Design method and design platform for urban rail transit industry

Abstract

The invention discloses an urban rail transit industrial design method and design platform, and belongs to the technical field of rail transit industrial design, the method comprises the following steps: obtaining multi-modal design input data, the multi-modal design input data comprising at least one of sketches, text descriptions and images; inputting the multi-modal design input data into a pre-trained generative deep learning model to obtain a generated design image; and performing multi-modal post-processing optimization on the generated design image to obtain a final design result. According to the method, multi-modal input such as sketches and text description can be quickly converted into high-quality effect pictures through the generative deep learning model, manual rendering and multiple iterations which are long in time consumption in a traditional process are avoided, and the design response speed is remarkably increased.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of urban rail transit industrial design method and design platform, belong to rail transit industrial design technical field. BACKGROUND

[0002] With the acceleration of urbanization, urban rail transit system is rapidly developing in the world, and higher requirements are put forward for the efficiency, quality and innovation of industrial design of train, station and the like. The traditional design method often relies on manual experience and manual operation, which is time-consuming, long in cycle and low in efficiency, especially when dealing with complex design requirements and changes, the traditional design means is not up to the task. Although modern design tools, such as computer-aided design (CAD), three-dimensional modeling, etc., have been widely used in rail transit design, they still rely on the subjective judgment and experience of designers. These tools cannot actively design innovation and optimization, and design iteration is often limited by time and resources. SUMMARY

[0003] The purpose of the present application is to overcome the deficiencies in the prior art, provide an urban rail transit industrial design method and design platform, which can quickly convert sketch, text description and other multi-modal input into high-quality rendering through a generative deep learning model, avoiding the time-consuming manual rendering and multiple iterations in the traditional process, and significantly improving the design response speed.

[0004] To achieve the above purpose, the present application adopts the following technical solutions: In a first aspect, the present application provides an urban rail transit industrial design method, comprising: Obtaining multi-modal design input data, the multi-modal design input data comprising at least one of sketch, text description and image; Inputting the multi-modal design input data into a pre-trained generative deep learning model to obtain a generated design image; Performing multi-modal post-processing optimization on the generated design image to obtain the final design result; Wherein, inputting the multi-modal design input data into a pre-trained generative deep learning model comprises: When the input data is a sketch, a high-quality rendering is generated through image recognition and deep learning technology; When the input data is a text description, the semantic is parsed through natural language processing technology and the corresponding visual design image is generated; When the input data is an image, the input image is converted into a specified design style through style transfer technology.

[0005] Further, the training method of the pre-trained generative deep learning model comprises: Acquire a training dataset for urban rail transit industrial design, which includes renderings, three-dimensional models, logo designs, and project information. Clean, label, and enhance the training dataset. Distributed training of the generative deep learning model using the processed training dataset. Evaluate and deploy the trained model.

[0006] Further, the multi-modal post-processing optimization of the generated design image includes detail enhancement, color correction, and material optimization.

[0007] Further, the distributed training of the generative deep learning model uses a multi-node parallel computing architecture to split complex computing tasks into multiple subtasks and use multiple computing nodes for parallel processing.

[0008] Further, the method further includes iterative optimization of the generative deep learning model based on historical design data and actual project feedback.

[0009] In a second aspect, the present application provides an urban rail transit industrial design platform, comprising: A data input module for acquiring multi-modal design input data, including at least one of sketches, text descriptions, and images. A generative deep learning model module pre-trained with a generative deep learning model for receiving multi-modal design input data and generating design images. A post-processing optimization module for multi-modal post-processing optimization of the generated design image to obtain the final design result. The generative deep learning model module includes: A sketch generation unit for generating high-quality renderings based on input sketches through a generative adversarial network. A text generation unit for generating visual design images based on natural language descriptions through semantic analysis. A style transfer unit for converting input images to a specified design style through style transfer technology.

[0010] Further, the style transfer unit supports template selection of multiple design styles and automatically adjusts color, material, and layout elements to adapt to the target style.

[0011] Further, it also includes an industrial design database module for storing and managing multi-modal design data to provide data support for model training.

[0012] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0013] Fourthly, the present invention provides a computer device, comprising: Memory, used to store computer programs / instructions; A processor for executing the computer program / instructions to implement the steps of any of the methods described above.

[0014] Fifthly, the present invention provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of any of the methods described above.

[0015] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: This invention provides an industrial design method and platform for urban rail transit. Through a generative deep learning model, it can quickly transform multimodal inputs such as sketches and text descriptions into high-quality renderings, avoiding the time-consuming manual rendering and multiple iterations in traditional processes, and significantly improving design response speed. By deeply integrating artificial intelligence with the industrial design process, this invention not only addresses various bottlenecks in existing technologies but also brings comprehensive and quantifiable significant progress in improving efficiency, reducing costs, ensuring quality, and stimulating innovation. It has significant value in promoting the intelligent upgrading of industrial design for urban rail transit equipment. Attached Figure Description

[0016] Figure 1 This is a flowchart of an urban rail transit industrial design method provided by an embodiment of the present invention; Figure 2 This is a schematic diagram of the sketch generation effect provided in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the conversion of language descriptions into conceptual diagrams according to embodiments of the present invention; Figure 4 This is a style transfer design diagram provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of data processing and model training provided in an embodiment of the present invention. Detailed Implementation

[0017] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0018] Example 1, as Figure 1 As shown in the figure, this embodiment introduces an industrial design method for urban rail transit, including: Acquire multimodal design input data, wherein the multimodal design input data includes at least one of sketches, text descriptions, and images; The multimodal design input data is input into a pre-trained generative deep learning model to obtain the generated design image; The generated design image is optimized through multimodal post-processing to obtain the final design result; The process of inputting the multimodal design input data into a pre-trained generative deep learning model includes: When the input data is a sketch, high-quality renderings are generated using image recognition and deep learning technologies. When the input data is a text description, natural language processing technology is used to parse the semantics and generate corresponding visual design images. When the input data is an image, style transfer technology is used to convert the input image into a specified design style.

[0019] The urban rail transit industrial design method provided in this embodiment involves the following steps in its application process: The process involves acquiring a generative deep learning model, inputting it into the model, and then using the model to perform various creative tasks, including sketch generation, form design optimization, and style transfer. By analyzing design drawings, renderings, and 3D models, deep learning and image recognition technologies are used to quickly generate design solutions, which are then adjusted and optimized based on designer feedback.

[0020] like Figure 2 As shown, the ability to generate renderings from sketches significantly reduces the time required for this process. Designers can input hand-drawn sketches or preliminary design plans into a generative deep learning model. The model quickly analyzes the sketches and generates high-quality renderings based on the designer's requirements. Utilizing powerful image recognition and deep learning capabilities, it extracts potential design elements from the sketches, including structural layout, color schemes, and material textures, and optimizes and adjusts them based on the designer's feedback. In this way, designers not only obtain more realistic and detailed renderings but also see the presentation effects of different design schemes, enabling them to make more informed decisions.

[0021] The advantage of generating renderings from sketches lies not only in improved work efficiency but also in providing designers with more design options. By quickly generating multiple renderings, designers can evaluate the advantages and disadvantages of different solutions in a short period of time, and then choose the most suitable design direction. This approach breaks the time constraints of the traditional design process, providing designers with greater creative space and more opportunities for innovation.

[0022] Furthermore, designers often need to communicate their design intentions through verbal descriptions, especially in the initial communication phase. Verbal descriptions help designers clarify their design thinking, but due to the subjective and complex nature of design language, many design details may not be fully expressed verbally. Traditionally, designers need to achieve this transformation through repeated sketching or discussions.

[0023] like Figure 3 As shown, this method directly transforms a designer's verbal description into an image. Designers only need to input a brief text description, and a generative deep learning model generates the corresponding design image. This process significantly reduces communication costs for designers, especially in team collaborations or communication with clients, allowing them to convey their design ideas more efficiently.

[0024] By combining natural language processing and image generation technologies, this transformation process becomes more precise. The platform not only understands the basic elements described by designers, including spatial layout, style preferences, and material requirements, but also personalizes the design based on the designer's detailed descriptions. For example, when a designer describes a space as having a "warm, modern, and minimalist" feel, the platform can automatically select appropriate color schemes, furniture styles, and materials to generate an image that matches the designer's intent. This approach significantly improves design accuracy and reduces the time spent on repeated communication and revisions.

[0025] Modern design often requires balancing multiple stylistic elements to create works that are both individualistic and aesthetically pleasing. However, when switching between different styles, designers often need to spend a significant amount of time and effort adjusting and optimizing to ensure the harmonious unity of the design.

[0026] like Figure 4 As shown, style transfer technology makes switching between different styles simpler and more efficient. Designers can select different style templates, and the generative deep learning model automatically adjusts elements such as color, material, and shape to achieve rapid and unified transformation of design styles, supporting cross-style integration and overall aesthetic optimization.

[0027] The advantage of style transfer lies in its ability to not only save designers time in redesigning but also help them break through the limitations of traditional styles, enabling more diverse creations. AI, through deep learning technology, can grasp the characteristics and patterns of different design styles, ensuring stylistic consistency across all elements. For example, modern styles often emphasize clean lines and open spatial layouts, while classical styles focus more on elaborate decorations and exquisite details. AI can optimize and adjust designs based on these characteristics.

[0028] To support the functional implementation of generative deep learning models, the platform has constructed a dataset covering multiple fields of urban rail transit data, including historical projects, design standards, and industry cases. This dataset comprehensively covers all aspects of rail transit design, from route planning, station design, and equipment selection to safety regulations and operation management. By cleaning, labeling, and evaluating this data, the platform can provide high-quality training data for generative deep learning models, ensuring the accuracy, completeness, and diversity of the data. This, in turn, improves the model's recognition accuracy, predictive ability, and innovation capabilities in the field of rail transit design.

[0029] For large-scale model training, the platform employs an advanced distributed computing architecture. This architecture breaks down complex computational tasks into multiple sub-tasks, utilizing multiple computing nodes for parallel processing, thereby significantly improving computational efficiency. The platform also collaborates with several institutions, leveraging their powerful computing resources to further accelerate the training process of large-scale deep learning models. These deep learning models, through continuous iterative optimization during training, can perform operations such as recognition, classification, generation, and optimization on a large number of design images, generating design solutions that meet practical needs.

[0030] like Figure 5 The diagram shown illustrates the data processing and model training process, which includes the following steps: Environment configuration: Configure the runtime environment required for training the model, and integrate the runtime and training environment with the remote AI platform.

[0031] Dataset collection: This includes collecting and managing training images, and uploading the images to the AI ​​platform for management.

[0032] Data preprocessing includes functions such as data acquisition, data clearing, and data annotation to prepare data for model training. Data annotation includes automatic image labeling, large model labeling, and manual labeling.

[0033] Model training: Includes capabilities such as model task creation, progress viewing, training management, and model management, meeting users' needs for self-service training of small models.

[0034] Model testing: Includes functions such as test task creation, configuration parameter setting, test result viewing, and test result list, providing test results for model selection.

[0035] Model deployment: Deploy and infer the model, and release the application.

[0036] Under certain special design requirements, such as complex geological conditions and abnormal traffic flow, generative deep learning models are combined with simulation algorithms to optimize and test multiple solutions and provide the optimal solution.

[0037] For large-scale model training, the platform employs a distributed computing architecture, leveraging powerful computing resources to train large-scale deep learning models. These models can perform operations such as recognition, classification, generation, and optimization on design images, and provide personalized solutions based on different design requirements.

[0038] In terms of multimodal image generation and processing, the platform can not only generate high-quality renderings based on text descriptions, but also repair, stylize and enhance the details of existing images, and continuously iterate and optimize design outputs by combining visual recognition and semantic understanding. Its core advantage lies in combining visual recognition and semantic understanding. Through visual recognition technology, the platform can automatically analyze design elements in images and identify the spatial relationships, functional distinctions, and structural layouts between these elements. Semantic understanding, on the other hand, uses natural language processing technology to enable the platform to understand the specific details of design requirements, such as the client's style preferences, functional requirements, and other specific conditions. In this way, the platform can accurately translate client requirements into visual design solutions and provide optimization suggestions.

[0039] Supported by multimodal image generation technology, the platform can not only generate entirely new design renderings but also fine-tune existing design images. Furthermore, it provides designers with a more intelligent design iteration solution. Through deep learning technology, the platform continuously optimizes design models based on historical design data and feedback from actual projects, ensuring that each generated image better meets the needs of the real-world application scenario. By combining these intelligent optimizations, designers can achieve more efficient creation in less time, reducing human error and unnecessary time consumption.

[0040] For example, when a designer inputs a design requirement for a vehicle interior, the platform can retrieve relevant design examples from its database and, combined with AI's reasoning capabilities, generate a design solution that meets the requirements. Simultaneously, the platform can optimize existing images, enhancing their detail and expressiveness to ensure a high-quality presentation of the design.

[0041] The platform implements a multi-node distributed deployment, ensuring service continuity even under high concurrency and high traffic. Data storage employs a distributed database architecture, further enhancing data reliability and access efficiency through data sharding and redundant backup mechanisms. Simultaneously, the platform features load balancing capabilities, dynamically allocating resources based on the pressure on different service modules to ensure stable operation even under uneven load conditions.

[0042] Furthermore, the platform boasts robust data security measures to safeguard user data privacy and security. All data transmissions are protected through encryption protocols to prevent interception or tampering during transmission. Simultaneously, the platform's access control system ensures that only authorized users can access sensitive data and functions. The system also generates detailed access logs and provides real-time monitoring and alerts for abnormal behavior, further mitigating potential security threats.

[0043] This embodiment utilizes a generative deep learning model to rapidly transform multimodal inputs such as sketches and text descriptions into high-quality renderings, avoiding the time-consuming manual rendering and multiple iterations of traditional processes, thus significantly improving design response speed. By deeply integrating artificial intelligence with industrial design processes, this invention not only addresses various bottlenecks in existing technologies but also brings comprehensive and quantifiable significant progress in improving efficiency, reducing costs, ensuring quality, and stimulating innovation, making it of great value in promoting the intelligent upgrading of urban rail transit equipment industrial design.

[0044] Example 2: This example provides an urban rail transit industrial design platform, including: A data input module is used to acquire multimodal design input data, wherein the multimodal design input data includes at least one of sketches, text descriptions, and images; The generative deep learning model module is pre-trained with a generative deep learning model, which is used to receive the multimodal design input data and generate design images; The post-processing optimization module is used to perform multimodal post-processing optimization on the generated design image to obtain the final design result; The generative deep learning model module includes: The sketch generation unit is used to generate high-quality renderings based on the input sketch using a generative adversarial network. The text generation unit is used to generate visual design images based on natural language descriptions through semantic parsing. The style transfer unit is used to convert an input image into a specified design style using style transfer techniques.

[0045] The style transfer unit supports template selection for multiple design styles and automatically adjusts colors, materials, and layout elements to adapt to the target style.

[0046] It also includes an industrial design database module for storing and managing multimodal design data, providing data support for model training.

[0047] The specific functions of each module described above are explained in the relevant content of the method in Embodiment 1, and will not be repeated here.

[0048] Example 3: This example provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described in Example 1.

[0049] Example 4: This example provides a computer device, including: Memory, used to store computer programs / instructions; A processor for executing the computer program / instructions to implement the steps of any of the methods described in Embodiment 1.

[0050] Example 5: This example provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the method described in any one of Examples 1.

[0051] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

[0052] Those skilled in the art will understand that embodiments of this disclosure can be provided as methods, systems, or computer program products. Therefore, this disclosure can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this disclosure can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0053] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0054] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0055] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this disclosure and not to limit its protection scope. Although this disclosure has been described in detail with reference to the above embodiments, those skilled in the art should understand that after reading this disclosure, they can still make various changes, modifications or equivalent substitutions to the specific implementation of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims.

Claims

1. An industrial design method for urban rail transit, characterized in that, include: Acquire multimodal design input data, wherein the multimodal design input data includes at least one of sketches, text descriptions, and images; The multimodal design input data is input into a pre-trained generative deep learning model to obtain the generated design image; The generated design image is optimized through multimodal post-processing to obtain the final design result; The process of inputting the multimodal design input data into a pre-trained generative deep learning model includes: When the input data is a sketch, high-quality renderings are generated using image recognition and deep learning technologies. When the input data is a text description, natural language processing technology is used to parse the semantics and generate corresponding visual design images. When the input data is an image, style transfer technology is used to convert the input image into a specified design style.

2. The urban rail transit industrial design method according to claim 1, characterized in that, The training method for the pre-trained generative deep learning model includes: Obtain an industrial design training dataset for urban rail transit, which includes renderings, 3D models, signage designs, and project information. The training dataset is cleaned, labeled, and augmented. The generative deep learning model is trained in a distributed manner using the processed training dataset; The trained model is evaluated, tested, and deployed.

3. The urban rail transit industrial design method according to claim 1, characterized in that, The multimodal post-processing optimization of the generated design image includes detail enhancement, color correction, and material optimization.

4. The urban rail transit industrial design method according to claim 1, characterized in that, The distributed training of the generative deep learning model adopts a multi-node parallel computing architecture, which is used to break down complex computing tasks into multiple sub-tasks and process them in parallel using multiple computing nodes.

5. The urban rail transit industrial design method according to claim 1, characterized in that, The method further includes iteratively optimizing the generative deep learning model based on historical design data and actual project feedback.

6. An industrial design platform for urban rail transit, characterized in that, include: A data input module is used to acquire multimodal design input data, wherein the multimodal design input data includes at least one of sketches, text descriptions, and images; The generative deep learning model module is pre-trained with a generative deep learning model, which is used to receive the multimodal design input data and generate design images; The post-processing optimization module is used to perform multimodal post-processing optimization on the generated design image to obtain the final design result; The generative deep learning model module includes: The sketch generation unit is used to generate high-quality renderings based on the input sketch using a generative adversarial network. The text generation unit is used to generate visual design images based on natural language descriptions through semantic parsing. The style transfer unit is used to convert an input image into a specified design style using style transfer techniques.

7. The urban rail transit industrial design platform according to claim 6, characterized in that, The style transfer unit supports template selection for multiple design styles and automatically adjusts colors, materials, and layout elements to adapt to the target style.

8. The urban rail transit industrial design platform according to claim 6, characterized in that, It also includes an industrial design database module for storing and managing multimodal design data, providing data support for model training.

9. An electronic device, characterized in that, include: Memory, used to store computer programs / instructions; A processor for executing the computer program / instructions to implement the steps of the method according to any one of claims 1-5.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When executed by a processor, the computer program implements the steps of the method described in any one of claims 1-5.

Images