene is generated by using the plurality of the three-dimensional models of the facilities to be screened, and a plurality of target facility three-dimensional models are determined from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model, including: Inputting each three-dimensional model of the facility to be screened into the category model to determine the category sequence, and inputting the category sequence into the direction model to determine the direction sequence; Inputting the category sequence and the position sequence into a position model to determine the position sequence, and inputting the category sequence, the direction sequence, and the position sequence into a size model to determine the size sequence; Determine a plurality of target facility three-dimensional models from the plurality of facility three-dimensional models to be screened based on the category sequence; The placement position of each target facility three-dimensional model is determined according to the position sequence, the placement direction of each target facility three-dimensional model is determined according to the direction sequence, and the size of each target facility three-dimensional model is determined according to the size sequence.

6. The generation method according to claim 3, characterized in that: After generating the spatial design three-dimensional model that conforms to the text description, the generating method further includes: Based on the facility knowledge graph, each facility node in the spatial design three-dimensional model is encoded and added to the neural network graph to obtain a scene generation design neural network graph of the target building; Each facility node is input into a multi-layer perceptron to obtain a design scheme score of the spatial design three-dimensional model.

7. A device for generating a three-dimensional model of architectural space design, characterized in that: The generating device comprises: A spatial layout three-dimensional model generation module, used to obtain a plane structure diagram of a target building and text description information for interior design requirements of the target building, and generate a spatial layout three-dimensional model of the target building based on the plane structure diagram and the text description information; A module for determining the three-dimensional model of facilities to be screened is used to screen out a plurality of three-dimensional models of facilities to be screened that match the scene style of the scene from a pre-built facility model library based on the text description information for each scene in the three-dimensional model of the spatial layout; A target facility three-dimensional model determination module is used to generate an object sequence corresponding to the scene using the plurality of the three-dimensional models of the facilities to be screened, and determine a plurality of target facility three-dimensional models from the plurality of the three-dimensional models of the facilities to be screened based on the object sequence, as well as the spatial attributes of each target facility three-dimensional model; wherein the spatial attributes include the placement position, placement direction and size of the target facility three-dimensional model in the scene; A spatial design three-dimensional model generation module is used to place each target facility model in a model area corresponding to the scene in the spatial layout three-dimensional model based on the spatial attributes of each target facility three-dimensional model, so as to generate a spatial design three-dimensional model that meets the interior design requirements; When the spatial layout three-dimensional model generation module is used to generate the spatial layout three-dimensional model of the target building based on the plane structure diagram and the text description information, the spatial layout three-dimensional model generation module is also used to: Generate a layout fusion feature vector based on the plan layout features of the plan structure diagram and the text layout features of the text description information; Inputting the layout fusion feature vector into a pre-trained spatial layout generation model to obtain the spatial layout of the target building, and generating a scene function matrix based on the spatial layout, the layout fusion feature vector and the text layout feature; The scene function matrix is ​​input into a pre-trained three-dimensional generative adversarial network to generate a three-dimensional model of the spatial layout of the target building.

8. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are run by the processor, the steps of the method for generating a three-dimensional model of architectural space design as described in any one of claims 1 to 6 are executed.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for generating a three-dimensional model of architectural space design as described in any one of claims 1 to 6 are executed.

Images