Design image generation method and device, electronic equipment and storage medium

Abstract

This disclosure relates to a design image generation method, apparatus, electronic device, and storage medium. The method involves acquiring textual design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements for a target building, fusing the aforementioned multimodal information to obtain a fused design intent for the target building; parametrically translating the fused design intent to generate structured design parameters for the target building; extracting hierarchical features from the structured design parameters to generate hierarchical design parameters for the target building; and generating a target design image of the target building based on the hierarchical design parameters and at least one preset design style. Thus, by fusing, parametrically translating, and extracting multi-level controllable parameters from the multimodal design requirements of the target building, a design image with a corresponding design style is intelligently generated. This not only achieves the inheritance and innovation of design images but also improves the intelligence and efficiency of the design image generation method.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a design of an image generation method, apparatus, electronic device, and storage medium. Background Technology

[0002] Traditional architecture contains rich wisdom in construction and aesthetic value. How to inherit and innovate the characteristics of traditional architecture in contemporary architectural design is a key issue of concern in the architectural community.

[0003] In related technologies, design images of the buildings to be constructed are generally generated based on parametric traditional architectural design methods or case-based design reference methods. However, this approach relies on designers possessing traditional architectural knowledge and manually adjusting or extracting parameters to generate the design images. Therefore, current design image generation methods have low levels of intelligence and are inefficient. Summary of the Invention

[0004] To address the aforementioned technical problems, this disclosure provides a design image generation method, apparatus, electronic device, and storage medium.

[0005] In a first aspect, this disclosure provides a method for designing an image generation method, including: Obtain design requirement information for the target building, wherein the design requirement information includes text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn representation of building elements, and three-dimensional model design requirements based on three-dimensional model representation of building elements. The text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements are integrated to obtain the integrated design intent of the target building; The integrated design intent is parametrically translated to generate the structural design parameters of the target building; Hierarchical feature extraction is performed on the structured design parameters to generate hierarchical design parameters for the target building; Based on the hierarchical design parameters and at least one preset design style, a target design image of the target building is generated.

[0006] Secondly, this disclosure provides a design image generation apparatus, comprising: The first acquisition module is used to acquire the design requirement information of the target building, wherein the design requirement information includes text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn drawings, and three-dimensional model design requirements based on three-dimensional model representation of building elements. The information fusion module is used to fuse the text design requirements, the image design requirements, the hand-drawn design requirements, and the three-dimensional model design requirements to obtain the fused design intent of the target building. The first generation module is used to parametrically translate the fused design intent and generate the structural design parameters of the target building. The second generation module is used to extract hierarchical features from the structured design parameters to generate hierarchical design parameters for the target building. The third generation module is used to generate a target design image of the target building based on the hierarchical design parameters and at least one preset design style.

[0007] Thirdly, embodiments of this disclosure also provide an electronic device, the electronic device comprising: One or more processors; Storage device for storing one or more programs. When one or more programs are executed by one or more processors, the one or more processors implement the methods provided in the first aspect.

[0008] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method provided in the first aspect.

[0009] The technical solution provided in this disclosure has the following advantages compared with the prior art: This disclosure discloses a design image generation method, apparatus, electronic device, and storage medium. The method acquires design requirement information for a target building, including textual design requirements representing building elements, image-based design requirements, hand-drawn design requirements, and 3D model design requirements. The method fuses these requirements to obtain a fused design intent for the target building. This fused design intent is then parametrically translated to generate structured design parameters for the target building. Hierarchical feature extraction is performed on these structured design parameters to generate hierarchical design parameters for the target building. Based on these hierarchical design parameters and at least one preset design style, a target design image for the target building is generated. Thus, by fusing and parametrically translating the multimodal design requirements of the target building, transparent and controllable structured design parameters are generated. Multi-level controllable parameter extraction is then performed on these structured design parameters, and a design image conforming to the corresponding design style is intelligently generated based on the obtained hierarchical design parameters. This achieves both the inheritance and innovation of design images while improving the intelligence and efficiency of the design image generation method. Attached Figure Description

[0010] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0011] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic flowchart illustrating a design image generation method provided in this embodiment of the disclosure; Figure 2 This is a schematic diagram of the structure of an image generation device provided in an embodiment of the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0013] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0014] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0015] Related technologies also employ design graphics generation methods based on generative adversarial networks (GANs) to generate design images of buildings. This method utilizes GANs to achieve architectural style transfer, converting modern architectural images into images with traditional styles. Other related technologies employ design graphics methods based on diffusion models to generate design images of buildings. These methods guide the generation process by describing architectural styles and features through text prompts and utilize diffusion models to generate design images of buildings. However, the above methods have the following drawbacks: Disadvantage 1: Insufficient depth of architectural semantic understanding. The above methods mainly rely on style transfer at the visual level and lack an understanding of the deep semantics of traditional architecture. For example, they cannot understand the construction logic of "hip roof", the mechanical principles of "dougong" (bracket set) and the cultural connotations of "caisson ceiling", which may lead to the problem that the generated results are similar in form but not in spirit, or even produce erroneous forms that violate the rules of traditional architectural construction. Disadvantage 2: Lack of multimodal information fusion capability. Typically, building design requirements are represented in text, image, hand-drawn, or 3D model form. These methods can only handle single-modal input and cannot achieve a unified understanding and fusion of multimodal design intent, thus limiting the flexibility of design expression.

[0016] Disadvantage 3: Lack of parameter translation mechanism. Most of the methods mentioned above are end-to-end black-box generation, directly mapping from input conditions to output images, lacking an intermediate design parameter translation process. This makes it impossible for users to understand and intervene in the specific design decisions during the generation process, and it is also difficult to make fine-grained parameter adjustments to the generated results, which does not conform to the professional workflow of architectural design.

[0017] Disadvantage 4: Insufficient controllability and consistency in generation. The above methods face the problem of poor controllability in architectural design applications, making it difficult for users to accurately control the specific features of the generated building. There is also a lack of consistency between the results generated multiple times, making it difficult to form a systematic design image.

[0018] Disadvantage 5: Insufficient consideration of building codes and constraints. The design images of buildings need to meet various constraints such as functional requirements, structural safety, and building codes. The above methods mainly focus on visual effects and lack modeling of architectural constraints. The generated design images may have problems such as unreasonable functions, infeasible structures, and violations of codes, making them difficult to use directly in actual design.

[0019] Disadvantage 6: The strategy for integrating tradition and modernity is too simplistic. When dealing with the integration of traditional architectural features and modern functional requirements, the above methods usually adopt a simple style superposition strategy, lacking in-depth thinking and systematic solutions to the core issue of "how to meet modern functions while maintaining traditional charm".

[0020] To address the aforementioned problems, this embodiment provides a method for designing image generation. The following is a detailed explanation... Figure 1 The design image generation method provided in this disclosure is described below. In this disclosure, the design image generation method can be executed by an electronic device or a server. The electronic device may include a desktop computer, laptop, tablet, and other smart devices. The server may include a cloud server or a server cluster. This embodiment specifically uses an electronic device as an example to generate design images.

[0021] Figure 1 A schematic flowchart of a design image generation method provided in an embodiment of this disclosure is shown.

[0022] As shown in Figure 1, the design image generation method may include the following steps.

[0023] S110. Obtain the design requirements information of the target building, including text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn sketches, and three-dimensional model design requirements based on three-dimensional model representation of building elements.

[0024] The target building refers to a building with architectural requirements for style design. Design requirement information can be building information model elements.

[0025] To address the limitations of unimodal input, this embodiment acquires multimodal requirement information as the design requirement information for the target building. Specifically, the design requirement information includes requirements for building elements represented in text, image, hand-drawn, and 3D model formats.

[0026] In some embodiments, the design requirement information includes textual design requirements; correspondingly, obtaining the design requirement information of the target building includes: obtaining the textual requirement information of the target building; using a preset text understanding model to perform semantic understanding on the textual requirement information to obtain initial semantic information; performing specialized term expansion and semantic role labeling on the initial semantic information to obtain semantic annotation information; and performing intent vector encoding on the semantic annotation information to obtain the textual design requirements.

[0027] Among them, textual demand information refers to user demand content represented in text form.

[0028] Optionally, the pre-defined text understanding model can be a large language model finely tuned from a traditional architectural corpus.

[0029] In some embodiments, the design requirement information includes image design requirements; correspondingly, obtaining the design requirement information of the target building includes: obtaining the image requirement information of the target building; and using a preset visual analysis model to extract and analyze the style features of the image requirement information to obtain the image design requirements.

[0030] Among them, image demand information refers to user demand content represented in the form of images.

[0031] Specifically, when a user represents design requirements in the form of images, the electronic device uses the Vision Transformer (ViT) as a preset visual analysis model to extract multi-granular features from the image requirement information of the target building to obtain global style features and local component features. Then, it performs style feature parsing to automatically parse the style attributes of traditional buildings as image design requirements.

[0032] In some embodiments, the design requirement information includes hand-drawn design requirements; correspondingly, obtaining the design requirement information of the target building includes: obtaining the hand-drawn requirement information of the target building; standardizing the hand-drawn requirement information to obtain standardized hand-drawn requirement information; using a preset hand-drawn understanding model, identifying the design intent from the standardized hand-drawn requirement information, and mapping the architectural requirements in the design intent to the hand-drawn design requirements.

[0033] Among them, hand-drawn requirement information refers to user requirements represented by hand-drawn drawings (i.e., sketches).

[0034] Optionally, the normalization process specifically includes stroke smoothing and noise reduction, line type recognition (e.g., outline lines, annotation lines, auxiliary lines), and separation of graphic elements (e.g., building bodies, environmental elements, text annotations).

[0035] In some embodiments, the design requirements information includes 3D model design requirements; correspondingly, obtaining the design requirements information of the target building includes: obtaining the 3D model requirements information of the target building; performing site understanding and environmental analysis on the 3D model requirements to generate 3D model design requirements.

[0036] Among them, 3D model requirement information refers to user requirements represented in the form of a 3D model.

[0037] This allows for flexible selection of the input method for design requirements information based on the design stage and expression habits, or the use of multiple modalities to describe design requirements information together, improving the ease of use of the design image generation process and the freedom of design expression, and overcoming the limitations of single-modal input in existing methods.

[0038] S120. The text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements are integrated to obtain the integrated design intent of the target building.

[0039] To enhance the understanding of the deep semantics of traditional architecture, electronic devices can integrate multimodal input design requirements into a unified design intent representation, serving as the integrated design intent of the target building.

[0040] In some embodiments, the specific implementation method of S120 includes: semantically aligning the text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements describing the same object in the target building; performing semantic understanding, structural understanding, and decorative understanding on the semantically aligned text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements to generate semantic design intent, structural design intent, and decorative design intent; and performing fusion processing on the semantic design intent, structural design intent, and decorative design intent after information completion to generate a fused design intent.

[0041] Specifically, firstly, the electronic device uses a language-image-based contrastive learning method (CLI) to map textual design requirements and image design requirements describing the same object to similar representation spaces, aligning them. It also learns the correspondence between hand-drawn design requirements and 3D model design requirements, enabling effective guidance in the generation of hand-drawn images. Next, a hierarchical intent fusion mechanism fuses the aligned four design requirements at the semantic, structural, and detail levels, generating semantic design intent, structural design intent, and decorative design intent. Finally, after information completion for these three design intents, a fusion process is performed, including missing modality detection, parameter default value filling, and consistency verification, to generate a fused design intent.

[0042] Optionally, semantic layer fusion includes integrating high-level semantic understandings from various modalities regarding building type, style, and function. Structural layer fusion includes integrating structural understandings from various modalities regarding spatial layout and form organization. Detail layer fusion includes integrating detailed preferences from various modalities regarding components, decorations, materials, etc.

[0043] Optionally, the missing modality judgment specifically involves inferring possible visual features based on the text content if only a text description is provided. Default value parameter filling specifically involves filling default values ​​for unspecified design parameters based on typical characteristics of the building type and style. Consistency verification specifically involves checking for contradictions between multimodal inputs; if contradictions exist, the user is prompted for clarification or the inputs are processed according to priority rules.

[0044] In this way, by integrating multimodal design requirements information, it is possible to understand the deep semantics of architecture. For example, it is possible to truly understand professional concepts such as the horse-head wall features of "Hui-style", the hierarchical regulations of "official architecture", and the structural significance of "dougong protrusion", rather than just imitating the superficial visual style. This is conducive to generating more accurate and professional design graphics in an architectural sense, and avoids the problem of being similar in form but not in spirit.

[0045] S130. Parametrically translate the integrated design intent to generate structural design parameters for the target building.

[0046] To transform vague design intentions into structured design parameters, making the generation process of design graphics transparent and controllable, electronic devices can translate fused design intentions into structured design parameters according to pre-designed overall layout parameters.

[0047] Among them, structured design parameters refer to design parameters that are transparent and controllable in design software.

[0048] Optionally, the overall layout parameters include a four-layer structure.

[0049] First layer: Overall layout parameters. Courtyard structure parameters: number of courtyards, number of transverse courtyards, courtyard shape (square / rectangular / irregular); Axis parameters: main axis direction, number of axes, axis spacing; Building group organization: location of main buildings, configuration of auxiliary buildings, corridor connection method.

[0050] Second layer: Individual building parameters. Plan parameters: width (number of bays and dimensions of each bay), depth (number of spans and dimensions), plan shape; Section parameters: eaves column height, ridge height, platform height, floor height (for multi-story buildings); Functional parameters: building function type, spatial composition, circulation organization.

[0051] Third layer: Component configuration parameters. Roof parameters: Roof type (hipped roof, gable roof, overhanging roof, hard gable roof, pyramidal roof, etc.), roof slope, eaves depth, upturned curve; Dougong parameters: Whether to set dougong, number of paving layers, dougong type, number of brackets; Decoration parameters: Door and window type, railing style, platform base configuration.

[0052] Fourth layer: Decorative details parameters. Painted decoration parameters: painting grade (Hexi, Xuanzi, Suzhou style), theme pattern, color scheme; Carving parameters: carving location, carving subject, carving technique; Material parameters: main material, roofing material, decorative material, color tone.

[0053] In some embodiments, the specific implementation method of S130 includes: compressing and encoding the fusion design based on the intent encoding layer in the parametric translation network to obtain intent encoding parameters; predicting discrete parameters, continuous parameters, and structural parameters from the intent encoding parameters based on the parameter prediction layer in the parametric translation network; and applying proportional constraints, hierarchical constraints, and construction constraints to the discrete parameters, continuous parameters, and structural parameters based on the parameter consistency layer in the parametric translation network to obtain structured design parameters.

[0054] Optionally, for discrete parameters (such as roof shape), a classification prediction head can be used as the output probability distribution of the parametric prediction layer. For continuous parameters (such as size scale), a regression prediction head can be used as the output numerical value of the parametric prediction layer. For structural parameters (such as spatial layout), a sequence / graph generation head can be used as the output topology structure of the parametric prediction layer.

[0055] Optionally, the proportional constraints specifically define the reasonable range of the ratio between bay width and depth. The hierarchical constraints specifically define the correspondence between building hierarchy and component configuration. The structural constraints specifically define the matching rules between different components.

[0056] In this way, vague design intentions can be transformed into structured design parameters, so that the generated results have clear parameter correspondences, solving the problem of the "black box" uncontrollability of existing deep generation methods.

[0057] Furthermore, after executing S130, the electronic device can also display the structured design parameters to the user in an intuitive panel format, generate parameter sensitivity prompts and parameter preset templates.

[0058] Optionally, the panel format may include displaying parameter hierarchy in a tree structure and using controls such as sliders / drop-down boxes to enable interactive adjustments, thereby achieving the effect of real-time preview of parameter adjustments.

[0059] Optionally, parameter sensitivity hints are used to highlight key parameters that have a significant impact on the final result, guiding users to pay attention to them.

[0060] Optionally, the parameter preset templates are based on different styles, eras, and regions, and users can adjust them accordingly.

[0061] S140. Extract hierarchical features from the structured design parameters to generate hierarchical design parameters for the target building.

[0062] To improve the recognition accuracy of features at different levels, electronic devices can adopt a hierarchical extraction strategy from overall layout to details to obtain hierarchical design parameters of the target building.

[0063] In some embodiments, the specific implementation method of S140 includes: S1401, extracting layout design parameters from structured design parameters based on a preset layout generation network; S1402, extracting morphological design parameters from layout design parameters based on a preset morphological generation layer; S1403, extracting component design parameters from morphological design parameters based on a preset component generation network; S1404, extracting preference design parameters from component design parameters based on a preset decoration generation network; S1405, generating hierarchical design parameters by performing context passing and global coordination processing on layout design parameters, morphological design parameters, component design parameters, and preference design parameters based on a preset consistency control network.

[0064] The specific implementation method of S1401 includes: performing site analysis on the structured design parameters based on the site analysis layer in the preset layout generation network to obtain the buildable range and orientation information; and performing layout planning and layout adjustment on the buildable range and orientation information based on the layout map generation layer in the preset layout generation network to generate layout design parameters.

[0065] Optionally, the layout planning can specifically use a conditional diffusion model to generate a floor plan. Layout adjustments can specifically optimize the functional flow lines and spatial sequences of the generated layout.

[0066] The specific implementation method of S1402 includes: based on the mass model generation layer in the preset morphology generation layer, performing plan layout and section parameter analysis on the layout design parameters to generate a building mass model; based on the roof morphology generation layer in the preset morphology generation layer, performing roof morphology analysis on the layout design parameters to generate a three-dimensional roof morphology; based on the morphology variant generation layer in the preset morphology generation layer, performing morphology variant analysis on the layout design parameters to generate morphology variant parameters; and using the building mass model, the three-dimensional roof morphology, and the morphology variant parameters as morphology design parameters.

[0067] The specific implementation method of S1403 includes: based on the instantiation layer in the preset component generation network, instantiating the components found from the preset component library according to the parameters to obtain the instantiated component parameters; and supplementing the instantiated component parameters to the component supplementation layer in the preset component generation network to obtain the component design parameters.

[0068] The preset component library includes: a bracket component library (e.g., parametric models of brackets from different eras, levels, and regions), a door and window component library (e.g., various types such as partition windows, sill windows, and hinged windows), a railing component library (e.g., railings with walking sticks, flower railings, etc.), and a decorative component library (e.g., decorative components such as brackets, hanging flowers, and pendants).

[0069] Specifically, firstly, matching components are retrieved from a pre-built parametric traditional building component library based on parameters, and the retrieved components are instantiated according to the parameters and assembled into the building model to obtain instantiated component parameters. For special components that do not exist in the component library, a generative method is used to create them to supplement the instantiated component parameters and obtain component design parameters.

[0070] The specific implementation method of S1404 includes: processing the component design parameters based on the stylized diffusion layer in the preset decoration generation network to generate painted patterns; processing the component design parameters based on the carving generation layer in the preset decoration generation network to generate carving information; processing the component design parameters based on the material and texture layers in the preset decoration generation network to generate material maps; and using painted patterns, carving information, and material maps as decoration design information.

[0071] Optionally, the painted patterns can be further subdivided into different levels of complexity and themes. Carving information can include wood carving, stone carving, and brick carving. Material textures include wood grain, brick and stone textures, and tile surface effects.

[0072] In S1405, context passing specifically refers to using the features of the upper-level generated results as conditions for the lower-level generated results; global coordination processing specifically refers to performing global style coordination checks and fine-tuning based on style consistency loss data after generation.

[0073] In this way, by adopting a hierarchical extraction strategy from overall layout to details, and using appropriate methods for feature extraction at each level, the generation accuracy of each level is ensured, thereby solving the problem of inconsistent accuracy of existing methods in multi-scale generation.

[0074] S150. Based on hierarchical design parameters and at least one preset design style, generate a target design image of the target building.

[0075] To support different design styles, from traditional to innovative, electronic devices can integrate hierarchical design parameters with at least one preset design style to generate design images that conform to the corresponding design style.

[0076] Optionally, at least one preset design style may include a restoration style, a heritage style, an interpretation style, a fusion style, or an innovative style.

[0077] Among them, the Restoration Style faithfully reproduces traditional architectural forms, introducing modern technology in key areas, maintaining a traditional appearance while incorporating modern equipment and facilities internally, and using modern materials for concealed structural reinforcement. The Heritage Style retains the core characteristics of traditional architecture, appropriately simplifying or adjusting secondary elements, preserving iconic traditional features (such as roof forms and brackets), simplifying secondary decorative elements, and using materials that mimic traditional modern styles. The Interpretive Style extracts the design principles and aesthetic features of traditional architecture, reinterpreting them with modern techniques, abstracting the principles of proportion, rhythm, and layering of traditional architecture, expressing traditional artistic conception with modern materials and forms, and preserving the traditional spatial experience. The Fusion Style juxtaposes traditional and modern elements, creating a dialogue between old and new, combining traditional and modern architecture, pairing traditional and modern materials, and integrating traditional forms with modern functions. The Innovative Style draws inspiration from traditional architecture, creating entirely new architectural forms, extracting design inspiration from traditional architecture, breaking away from direct correspondence with traditional forms, and pursuing a modern expression of cultural spirit.

[0078] In some embodiments, the specific implementation method of S150 includes: S1501, constraining the hierarchical design parameters to generate constrained hierarchical design parameters; S1502, converting the constrained hierarchical design parameters into target design parameters corresponding to at least one design style, and generating a target design image based on the target design parameters.

[0079] The specific implementation method of S1501 includes: performing functional constraint processing, structural constraint processing, and code constraint processing on the hierarchical design parameters to generate constrained hierarchical design parameters; wherein, the functional constraint processing includes at least one of functional area constraint, streamline rationality constraint, and spatial sequence constraint; the structural constraint processing includes at least one of column grid rationality constraint, load transfer constraint, and component dimensional constraint; and the code constraint processing includes at least one of building code constraint and construction code constraint.

[0080] Among these, functional area constraints refer to checking whether the area of ​​each functional space meets the requirements. Circulation rationality constraints refer to analyzing whether the functional circulation is reasonable, avoiding circulation intersections and dead ends. Spatial sequence constraints refer to verifying whether the spatial sequence unique to traditional architecture is complete (e.g., the progressive relationship of courtyards).

[0081] Among these, column grid rationality constraints verify whether the column grid layout is reasonable and whether the span is within the structurally reasonable range. Load transfer constraints simplify the verification of the integrity of the load transfer path. Component dimensional constraints verify whether the component dimensions conform to the provisions of traditional building codes. Building code constraints verify whether the basic requirements of current building design codes are met: fire separation distance; evacuation distance; accessibility design; and sunlight requirements. Construction code constraints verify whether the rules of traditional building construction methods are followed: component proportions; grade correspondence; and regional suitability.

[0082] Furthermore, in addition to the hierarchical design parameters after constraints, suggestions for correcting violations can also be generated. These suggestions include: problem location: indicating the specific location and parameters of the violation; correction direction: suggesting the direction and scope of adjustment; and automatic correction options: providing automatic correction schemes for the user to choose from.

[0083] Furthermore, after generating the target design image, the fusion effect of the target design image can be evaluated. Specifically, the effect of different fusion strategies can be pre-evaluated before generation: the harmony of the fusion scheme can be predicted to help users select a suitable fusion strategy.

[0084] In this way, based on the design style that moves from restoration to innovation, and with refined control of integration parameters, it supports various design needs that combine inheritance and innovation, and solves the problem that the generation of existing design styles is relatively monotonous.

[0085] This disclosure discloses a design image generation method that obtains design requirement information for a target building. This design requirement information includes textual design requirements representing building elements, image-based design requirements, hand-drawn design requirements, and 3D model design requirements. The method fuses these requirements to obtain a fused design intent for the target building. It then parametrically translates this fused design intent to generate structured design parameters for the target building. Hierarchical feature extraction is performed on these structured design parameters to generate hierarchical design parameters for the target building. Finally, based on these hierarchical design parameters and at least one preset design style, a target design image for the target building is generated. Thus, by fusing and parametrically translating the multimodal design requirements of the target building, transparent and controllable structured design parameters are generated. Multi-level controllable parameter extraction is then performed on these structured design parameters, and a design image conforming to the corresponding design style is intelligently generated based on the obtained hierarchical design parameters. This method achieves both the inheritance and innovation of design images while improving the intelligence and efficiency of the design image generation method.

[0086] This disclosure also provides a design image generation apparatus for implementing the above-described design image generation method, which is described below in conjunction with... Figure 2 The following description is provided. In this embodiment, the design image generation apparatus can be executed by an electronic device or a server. The electronic device may include a desktop computer, laptop, tablet, or other smart devices. The server may include a cloud server or a server cluster.

[0087] Figure 2 A schematic diagram of the structure of an image generation apparatus provided in an embodiment of this disclosure is shown.

[0088] like Figure 2 As shown, the image generation device 200 may include: The first acquisition module 210 is used to acquire the design requirement information of the target building, wherein the design requirement information includes text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn drawings, and three-dimensional model design requirements based on three-dimensional model representation of building elements. The information fusion module 220 is used to fuse the text design requirements, the image design requirements, the hand-drawn design requirements and the three-dimensional model design requirements to obtain the fused design intent of the target building; The first generation module 230 is used to parametrically translate the fused design intent and generate the structural design parameters of the target building; The second generation module 240 is used to perform hierarchical feature extraction on the structured design parameters to generate hierarchical design parameters for the target building. The third generation module 250 is used to generate a target design image of the target building based on the hierarchical design parameters and at least one preset design style.

[0089] This disclosure discloses a design image generation apparatus that acquires design requirement information for a target building. This design requirement information includes textual design requirements representing building elements, image-based design requirements representing building elements, hand-drawn design requirements representing building elements, and three-dimensional model design requirements representing building elements. The apparatus fuses these textual, image, hand-drawn, and three-dimensional model design requirements to obtain a fused design intent for the target building. It then parametrically translates this fused design intent to generate structured design parameters for the target building. Hierarchical feature extraction is performed on these structured design parameters to generate hierarchical design parameters for the target building. Finally, based on these hierarchical design parameters and at least one preset design style, a target design image of the target building is generated. Thus, by fusing and parametrically translating the multimodal design requirements of the target building, transparent and controllable structured design parameters are generated. Multi-level controllable parameter extraction is then performed on these structured design parameters, and a design image conforming to the corresponding design style is intelligently generated based on the obtained hierarchical design parameters. This approach achieves both the inheritance and innovation of design images while improving the intelligence and efficiency of the design image generation method.

[0090] In some embodiments of this disclosure, the design requirement information includes the textual design requirements; the first acquisition module 210 includes: The first acquisition unit is used to acquire textual requirement information for the target building; The semantic understanding unit is used to perform semantic understanding on the text requirement information using a preset text understanding model to obtain initial semantic information; The annotation unit is used to perform specialized word expansion and semantic role annotation on the initial semantic information to obtain semantic annotation information; The first determining unit is used to perform intent vector encoding on the semantic annotation information to obtain the text design requirements.

[0091] In some embodiments of this disclosure, the design requirement information includes the image design requirements; the first acquisition module 210 includes: The second acquisition unit is used to acquire image requirement information of the target building; The second determining unit is used to extract and analyze the style features of the image requirement information using a preset visual analysis model to obtain the image design requirements.

[0092] In some embodiments of this disclosure, the design requirement information includes the hand-drawn design requirements; the first acquisition module 210 includes: The third acquisition unit is used to acquire hand-drawn drawing requirements for the target building. The standardization unit is used to standardize the hand-drawing requirement information to obtain standardized hand-drawing requirement information. The third determining unit is used to identify the design intent from the standardized hand-drawn drawing requirements information using a preset hand-drawn drawing understanding model, and to map the architectural requirements in the design intent to the hand-drawn drawing design requirements.

[0093] In some embodiments of this disclosure, the design requirement information includes the three-dimensional model design requirements; the first acquisition module 210 includes: The fourth acquisition unit is used to acquire the required information for the three-dimensional model of the target building; The fourth determining unit is used to perform site understanding and environmental analysis on the requirements of the three-dimensional model, and generate the design requirements of the three-dimensional model.

[0094] In some embodiments of this disclosure, the information fusion module 220 includes: The semantic alignment unit is used to semantically align the text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements that describe the same object in the target building. The first generation unit is used to perform semantic understanding, structural understanding, and decorative understanding on semantically aligned text design requirements, semantically aligned image design requirements, semantically aligned hand-drawn design requirements, and semantically aligned 3D model design requirements, and generate semantic design intent, structural design intent, and decorative design intent. The second generation unit is used to perform information completion and fusion processing on the semantic design intent, the structural design intent and the decorative design intent to generate the fused design intent.

[0095] In some embodiments of this disclosure, the first generation module 230 includes: The compression coding unit is used to perform compression coding on the fusion design based on the intent coding layer in the parametric translation network to obtain intent coding parameters. The prediction unit is used to predict discrete parameters, continuous parameters, and structural parameters from the intent encoding parameters based on the parameter prediction layer in the parameterized translation network. The fifth determining unit is used to apply proportional constraints, hierarchical constraints, and construction constraints to the discrete parameters, the continuous parameters, and the structural parameters based on the parameter consistency layer in the parameterized translation network, so as to obtain the structured design parameters.

[0096] In some embodiments of this disclosure, the second generation module 240 includes: The first extraction unit is used to extract layout design parameters from the structured design parameters based on a preset layout generation network. The second extraction unit is used to extract morphological design parameters from the layout design parameters based on a preset morphological generation layer. The third extraction unit is used to extract component design parameters from the morphological design parameters based on a preset component generation network. The fourth extraction unit is used to extract preferred design parameters from the component design parameters based on a preset decoration generation network; The third generation unit is used to generate the hierarchical design parameters by performing context passing and global coordination processing on the layout design parameters, the shape design parameters, the component design parameters and the preference design parameters based on a preset consistency control network.

[0097] In some embodiments of this disclosure, the first extraction unit is specifically used for: Based on the site analysis layer in the preset layout generation network, site analysis is performed on the structured design parameters to obtain the buildable area and orientation information; Based on the layout diagram generation layer in the preset layout generation network, the constructable area and the orientation information are used for layout planning and adjustment to generate the layout design parameters.

[0098] In some embodiments of this disclosure, the second extraction unit is specifically used for: Based on the volume model generation layer in the preset morphology generation layer, the layout design parameters are analyzed for plan layout and section parameters to generate a building volume model. Based on the roof shape generation layer in the preset shape generation layer, the layout design parameters are analyzed to generate a three-dimensional roof shape. Based on the morphology variant generation layer in the preset morphology generation layer, morphology variant analysis is performed on the layout design parameters to generate morphology variant parameters; The building mass model, the three-dimensional shape of the roof, and the parameters of the shape variants are used as the shape design parameters.

[0099] In some embodiments of this disclosure, the third extraction unit is specifically used for: Based on the instantiation layer in the preset component generation network, the components found from the preset component library are instantiated according to parameters to obtain instantiated component parameters. The component supplement layer in the preset component generation network is used to supplement the instantiated component parameters to obtain the component design parameters.

[0100] In some embodiments of this disclosure, the fourth extraction unit is specifically used for: Based on the stylized diffusion layer in the preset decoration generation network, the component design parameters are processed to generate painted patterns; Based on the carving generation layer in the preset decoration generation network, the component design parameters are processed to generate carving information; Based on the material and texture layers in the preset decoration generation network, the component design parameters are processed to generate a material texture map; The painted patterns, the engraving information, and the material texture are used as the decorative design information.

[0101] In some embodiments of this disclosure, the third generation module 250 includes: A constraint unit is used to constrain the hierarchical design parameters and generate constrained hierarchical design parameters. The fourth generation unit is used to convert the constrained hierarchical design parameters into target design parameters corresponding to the at least one design style, and generate the target design image based on the target design parameters.

[0102] It should be noted that, Figure 2 The design image generation device 200 shown can perform... Figure 1 The various steps in the method embodiment shown are implemented. Figure 1 The processes and effects in the method embodiments shown are not described in detail here.

[0103] Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure is shown.

[0104] like Figure 3 As shown, the electronic device may include a processor 301 and a memory 302 storing computer program instructions.

[0105] Specifically, the processor 301 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0106] Memory 302 may include a large-capacity storage for advertising or instructions. For example, and not limitingly, memory 302 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 302 may include removable or non-removable (or fixed) media. Where appropriate, memory 302 may be internal or external to the integrated gateway device. In a particular embodiment, memory 302 is non-volatile solid-state memory. In a particular embodiment, memory 302 includes read-only memory (ROM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (Electrically Programmable ROM, EPROM), an electrically erasable programmable PROM (EEPROM), an electrically alterable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0107] The processor 301 acquires and executes computer program instructions stored in the memory 302 to perform the steps of the design image generation method provided in the embodiments of this disclosure.

[0108] In one example, the electronic device may also include a transceiver 303 and a bus 304. Wherein, as... Figure 3 As shown, the processor 301, memory 302 and transceiver 303 are connected via bus 304 and communicate with each other.

[0109] Bus 304 may include hardware, software, or both. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industrial Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 304 may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application considers any suitable bus or interconnection.

[0110] The following are embodiments of a computer-readable storage medium provided in this disclosure. This computer-readable storage medium and the design image generation methods of the above embodiments belong to the same inventive concept. For details not described in detail in the embodiments of the computer-readable storage medium, please refer to the embodiments of the design image generation methods described above.

[0111] This embodiment provides a storage medium containing computer-executable instructions, which, when executed by a computer processor, are used to perform a design image generation method, including: Obtain design requirement information for the target building, wherein the design requirement information includes text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn representation of building elements, and three-dimensional model design requirements based on three-dimensional model representation of building elements. The text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements are integrated to obtain the integrated design intent of the target building; The integrated design intent is parametrically translated to generate the structural design parameters of the target building; Hierarchical feature extraction is performed on the structured design parameters to generate hierarchical design parameters for the target building; Based on the hierarchical design parameters and at least one preset design style, a target design image of the target building is generated.

[0112] Of course, the computer-executable instructions provided in the embodiments of this disclosure are not limited to the above-described method operations, but can also perform related operations in the design image generation method provided in any embodiment of this disclosure.

[0113] Based on the above description of the implementation methods, those skilled in the art can clearly understand that this disclosure can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer cloud platform (which may be a personal computer, server, or network cloud platform, etc.) to execute the design image generation method provided in the various embodiments of this disclosure.

[0114] Note that the above description is merely a preferred embodiment and the technical principles employed in this disclosure. Those skilled in the art will understand that this disclosure is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of this disclosure. Therefore, although this disclosure has been described in detail through the above embodiments, it is not limited to the above embodiments. Many other equivalent embodiments may be included without departing from the concept of this disclosure, and the scope of this disclosure is determined by the scope of the appended claims.

Claims

1. A design image generation method, characterized in that, include: Obtain design requirement information for the target building, wherein the design requirement information includes text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn representation of building elements, and three-dimensional model design requirements based on three-dimensional model representation of building elements. The text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements are integrated to obtain the integrated design intent of the target building; The integrated design intent is parametrically translated to generate the structural design parameters of the target building; Hierarchical feature extraction is performed on the structured design parameters to generate hierarchical design parameters for the target building; Based on the hierarchical design parameters and at least one preset design style, a target design image of the target building is generated.

2. The method according to claim 1, characterized in that, The design requirements information includes the textual design requirements; The acquisition of design requirements information for the target building includes: Obtain the textual requirements information for the target building; Using a pre-defined text understanding model, the text requirement information is semantically understood to obtain initial semantic information; The initial semantic information is expanded using specialized terms and labeled with semantic roles to obtain semantic annotation information; The semantic annotation information is encoded into intent vectors to obtain the text design requirements.

3. The method according to claim 1, characterized in that, The design requirements information includes the image design requirements; The acquisition of design requirements information for the target building includes: Obtain the image requirements information for the target building; Using a pre-defined visual analysis model, style features are extracted and analyzed from the image requirement information to obtain the image design requirements.

4. The method according to claim 1, characterized in that, The design requirements information includes the hand-drawn design requirements; The acquisition of design requirements information for the target building includes: Obtain the hand-drawn sketch requirements for the target building; The hand-drawing requirement information is standardized to obtain standardized hand-drawing requirement information; Using a pre-defined hand-drawn drawing understanding model, the design intent is identified from the standardized hand-drawn drawing requirements information, and the architectural requirements in the design intent are mapped to the hand-drawn drawing design requirements.

5. The method according to claim 1, characterized in that, The design requirements information includes the 3D model design requirements; The acquisition of design requirements information for the target building includes: Obtain the required information for the 3D model of the target building; The site understanding and environmental analysis of the three-dimensional model requirements are performed to generate the three-dimensional model design requirements.

6. The method according to claim 1, characterized in that, The process of fusing the text design requirements, the image design requirements, the hand-drawn design requirements, and the 3D model design requirements to obtain the integrated design intent of the target building includes: Semantic alignment is performed on the text design requirements, image design requirements, hand-drawn design requirements, and 3D model design requirements describing the same object in the target building; Semantic understanding, structural understanding, and decorative understanding are performed on semantically aligned text design requirements, semantically aligned image design requirements, semantically aligned hand-drawn design requirements, and semantically aligned 3D model design requirements to generate semantic design intent, structural design intent, and decorative design intent. After information completion of the semantic design intent, the structural design intent, and the decorative design intent, a fusion process is performed to generate the fused design intent.

7. The method according to claim 1, characterized in that, The step of parametrically translating the fused design intent to generate the structural design parameters of the target building includes: Based on the intent coding layer in the parametric translation network, the fusion design is compressed and encoded to obtain intent coding parameters; Based on the parameter prediction layer in the parameterized translation network, discrete parameters, continuous parameters, and structural parameters are predicted from the intent encoding parameters; Based on the parameter consistency layer in the parametric translation network, proportional constraints, hierarchical constraints, and construction constraints are applied to the discrete parameters, the continuous parameters, and the structural parameters to obtain the structured design parameters.

8. The method according to claim 1, characterized in that, The step of extracting hierarchical features from the structured design parameters to generate hierarchical design parameters for the target building includes: Based on a preset layout generation network, layout design parameters are extracted from the structured design parameters; Based on the preset shape generation layer, shape design parameters are extracted from the layout design parameters; Based on a preset component generation network, component design parameters are extracted from the morphological design parameters; Based on a preset decoration generation network, preferred design parameters are extracted from the component design parameters; Based on a preset consistency control network, the layout design parameters, the shape design parameters, the component design parameters, and the preference design parameters are processed through context passing and global coordination to generate the hierarchical design parameters.

9. The method according to claim 8, characterized in that, The preset layout generation network extracts layout design parameters from the structured design parameters, including: Based on the site analysis layer in the preset layout generation network, site analysis is performed on the structured design parameters to obtain the buildable area and orientation information; Based on the layout diagram generation layer in the preset layout generation network, the constructable area and the orientation information are used for layout planning and adjustment to generate the layout design parameters.

10. The method according to claim 8, characterized in that, The preset shape generation layer extracts shape design parameters from the layout design parameters, including: Based on the volume model generation layer in the preset morphology generation layer, the layout design parameters are analyzed for plan layout and section parameters to generate a building volume model. Based on the roof shape generation layer in the preset shape generation layer, the layout design parameters are analyzed to generate a three-dimensional roof shape. Based on the morphology variant generation layer in the preset morphology generation layer, morphology variant analysis is performed on the layout design parameters to generate morphology variant parameters; The building mass model, the three-dimensional shape of the roof, and the parameters of the shape variants are used as the shape design parameters.

11. The method according to claim 8, characterized in that, The component generation network based on the preset component generation network extracts component design parameters from the morphological design parameters, including: Based on the instantiation layer in the preset component generation network, the components found from the preset component library are instantiated according to parameters to obtain instantiated component parameters. The component supplement layer in the preset component generation network is used to supplement the instantiated component parameters to obtain the component design parameters.

12. The method according to claim 8, characterized in that, The preset decoration generation network extracts preferred design parameters from the component design parameters, including: Based on the stylized diffusion layer in the preset decoration generation network, the component design parameters are processed to generate painted patterns; Based on the carving generation layer in the preset decoration generation network, the component design parameters are processed to generate carving information; Based on the material and texture layers in the preset decoration generation network, the component design parameters are processed to generate a material texture map; The painted patterns, the engraving information, and the material texture are used as the decorative design information.

13. The method according to claim 1, characterized in that, The step of generating a target design image of the target building based on the hierarchical design parameters and at least one preset design style includes: The hierarchical design parameters are constrained to generate constrained hierarchical design parameters; The constrained hierarchical design parameters are transformed into target design parameters corresponding to the at least one design style, and the target design image is generated based on the target design parameters.

14. The method according to claim 13, characterized in that, The step of constraining the hierarchical design parameters to generate constrained hierarchical design parameters includes: The hierarchical design parameters are subjected to functional constraint processing, structural constraint processing, and specification constraint processing to generate constrained hierarchical design parameters. The functional constraint processing includes at least one of functional area constraints, streamline rationality constraints, and spatial sequence constraints; the structural constraint processing includes at least one of column grid rationality constraints, load transfer constraints, and component dimensional constraints; and the specification constraint processing includes at least one of building code constraints and construction code constraints.

15. An image generation device, characterized in that, include: The first acquisition module is used to acquire the design requirement information of the target building, wherein the design requirement information includes text design requirements based on text representation of building elements, image design requirements based on image representation of building elements, hand-drawn design requirements based on hand-drawn drawings, and three-dimensional model design requirements based on three-dimensional model representation of building elements. The information fusion module is used to fuse the text design requirements, the image design requirements, the hand-drawn design requirements, and the three-dimensional model design requirements to obtain the fused design intent of the target building. The first generation module is used to parametrically translate the fused design intent and generate the structural design parameters of the target building. The second generation module is used to extract hierarchical features from the structured design parameters to generate hierarchical design parameters for the target building. The third generation module is used to generate a target design image of the target building based on the hierarchical design parameters and at least one preset design style.

16. An electronic device, characterized in that, include: processor; Memory, used to store executable instructions; The processor is configured to retrieve the executable instructions from the memory and execute the executable instructions to implement the method of any one of claims 1-14.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that, The storage medium stores a computer program that, when executed by a processor, causes the processor to implement the method described in any one of claims 1-14.

Images