Building interface style migration and optimization method based on multi-machine learning model coupling

Abstract

The invention provides a building interface style migration and optimization method based on multi-machine learning model coupling. According to the method, a building interface style migration model is constructed based on various machine learning technologies, so that different building style characteristics and a mapping relation thereof are obtained. Based on a de-noising diffusion probability model and a cyclic generative adversarial network, high-quality style migration among different architectural styles is realized, the stability and detail quality of an architectural interface design scheme are optimized through a neural network, and a multi-objective optimization method is combined to generate an architectural interface layout scheme meeting urban design requirements. Compared with the prior art, the method has remarkable advantages in the aspects of building interface style migration precision, generation scheme adaptability and optimization design efficiency. Meanwhile, due to the reversibility of the design process, the generation result of the scheme can be returned to any step at any time to be optimized until the generated image meets the construction design requirement.

Description

Technical Field

[0001] The present invention belongs to the cross - technical field of computer vision and architectural design, and particularly relates to a method for architectural interface style transfer and optimization based on the coupling of multiple machine learning models. This method uses the coupling technology of multiple machine learning models to extract, analyze, and transfer the style features of building facades, and can be widely applied to fields such as building facade renovation, historical block style protection, urban renewal, and virtual reality building generation. Background Technique

[0002] With the continuous acceleration of the urbanization process, as an important part of the urban image and style, the design and renovation of building facades are gradually becoming the core issues in the fields of urban planning and architectural design. Building facades not only carry historical culture and regional characteristics but also affect the visual order of urban space and the living experience of residents. However, traditional architectural interface design mainly relies on designers' experience and manual drawing. Although this method can reflect personalized design concepts, it is difficult to complete large - scale data processing and style transfer in a short time. Especially in scenarios such as urban renewal, old city renovation, and modern architectural style optimization, traditional methods often show many limitations such as long design cycles, unstable effects, and difficulty in quantifying optimization plans. In addition, the diversity and complexity of architectural interface design make it difficult to ensure the coherence of architectural styles and environmental adaptability in practical applications, further restricting the innovative design and intelligent optimization of architectural interfaces.

[0003] In recent years, the rapid development of machine learning technology has brought unprecedented changes to architectural interface design, especially showing great potential in data - driven design methods and automated optimization. By collecting, processing, and analyzing a large amount of building interface image data with high precision and combining advanced machine learning models, it is possible to effectively mine the feature distribution patterns of architectural styles and establish mapping relationships between different styles. Training the style transfer model using the denoising diffusion probability model can not only accurately extract the visual features of architectural styles but also achieve smooth conversion of style features, so that different architectural elements maintain structural consistency during the style transformation process. At the same time, the introduction of the Cycle - Generative Adversarial Network (CycleGAN) makes the style transfer process of architectural interfaces more stable and adaptable, enabling two - way conversion between different architectural styles without aligned data. In addition, by combining neural network control optimization (such as ControlNet) with multi - objective optimization algorithms, the detail quality and overall stability of architectural interface generation can be further improved, and in - depth optimization can be carried out in aspects such as architectural aesthetics, functionality, and environmental adaptability. The integrated application of the above - mentioned technologies not only greatly improves the intelligence and automation level of architectural interface design but also provides a reliable scientific basis for the evaluation and feedback of architectural plans in actual construction, thus promoting the digital and intelligent transformation of the fields of architectural design, urban planning, and construction. Summary of the Invention

[0004] The object of the present invention is to solve the problems in the prior art, and a method for building interface style transfer and optimization based on the coupling of multiple machine learning models is proposed.

[0005] The present invention is realized through the following technical solutions. The present invention proposes a method for building interface style transfer and optimization based on the coupling of multiple machine learning models, and the method includes the following steps:

[0006] (a) Collect building interface data of the target street and perform image preprocessing;

[0007] (b) Use the denoising diffusion probabilistic model DDPM to train the style transfer model and establish the mapping relationship between different building styles;

[0008] (c) Perform style conversion through the cyclic generative adversarial network CycleGAN and optimize the interface layout;

[0009] (d) Combine the ControlNet neural network to optimize the generation of the building interface and improve the image quality and stability;

[0010] (e) Use the multi-objective optimization algorithm to screen the best building interface design scheme;

[0011] (f) Evaluate and optimize the effect of the selected building interface design scheme.

[0012] Further, the step (a) is specifically:

[0013] (1) Use the drone oblique photography or the fixed street view camera system to obtain the building interface image data;

[0014] (2) Perform coordinate correction, denoising, image cropping and enhancement processing on the collected image data to ensure that the data meets the requirements of subsequent style transfer and optimization.

[0015] Further, the step (b) is specifically:

[0016] (1) Apply DDPM to the preprocessed building interface data for training and extract the deep features of each building style;

[0017] (2) Establish the mapping relationship between different building styles, so as to provide basic data support for style conversion.

[0018] Further, the step (c) is specifically:

[0019] (1) Use CycleGAN to realize the style conversion of the building interface image, so that the generated image is naturally connected in style;

[0020] (2) Combine the parametric modeling method to optimize the layout of the converted building interface to ensure that the generated results comply with urban design specifications.

[0021] Further, the specific steps of step (d) are as follows:

[0022] (1) Use the ControlNet neural network to constrain and correct the building interface images generated during the style transfer process to improve the image detail quality;

[0023] (2) Combine the user's visual feedback to dynamically adjust the key parameters during the generation process to ensure the stability of the generated results.

[0024] Further, the specific steps of step (e) are as follows:

[0025] (1) Construct an optimization model based on multiple objective functions of architectural aesthetics, architectural practicality, and environmental adaptability;

[0026] (2) Use the multi-objective optimization algorithm to batch-screen the generated building interface design schemes to determine the optimal scheme.

[0027] Further, the effect evaluation and optimization of the selected building interface design scheme in step (f) include: inviting stakeholders to evaluate the selected scheme and further optimize according to the evaluation results to ensure that the aesthetics, practicality, and environmental adaptability of the scheme reach the best balance.

[0028] The present invention proposes a building interface style transfer and optimization system based on the coupling of multiple machine learning models, and the system includes:

[0029] Collection module: Collect the building interface data of the target street and perform image preprocessing;

[0030] Training module: Use the denoising diffusion probabilistic model DDPM to train the style transfer model and establish the mapping relationship between different architectural styles;

[0031] Style conversion module: Perform style conversion through the cyclic generative adversarial network CycleGAN and optimize the interface layout;

[0032] Interface optimization module: Combine the ControlNet neural network to optimize the generation of the building interface and improve the image quality and stability;

[0033] Screening module: Use the multi-objective optimization algorithm to screen the best building interface design scheme;

[0034] Evaluation and optimization module: Perform effect evaluation and optimization on the selected building interface design scheme.

[0035] The present invention also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for architectural interface style transfer and optimization based on the coupling of multiple machine learning models are implemented.

[0036] The present invention also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the method for architectural interface style transfer and optimization based on the coupling of multiple machine learning models are implemented.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] Through the organic combination of the foregoing steps, the present invention can effectively improve the accuracy of architectural interface style transfer, optimize the adaptability of the generated scheme, and improve the efficiency of optimization design. It successfully solves the problems of low efficiency, unstable conversion, and difficult optimization existing in traditional architectural interface style transfer. While taking into account the aesthetic properties of architectural interface design, it improves its practicability and construction feasibility, and can be widely applied to multiple fields such as urban renewal, building renovation, virtual city modeling, historical block protection, and intelligent building design. At the same time, due to the reversibility of the design process, the generated result of the scheme can be returned to any step at any time for optimization until the generated image meets the construction design requirements. Description of the Drawings

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0040] Figure 1 It is a flowchart of the method for architectural interface style transfer and optimization based on the coupling of multiple machine learning models according to the present invention.

[0041] Figure 2 It is a logical schematic diagram of the generation of architectural interface design for the embodiment.

[0042] Figure 3 It is a schematic diagram of the training and application of the style transfer model.

[0043] Figure 4 It is a schematic diagram of the generated scheme for architectural interface design. Detailed Embodiments

[0044] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0045] Combined with Figures 1-4 , the present invention proposes a method for building interface style transfer and optimization based on the coupling of multiple machine learning models, and the method includes the following steps:

[0046] (a) Collect the building interface data of the target street and perform image preprocessing;

[0047] The specific steps of step (a) are as follows:

[0048] (1) Use unmanned aerial vehicle (UAV) oblique photography or a fixed street view camera system to obtain building interface image data;

[0049] (2) Perform coordinate correction, denoising, image cropping and enhancement processing on the collected image data to ensure that the data meets the requirements of subsequent style transfer and optimization.

[0050] In step (a), the building interface image data of the target area is obtained through UAV aerial photography, street view acquisition system or other remote sensing devices. During the acquisition process, computer vision technology is used for image correction to eliminate the deformation caused by factors such as perspective distortion and lens distortion. An automatic occlusion detection algorithm is adopted to eliminate the building interface images with more than 10% occlusion by trees, wires, pedestrians, etc. to ensure the data quality. Based on noise removal, color enhancement, and edge detection preprocessing techniques, the clarity, color consistency, and recognizability of building features are improved, and the building interfaces with more than 10% occlusion are screened out. In order to ensure the unity and adaptability of the input data, each building monomer is segmented and uniformly cropped into training data of 1024×1024px to ensure the unity and adaptability of the input data, and finally a high-quality building interface dataset is constructed.

[0051] (b) Use the denoising diffusion probabilistic model (DDPM) to train the style transfer model and establish the mapping relationship between different building styles;

[0052] The specific steps of step (b) are as follows:

[0053] (1) Apply DDPM to the preprocessed building interface data for training and extract the deep features of each building style;

[0054] (2) Establish the mapping relationship between different building styles, thereby providing basic data support for style conversion.

[0055] In step (b), the preprocessed building interface image data is subjected to deep learning training using DDPM to extract the core features of different architectural styles. This model learns the variation patterns of different architectural styles in terms of spatial layout, materials, colors, and detailed features by gradually adding noise to the input images and performing reverse denoising training, thereby establishing a mapping relationship between architectural styles. During the training process, a self-supervised learning mechanism is used to optimize the model parameters, enabling it to gradually recover building interface images that conform to the target style from the noise distribution. Compared with traditional style transfer methods, DDPM performs better in terms of the smoothness and consistency of style conversion and can generate higher-quality building interface style transfer results.

[0056] (c) Perform style conversion through the CycleGAN (Cyclic Generative Adversarial Network) and optimize the interface layout;

[0057] The specific steps of step (c) are as follows:

[0058] (1) Use CycleGAN to achieve style conversion of building interface images, making the generated images achieve natural connection in style;

[0059] (2) Combine parametric modeling methods to optimize the layout of the converted building interface to ensure that the generated results comply with urban design specifications.

[0060] In step (c), based on the style features extracted by DDPM, CycleGAN is used for building interface style transfer, making the conversion between different styles more natural and coherent. Since CycleGAN is trained in an unsupervised learning manner and does not require paired data before and after style, it can more efficiently handle the diverse building style conversion requirements in actual scenarios. To improve the structural stability and style coordination of the generated building interface, Structural Consistency Loss and Feature Mapping Loss are additionally introduced to ensure that the building interface after style conversion still conforms to the proportional relationship and spatial hierarchy of the original building; during the style conversion process, the building interface is optimized and adjusted in terms of spatial scale, window layout, decorative elements, etc. through parametric modeling, procedural generation, or other layout optimization algorithms to ensure the rationality of the building interface after style transfer in terms of visual aesthetics and spatial function.

[0061] (d) Combine the ControlNet neural network to optimize the generation of the building interface and improve the image quality and stability;

[0062] The specific steps of step (d) are as follows:

[0063] (1) Use the ControlNet neural network to constrain and correct the generated building interface images during the style transfer process, improving the image detail quality;

[0064] (2) Combine the user's visual feedback to dynamically adjust the key parameters during the generation process, ensuring the stability of the generated results.

[0065] In step (d), the ControlNet neural network is introduced to provide additional structural constraints during the image generation process, ensuring that the original form and spatial characteristics of the building are still maintained after the building interface style transfer. ControlNet can accept the input of auxiliary information such as edge detection, depth maps, and line drawings, and combine the outputs of the diffusion model and CycleGAN to finely adjust the generated building interface during the style transfer process. In practical applications, ControlNet can effectively prevent problems such as deformation of building elements and loss of details during the style transfer process, and enhance the structural stability of the interface. To further improve the detail quality of the generated building interface, a super-resolution module is added during the ControlNet optimization process to perform high-definition reconstruction on key areas (such as windows, porches, eaves, etc.), improving the clarity and realism of the final building interface image.

[0066] (e) Use a multi-objective optimization algorithm to screen the best building interface design scheme;

[0067] The specific steps of step (e) are as follows:

[0068] (1) Construct an optimization model based on multiple objective functions of architectural aesthetics, architectural practicality, and environmental adaptability;

[0069] (2) Use a multi-objective optimization algorithm to batch-screen the generated building interface design schemes to determine the optimal scheme.

[0070] In step (e), a multi-objective optimization algorithm is used to screen and optimize the building interface scheme to ensure that the generated scheme reaches an optimal state in terms of aesthetics, practicality, and environmental adaptability. During the optimization process, multiple evaluation indicators are set, including the visual coordination, structural rationality, material matching degree, and environmental adaptability of the building interface. Genetic algorithms, particle swarm optimization algorithms, or other intelligent optimization methods are used to automatically evaluate and select different building style conversion schemes. During the optimization process, in order to improve the stability of the scheme, the Fuzzy Comprehensive Evaluation method is introduced, and combined with the experience weights of building experts, the generated results are evaluated and sorted in multiple dimensions, so as to screen out the best building interface design scheme. In addition, through a dynamic parameter adjustment mechanism, during the optimization process, the evaluation criteria can be adjusted according to the requirements of different cities, climate conditions, and building codes to meet the needs of different application scenarios. This optimization process ensures that the final scheme can reach an optimal state in terms of style coordination, structural rationality, and actual construction feasibility.

[0071] (f) Evaluate and optimize the effects of the selected building interface design scheme.

[0072] The evaluation and optimization of the effects of the selected building interface design scheme in step (f) include: inviting stakeholders to evaluate the selected scheme and further optimizing it according to the evaluation results to ensure that the aesthetics, practicality, and environmental adaptability of the scheme reach the best balance.

[0073] In step (f), the actual effect of the optimized and screened building interface design scheme is evaluated, the performance of the building interface scheme in terms of lighting, shadow, color matching, etc. is analyzed, and the detailed design of the scheme is further optimized to ensure the implementation effect of the final design scheme. During the evaluation process, methods such as lighting simulation, shadow analysis, and material matching detection are used to analyze the performance of the building interface scheme under different lighting conditions. At the same time, combined with street view simulation technology, the visual effects of the building interface under different times and weather conditions are simulated. Evaluate whether there are potential problems in the actual construction process of the building interface, and combine the expert review and public feedback mechanism to finally adjust and optimize the scheme. Finally, the optimized building interface style transfer scheme will be applied to the actual construction to improve the intelligent level of building interface design and contribute to urban renewal and the optimization of building styles. Due to the reversibility of the design process, the generated results of the scheme can be returned to any step at any time for optimization. By repeating the above steps (a) to (e) in the same computing environment until the generated image meets the construction design requirements. The method can effectively improve the accuracy of building interface style transfer, optimize the adaptability of the generated scheme, and improve the efficiency of optimization design. It can be widely applied to many fields such as urban renewal, building renovation, virtual city modeling, historical block protection, and building intelligent design, and has good applicability and promotion prospects.

[0074] The method can effectively improve the accuracy of building interface style transfer, optimize the adaptability of the generated scheme, and improve the efficiency of optimization design. At the same time, due to the reversibility of the design process, the generated results of the scheme can be returned to any step at any time for optimization until the generated image meets the construction design requirements.

[0075] The present invention proposes a building interface style transfer and optimization system based on the coupling of multiple machine learning models, and the system includes:

[0076] Collection module: Collect the building interface data of the target street and perform image preprocessing;

[0077] Training module: Use the denoising diffusion probabilistic model DDPM to train the style transfer model and establish the mapping relationship between different building styles;

[0078] Style conversion module: Perform style conversion through the cyclic generative adversarial network CycleGAN and optimize the interface layout;

[0079] Interface optimization module: Combine the ControlNet neural network to optimize the generation of the building interface and improve the image quality and stability;

[0080] Screening module: Use the multi-objective optimization algorithm to screen the best building interface design scheme;

[0081] Evaluation and Optimization Module: Evaluate and optimize the effects of the selected building interface design solutions.

[0082] The present invention proposes a building interface style transfer and optimization method based on the coupling of multiple machine learning models. Through the systematic coupling of various machine learning technologies, a complete intelligent design process for building interfaces is formed. By extracting building style features through relevant models and technologies, automatic style transfer and optimization are achieved, which not only improves the design efficiency of building interfaces but also makes the style conversion more natural and coherent. At the same time, this method has high flexibility and reversibility during the optimization process. The generated solutions can be returned to any design stage for adjustment at any time to ensure that the finally generated building interface solutions meet the actual construction requirements. In addition, this method can also adapt to different building styles and regional environments, making it have broad application prospects in multiple fields such as historical block protection, urban renewal, and intelligent building design.

[0083] Embodiment

[0084] The present invention provides a building interface style transfer and optimization method based on the coupling of multiple machine learning models. The generation logic of the building interface design in the embodiment is as Figure 2 shown and includes the following steps:

[0085] Step (a), collect the building interface data of the target street and perform image preprocessing.

[0086] The specific content of step (a) is as follows:

[0087] (1) Data collection and acquisition. Obtain the building interface image data of the target area through drone aerial photography, street view collection systems, or other remote sensing devices.

[0088] (2) Image correction and occlusion removal. Use computer vision technology to correct the image to eliminate deformations caused by factors such as perspective distortion and lens distortion, and remove building interface images with more than 10% occlusion.

[0089] (3) Image preprocessing and standardization. Adopt preprocessing technologies such as noise removal, color enhancement, and edge detection to improve the image quality, and uniformly crop it to a standard size of 1024×1024px to construct a high-quality building interface dataset.

[0090] Step (b), train a style transfer model using the denoising diffusion probabilistic model (DDPM) to establish the mapping relationship between different building styles. Among them, the training and application of the style transfer model are shown as Figure 3 shown.

[0091] The specific content of step (b) is as follows:

[0092] (1) Visual understanding and language generation. Use the BLIP natural language processing algorithm to conduct visual understanding on the training samples, analyze the content and features of the building interface images, and at the same time generate corresponding text descriptions (trigger words) to establish the association between images and language.

[0093] (2) Construct style mapping relationships. Substitute the obtained "image-trigger word" pairs into the diffusion model to construct the mapping relationships between different architectural styles, enabling the model to learn the features and conversion rules of various architectural styles.

[0094] (3) Step-by-step regression training. Based on the generation ability of the diffusion model, adopt a step-by-step regression training strategy to enable the model to gradually recover the building interface images that conform to the target style from random noise, and at the same time improve the accuracy and stability of style conversion.

[0095] (4) Manual tuning and optimization. Combine manual tuning during the training process to optimize key parameters (such as denoising steps, learning rate, regularization terms, etc.) to improve the quality of the final style transfer model and make the generated results more in line with architectural aesthetics and actual application requirements.

[0096] Step (c), perform style conversion through a CycleGAN (Cycle Generative Adversarial Network) and optimize the interface layout.

[0097] The specific content of the said step (c) is as follows:

[0098] (1) Style conversion optimization. Based on the style features extracted by DDPM, use CycleGAN to optimize the architectural interface style transfer to make the style conversion more natural and coherent.

[0099] (2) Enhance structural stability. Introduce structural preservation loss and feature mapping loss to ensure that the converted interface conforms to the proportional relationship and spatial hierarchy of the original building.

[0100] (3) Detail adjustment and layout optimization. Combine methods such as parametric modeling and procedural generation to optimize aspects such as window layout, decorative elements, and material details to ensure aesthetics, functionality, and construction feasibility.

[0101] Step (d), optimize the generation of the building interface by combining with the ControlNet neural network to improve the image quality and stability. Among them, the schematic diagram of the building interface design generation scheme is as Figure 4 shown.

[0102] The specific content of the said step (d) is as follows:

[0103] (1) Structural constraint optimization. Introduce the ControlNet neural network to provide additional structural constraints for the style transfer process to ensure that the building interface maintains its original form and spatial features.

[0104] (2) Detail enhancement and stability control. Utilize auxiliary information inputs such as edge detection, depth maps, and line drawings to finely adjust the results of the architectural interface style transfer.

[0105] (3) High-resolution enhancement. Add a high-resolution enhancement module to perform high-definition reconstruction on key areas and improve the clarity and realism of the final architectural interface image.

[0106] Step (e): Use a multi-objective optimization algorithm to screen the best architectural interface design scheme.

[0107] The specific content of step (e) is as follows:

[0108] (1) Evaluation index setting. Set multiple evaluation indexes to optimize the architectural interface scheme.

[0109] (2) Intelligent optimization algorithm screening. Use intelligent optimization methods such as genetic algorithms and particle swarm optimization algorithms to automatically evaluate and screen different style conversion schemes.

[0110] (3) Dynamic parameter adjustment. Through a dynamic parameter adjustment mechanism, optimize the evaluation criteria according to different cities, climate conditions, and building code requirements to adapt to different application scenarios.

[0111] Step (f): Evaluate and optimize the effect of the selected architectural interface design scheme.

[0112] The specific content of step (f) is as follows:

[0113] (1) Street view simulation test. Combine street view simulation technology to simulate the visual effects of the architectural interface under different times and weather conditions.

[0114] (2) Fuzzy comprehensive evaluation. Introduce expert experience weights to conduct multi-dimensional evaluation and ranking of the generated results.

[0115] (3) Iterative optimization and adaptability enhancement. The scheme can return to any step at any time for optimization to ensure that the final result meets the construction design requirements.

[0116] Through the above technical solutions, the present invention provides an architectural interface style transfer and optimization method based on the coupling of multiple machine learning models, which can effectively improve the accuracy of architectural interface style transfer, optimize the adaptability of the generated scheme, and improve the efficiency of optimization design, successfully solving the problems of low efficiency, unstable conversion, and difficult optimization existing in traditional architectural interface style transfer. At the same time, due to the reversibility of the design process, the generated result of the scheme can return to any step at any time for optimization until the generated image meets the construction design requirements.

[0117] In this embodiment, the proposed method for architectural interface style transfer and optimization coupled with multiple machine learning models can operate solely based on street interface image data without the need for complex external auxiliary information. It has good applicability and broad prospects for popularization and can be widely applied to multiple fields such as urban renewal, building renovation, virtual city modeling, historical block protection, and intelligent building design.

[0118] The present invention also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for architectural interface style transfer and optimization based on the coupling of multiple machine learning models are implemented.

[0119] The present invention also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the method for architectural interface style transfer and optimization based on the coupling of multiple machine learning models are implemented.

[0120] The memory in the embodiments of this application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read only memory (ROM), a programmable ROM (PROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DRRAM). It should be noted that the memory of the method described in the present invention is intended to include but not be limited to these and any other suitable types of memory.

[0121] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from a website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a high-density digital video disc (DVD)), or a semiconductor medium (such as a solid state disc (SSD)), etc.

[0122] In the implementation process, the steps of the above method can be completed by the integrated logic circuit of the hardware in the processor or the instructions in the form of software. The steps of the method disclosed in the embodiments of the present application can be directly embodied as being executed by the hardware processor or completed by the combination of the hardware and software modules in the processor. The software module can be located in a mature storage medium in the art such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0123] It should be noted that the processor in the embodiments of the present application may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method embodiments can be completed by the integrated logic circuit in the hardware of the processor or instructions in software form. The above-mentioned processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by a combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method.

[0124] The above has introduced in detail a method for building interface style transfer and optimization based on the coupling of multiple machine learning models. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A method for building interface style transfer and optimization based on the coupling of multiple machine learning models, characterized in that The method includes the following steps: (a) Collect data of the building interfaces on the target street and perform image preprocessing; (b) Use the Denoising Diffusion Probabilistic Model (DDPM) to train a style transfer model and establish the mapping relationship between different architectural styles; (c) Perform style conversion through the Cyclic Generative Adversarial Network (CycleGAN) and optimize the interface layout; (d) Combine the ControlNet neural network to optimize the generation of building interfaces and improve the image quality and stability; (e) Use a multi-objective optimization algorithm to screen the best building interface design scheme; (f) Evaluate and optimize the effects of the selected building interface design scheme.

2. The method according to claim 1, wherein The specific content of step (a) is as follows: (1) Use an unmanned aerial vehicle (UAV) oblique photography or a fixed street view camera system to obtain image data of the building interfaces; (2) Perform coordinate correction, denoising, image cropping, and enhancement processing on the collected image data to ensure that the data meets the requirements of subsequent style transfer and optimization.

3. The method according to claim 1, wherein The specific content of step (b) is as follows: (1) Apply DDPM to the preprocessed building interface data for training and extract the deep features of each architectural style; (2) Establish the mapping relationship between different architectural styles to provide basic data support for style conversion.

4. The method according to claim 1, wherein The specific content of step (c) is as follows: (1) Use CycleGAN to achieve style conversion of the building interface images, making the generated images achieve natural connection in style; (2) Combine parametric modeling methods to optimize the layout of the converted building interfaces to ensure that the generated results meet the urban design specifications.

5. The method according to claim 1, wherein The specific content of step (d) is as follows: (1) Use the ControlNet neural network to constrain and correct the building interface images generated during the style transfer process to improve the image detail quality; (2) Combine the user's visual feedback to dynamically adjust the key parameters during the generation process to ensure the stability of the generated results.

6. The method according to claim 1, characterized in that The specific content of step (e) is as follows: (1) Construct an optimization model according to multiple objective functions such as architectural aesthetics, architectural practicality, and environmental adaptability; (2) Use a multi-objective optimization algorithm to batch-screen the generated building interface design schemes and determine the optimal scheme.

7. The method according to claim 1, characterized in that, The evaluation and optimization of the selected building interface design scheme in step (f) include: inviting stakeholders to evaluate the selected scheme and further optimize according to the evaluation results to ensure that the aesthetics, practicality, and environmental adaptability of the scheme reach the best balance.

8. A building interface style transfer and optimization system based on the coupling of multiple machine learning models, characterized in that, The system includes: A collection module: collect data of the building interfaces on the target street and perform image preprocessing; A training module: use the Denoising Diffusion Probabilistic Model (DDPM) to train a style transfer model and establish the mapping relationship between different architectural styles; A style conversion module: perform style conversion through the Cyclic Generative Adversarial Network (CycleGAN) and optimize the interface layout; An interface optimization module: combine the ControlNet neural network to optimize the generation of building interfaces and improve the image quality and stability; A screening module: use a multi-objective optimization algorithm to screen the best building interface design scheme; An evaluation and optimization module: evaluate and optimize the effects of the selected building interface design scheme.

9. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium for storing computer instructions, characterized in that, When the computer instructions are executed by the processor, the steps of the method according to any one of claims 1-7 are implemented.

Images