Interactive footwear design and platemaking optimization system based on AI

Abstract

The invention discloses an interactive footwear design and platemaking optimization system based on AI, which relates to the technical field of digital interaction and comprises an auxiliary recommendation module, an AI interactive design module, a platemaking simulation module and a parameter optimization module. The auxiliary recommendation module is used for performing feature extraction processing on the shoe design related data by using a machine learning method to obtain recommended shoe design parameters; the AI interactive design module is used for generating a design reference drawing of the shoes, performing visual display and contour extraction on the design reference drawing, generating designed shoes with corresponding shoe sizes according to the three-dimensional shoe style data and the shoe size data, and performing local parameter modification according to the try-on feedback data; the plate-making simulation module is used for splitting the designed shoes, determining sewing procedures, outputting 3D printing parameters and driving the corresponding production line models to execute simulation production tasks to perform plate-making simulation on the obtained simulation shoe models, the design cost can be reduced, and the design quality and efficiency of the shoe products can be improved.

Description

Technical Field

[0001] This application relates to the field of digital interaction technologies, and particularly to an AI-based interactive footwear design and pattern-making optimization system. Background Art

[0002] The traditional way of footwear styling design by hand-drawing is difficult to meet the requirements of the rapid change of fashion trends, the growth of design information, and collaborative design. Moreover, the design mode relying on a single graphics or image simulation technology has obvious limitations in terms of design efficiency. In addition, as an important link in the design and production of the footwear industry, footwear pattern-making technology and efficiency directly affect the final shaping of footwear and even the success or failure of footwear enterprises' production. Most of the current footwear auxiliary design systems, footwear recommendation systems, and footwear pattern-making systems on the market are independent of each other and lack effective integration. Among them, the footwear auxiliary design system only focuses on facilitating the process simplification for designers when designing footwear, and cannot provide inspiration and ideas for designers. The footwear recommendation system is more oriented to consumers and cannot accurately provide designers with a time-sensitive fashion trend prediction and element recommendation function. In addition, the footwear pattern-making system cannot timely simulate and adjust the pattern-making parameters, which prolongs the production cycle and may affect the quality of the finished product. Summary of the Invention

[0003] In view of this, this application provides an AI-based interactive footwear design and pattern-making optimization system, which solves the problem that the current footwear auxiliary design systems, footwear recommendation systems, and footwear pattern-making systems on the market lack effective integration. And the existing footwear auxiliary design systems only focus on facilitating the process simplification for designers when designing footwear, and cannot provide inspiration and ideas for designers. And the footwear recommendation system is more oriented to consumers and cannot accurately provide designers with a time-sensitive fashion trend prediction and element recommendation function. And the footwear pattern-making system cannot timely simulate and adjust the pattern-making parameters, which prolongs the production cycle and may affect the quality of the finished product.

[0004] To achieve the above object, the present invention provides the following technical solution: An AI-based interactive footwear design and pattern-making optimization system is realized, which effectively integrates an auxiliary recommendation module, an AI interactive design module, a pattern-making simulation module, and a parameter optimization module. It can achieve the rapid prediction of footwear fashion trends and element recommendation, as well as generate reference diagrams for shoe design and extract sketch contours, providing inspiration and a creative basis for designers. And it can perform parameter interactive modification on a virtual footwear model for footwear design tasks. It can also decompose a virtual footwear model and perform simulation control on the pattern-making process for footwear pattern-making tasks.

[0005] Among them, the auxiliary recommendation module obtains data related to footwear design on the online network platform through web crawling and data cleaning, and uses machine learning methods to perform feature extraction and feature screening on the data related to footwear design according to the trigger instruction signal of the user, obtaining recommended footwear design parameters, and sending the footwear design parameters to the AI interactive design module. The footwear design parameters include popular trend keyword groups and popular element matrices; The AI interactive design module calls the popular trend keyword groups, popular element matrices and material parameter libraries output by the auxiliary recommendation module through data call functions and logical execution instructions, maps the text to the latent space, generates an initial image by the StyleGAN3 generator, and then corrects the problem of insufficient sole ground contact area through the FootNet model, outputting multiple corresponding design reference diagrams.

[0006] The AI interactive design module is also used for visual display and contour extraction of the design reference diagrams, facilitating the provision of inspiration and creative basis for designers. After the user optimizes the sketch on SolidWorks based on the reference diagram and converts the sketch into a 3D model, the design sketch data and 3D shoe model data are uploaded to the AI interactive design module through the human-computer interaction interface. It generates design shoes corresponding to the shoe size according to the 3D shoe model data and the shoe size data pre-stored in the shoe size database, and performs local parameter modification according to the fitting feedback data sent by the parameter optimization module to generate new design shoes; The plate-making simulation module is connected to the AI interactive design module. The plate-making simulation module calls the 3D model data of the design shoes output by the AI interactive design module through data call functions and logical execution instructions, disassembles the designed shoes into pieces using the piecewise model, and determines the sewing process according to the optimization model, facilitating the use as the data basis for subsequent simulation. According to the 3D model size parameters, the printing materials of the upper and sole selected by the user, and the printing equipment, the printing layer thickness parameters and filling rate of the upper and sole are output. The trial model is printed on a 3D printer according to the exported data. After the trial wearer tries on the shoes, the fitting parameters are uploaded to the AI interactive design module through the parameter optimization module to modify the 3D model data, and then the plate-making simulation is carried out using the modified plate-making parameters. It is also used to form a simulated production line according to the equipment model in the plate-making model database, adjust the model drive parameters according to the process flow of plate-making with different materials, and drive the corresponding production line model to execute the simulated production task according to the model drive parameters to perform plate-making simulation on the obtained simulated footwear model, display the simulation control effect, and store and update the management of the plate-making data of different shoes.

[0007] Furthermore, the auxiliary recommendation module includes a data preprocessing unit and a feature extraction unit; The data preprocessing unit is used to clean and label the structured and unstructured data collected through the data interface, slice and store the cleaned data according to the data type, and establish a time series index and a spatial index; The feature extraction unit is used to construct a multi-modal fusion feature extraction model. The multi-modal fusion feature extraction model includes a visual feature extraction module, a semantic feature extraction module, a time series feature extraction module, and a prediction element output module. The visual feature extraction module is used to extract the visual features of the shoe design image data by using a visual feature extraction method based on ResNet-152 and a channel attention mechanism. The semantic feature extraction module uses the BERT model and the TF-IDF weighting method to process the text data related to shoe descriptions and construct semantic features. The time series feature extraction module uses the Prophet decomposition method to extract the time series features of shoes such as trend terms, seasonal terms, and holiday effects. The prediction element output module is connected to the visual feature extraction module, the semantic feature extraction module, and the time series feature extraction module, and is used to output the shoe fashion element features.

[0008] Furthermore, the AI interaction design module includes an image generation unit. The image generation unit is used to generate a design reference diagram of shoes by using an image generation model. The image generation model includes a sampling module, a StyleGAN3 generator module, a conditional control module, and a constraint module; The sampling module is used to sample a 512-dimensional vector z from a random Gaussian distribution and input the vector z into the StyleGAN3 generator module. The StyleGAN3 generator module obtains a 16*512-dimensional latent vector w through a mapping network. The StyleGAN3 generator module is also used to receive the vector w* fed back by the conditional control module and input the vector w* into the generator G, then the edited image M is obtained; A multi-attribute classifier is used to score the edited image M on the semantic attributes to be edited, so as to guide the model training based on the backpropagation algorithm, and a FootNet discriminator is introduced to detect the symmetry of shoelace holes and the rationality of the sole contact area, and the physical constraint Diffusion model is added with mass conservation; The conditional control module is used to input the given semantic attribute vector c and the latent vector w into a latent vector editing network constructed by a neural network to obtain the vector w*. The vector w* contains the semantic information that needs to be changed; The constraint module is used to add a mass conservation constraint of shoe weight by using the Diffusion model.

[0009] Even further, the AI interaction design module includes an image contour extraction unit and an optimization design unit; The image contour extraction module is used to obtain the design reference diagram, perform non-uniform illumination correction on the design reference diagram, then extract high-frequency textures and low-frequency structures using FFT filtering, perform artifact detection, and use the canny edge detection method to detect the edge contours of the selected design reference diagram, strengthen specific angle lines through a directionally controllable filter, and output a binary edge map and a structural line classification map; The optimization design unit is used to retrieve the design sketch data and 3D shoe model data uploaded by the designer, automatically adjust the 3D shoe model data according to the preset foot size information corresponding to each shoe size, generate shoe product data corresponding to different shoe sizes, and modify local parameters according to the trial wear feedback data to generate new designed shoe product data.

[0010] Furthermore, the AI interaction design module further includes a visualization display module, which is used to visually display the design sketch data, 3D shoe model data, and interaction parameters.

[0011] Furthermore, the plate making simulation module includes a decomposition unit, a printing parameter management unit, and a plate making simulation module; The decomposition unit is used to obtain the 3D shoe product data and the corresponding material properties, identify the geometric features of the shoe through MeshCNN, detect the upper, sole, and decorative strip components, and obtain the generated cutting boundary generated by the Blender geometric node system to obtain a 2D cutting vector diagram and a sewing seam relationship matrix. It is also used to output an optimized process flow according to the optimization model, objective function, and constraint conditions; The printing parameter management unit is used to select the printing materials for the upper and sole and the printing equipment according to the user interaction panel, and output the printing layer thickness parameters and filling rate for the upper and sole; The plate making simulation module is used to form a simulated production line according to the equipment model in the plate making model database, adjust the model drive parameters according to the process parameters and process flow, and drive the corresponding production line model to execute the simulated production task according to the model drive parameters to perform plate making simulation on the obtained simulated shoe model, and display the simulation control data. It is also used to store and update the management of the plate making data of different shoe products.

[0012] Further, the parameter optimization module is used to obtain the sensor and user survey form parameters through the data interface, obtain the local pressure peak value and foot sliding displacement of the shoe product, and perform parameter optimization through the backpropagation algorithm: , where E is the comfort error function, η is the learning rate η = 0.01, α is the momentum term, α = 0.9, and the output is the adjustment of the shoe last width and the correction of the midsole hardness gradient.

[0013] It can be seen from the above technical solutions that the advantages of the present invention are: In this application, by integrating the auxiliary recommendation module, the AI interaction design module, and the plate-making simulation module, and through multiple AI computing models for data learning and processing, designers can quickly obtain current popular trend information and design reference diagram information. Moreover, the figure contour extraction technology greatly shortens the design cycle of designers, improves the efficiency and accuracy of footwear design, and avoids the shortcomings that the existing footwear auxiliary design system cannot provide inspiration and ideas for designers, as well as the problem that the footwear recommendation system cannot accurately provide designers with timely popular trend predictions and element recommendations. At the same time, through the plate-making simulation module to simulate and regulate the plate-making parameters, the production cycle is shortened and the production cost is reduced. Brief Description of the Drawings

[0014] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application.

[0015] Figure 1 It is a schematic diagram of the composition structure of this application.

[0016] Figure 2 It is a schematic diagram of the steps of the popular trend and element prediction process of this application.

[0017] Figure 3 It is a schematic diagram of the steps of the design reference diagram generation and contour extraction process of this application.

[0018] Figure 4 It is a schematic diagram of the composition structure of the plate-making simulation module of this application. Detailed Embodiment

[0019] To make the purpose, technical solution, and advantages of this application clearer and more understandable, the following further details this application in combination with the embodiments and the drawings. Herein, the schematic embodiments of this application and their descriptions are used to explain this application, but do not serve as a limitation to this application.

[0020] Refer to Figures 1 to 4 the illustrated embodiment, the AI-based interactive footwear design and plate-making optimization system is developed based on a custom system framework. This system specifically includes: an auxiliary recommendation module, an AI interaction design module, a plate-making simulation module, and a parameter optimization module.

[0021] Among them, the auxiliary recommendation module obtains footwear design-related data on the online network platform through data crawling and data cleaning, and obtains footwear design parameters based on the auxiliary recommendation model and the storage database. The footwear design parameters include popular trend keyword groups and popular element matrices.

[0022] The auxiliary recommendation module, as a fashion trend element extractor, is mainly used to obtain data related to footwear design, and perform feature extraction and feature screening on the footwear design-related data using machine learning methods based on the trigger instruction signal of the user, and provide footwear design parameters for the AI interaction design module. The AI interaction design module is mainly used to generate a design reference diagram of the shoe using deep learning methods, and perform visual display and contour extraction on the design reference diagram. It is convenient to provide inspiration and a creative basis for designers. After the user optimizes the sketch on SolidWorks based on this reference diagram and converts the sketch into a 3D model, the design sketch data and 3D shoe model data are uploaded to the AI interaction design module through the human-computer interaction interface. It is also used to generate the designed shoe product corresponding to the shoe size based on the 3D shoe model data and the shoe size data pre-stored in the shoe size database, and perform local parameter modification according to the trial-wearing feedback data sent by the parameter optimization module to generate a new designed shoe product. In this embodiment, the hardware for system function implementation includes a data acquisition server (dual-way Intel Xeon Silver 4310 processor, 128GB DDR4 ECC memory, 10Gbps network card), a machine learning node (NVIDIA A100 GPU cluster), a feature database (distributed NAS storage system), and an edge computing gateway, as well as a deep learning workstation (2×NVIDIA RTX 6000 Ada GPU), a visual interaction terminal (32-inch 4K touch screen), an intelligent fabric printer, and a trial-wearing feedback sensing system (intelligent insole with 32-point pressure, Bluetooth 5.3 transmission). The data acquisition server is used to crawl footwear design data from e-commerce platforms / social media / patent libraries in real time. The machine learning node is used to run feature extraction algorithms and related computational models. The feature database is used to store structured design features and unstructured data. The deep learning workstation is used to run the StyleGAN3 generative adversarial network to achieve parallel generation of multiple design diagrams (4-6 concept diagrams per second). The visual interaction terminal supports designers' gesture operations. The trial-wearing feedback sensing system is used to monitor the wearing comfort data in real time.

[0023] Data acquisition terminal, auxiliary recommendation module server cluster, AI interaction design workstation, human-computer interaction terminal, parameter optimization console, digital production equipment, trial-wearing feedback sensor network In this embodiment, a large amount of data is required to support the model for operation. When predicting the popular styles of shoes, the system crawls structured data such as e-commerce platform sales data (SKU attributes, sales volume time series), Pantone quarterly color report (RGB color value + popularity index), and material market quotes (leather / synthetic material futures prices) according to the keyword fields and data volume set by the administrator. For example, data volume: at least 3 years of historical data, keyword fields: SKU attributes, sales volume, price, timestamp; and obtain unstructured data: Instagram / TikTok pictures (Hashtag-related shoe styles), fashion show videos (action recognition to extract design elements), Google search trends (keywords "sports shoes", "year", and sentiment polarity tags). For unstructured data, the corresponding pictures and texts can be crawled through the API. When cleaning the data, the U2-Net network model is used to accurately segment and remove the background from the image data, and key point annotation is performed. For example, 36 feature points such as the toe, heel, and shoelaces are annotated; for the text, the cleaner(text) function is used to remove non-shoe-related descriptions. When storing the data, the MongoDB sharded cluster method is adopted, and sharding is performed according to the data type (sales, images, texts) to establish a time series index (date field) and a spatial index (image feature vector) for subsequent data applications.

[0024] Specifically, the auxiliary recommendation module includes a data preprocessing unit and a feature extraction unit; the data preprocessing unit is used to clean and annotate the structured and unstructured data collected through the data interface, and store the cleaned data by sharding according to the data type, and establish a time series index and a spatial index.

[0025] The feature extraction unit is mainly used to construct a multi-modal fusion feature extraction model. The multi-modal fusion feature extraction model includes a visual feature extraction module, a semantic feature extraction module, a time series feature extraction module, and a prediction element output module. The visual feature extraction module uses a visual feature extraction method based on ResNet-152 and channel attention mechanism to extract the visual features of shoe design image data from the shoe picture dataset. The semantic feature extraction module uses the BERT model and the TF-IDF weighting method to process the text data related to shoe descriptions, and constructs a 2000 shoe design term dictionary (such as "mono yarn", "EVA midsole"). To construct semantic features, the time series feature extraction module uses the Prophet decomposition method to extract the time series features of shoes such as trend terms, seasonal terms, and holiday effects. The prediction element output module is connected to the visual feature extraction module, the semantic feature extraction module, and the time series feature extraction module, and the prediction element output module is used to output the popular element features of shoes.

[0026] In this embodiment, each model is trained on the corresponding training server, and the loss function adopted by the multi-modal fusion feature extraction model is the LOSS function. During application, the user inputs control instructions through interactive hardware such as a mouse and keyboard, and the computer logic execution module realizes data calling and processing. First, the input data includes: Images: 3,000 latest trendy shoes on Instagram, Text: The search volume of "thick sole design" has increased by 150% monthly, Time series: The price of EVA material has decreased by 8% month-on-month. Model output: Prediction elements: Thick sole coefficient: 0.78 (confidence level 92%), Probability of using fluorescent color: 0.65, Demand for mesh material: Up 23%.

[0027] Specifically, the AI interaction design module includes an image generation unit. The image generation unit is used to generate a design reference diagram of shoes using an image generation model. The image generation model includes a sampling module, a StyleGAN3 generator module, a conditional control module, and a constraint module. The sampling module is used to sample a 512-dimensional vector z from a random Gaussian distribution and input the vector z into the StyleGAN3 generator module. The StyleGAN3 generator module obtains a 16*512-dimensional latent vector w through a mapping network. The StyleGAN3 generator module is also used to receive the vector w* fed back by the conditional control module and input the vector w* into the generator G, then the edited image M is obtained. A multi-attribute classifier is used to score the edited image M on the semantic attributes to be edited, so as to guide the model training based on the backpropagation algorithm, and the FootNet discriminator is introduced to detect the symmetry of shoelace holes and the rationality of the sole grounding area, and the physical constraint Diffusion model adds mass conservation; The conditional control module is used to input the given semantic attribute vector c and the latent vector w into the latent vector editing network constructed by a neural network to obtain the vector w*, and the vector w* contains the semantic information that needs to be changed. The constraint module is used to add the mass conservation constraint of the shoe weight using the Diffusion model. The semantic attribute vector includes the above-mentioned trend keywords.

[0028] In this embodiment, the reference image dataset and popular elements in the image database are obtained. According to the trend prediction result of "outdoor functional style + reflective strips + Vibram sole", the generation process is as follows: The CLIP model maps the text to the latent space Z_clip; The StyleGAN3 generator generates an initial image (512×512); The diffusion model adds the reflective strip material (Substance texture mapping); The FootNet corrects the problem of insufficient sole grounding area, and then the model outputs: 3 candidate designs.

[0029] The AI interaction design module includes an image contour extraction unit and an optimization design unit; the image contour extraction module is used to obtain a design reference diagram, perform non-uniform illumination correction on the design reference diagram, then use FFT filtering to extract high-frequency textures and low-frequency structures, and perform artifact detection, and use the canny edge detection method to detect the edge contour of the selected design reference diagram, strengthen specific angle lines through a directionally controllable filter, and output a binary edge map and a structure line classification map, and the binary edge map and the structure line classification map include contours, stitches, decorations, etc.; the optimization design unit is used to retrieve the design sketch data and 3D shoe model data uploaded by the designer, and automatically adjust the 3D shoe model data according to the preset foot size information corresponding to each shoe size, generate shoe product data corresponding to different shoe sizes, and perform local parameter modification according to the fitting feedback data to generate new designed shoe product data.

[0030] In this embodiment, before extracting the image contour, it is necessary to preprocess the image. The problem of uneven brightness in the generated image can be solved through non-uniform illumination correction (such as local overexposure generated by GAN). Use FFT filtering to extract high-frequency textures and low-frequency structures. The high-frequency textures include shoelace braided patterns, etc., and the low-frequency structures include contour lines. Use a UNet-based artifact detection network for detection. When training this detection model, the training data includes 10,000 pairs of generated images and manually annotated artifact regions to ensure the accuracy of the model. After obtaining the binary edge map and the structure line classification map, it is also necessary to optimize the structure line, perform vector line extraction (sub-pixel edge localization, Bezier curve fitting), and semantic perception optimization (structure line classification optimization, physical constraint injection), output a vectorized map, and use DXF output adaptation.

[0031] The AI interaction design module also includes a visualization display module, which is used to visually display the design sketch data, 3D shoe model data, and interaction parameters.

[0032] The plate-making simulation module is connected to the AI interaction design module. The plate-making simulation module is used to disassemble the designed shoe product into pieces, determine the sewing process, output 3D printing parameters for fitting and printing, and perform plate-making simulation according to the plate-making parameters modified after fitting. It is also used to form a simulated production line according to the equipment model in the plate-making model database, adjust the model drive parameters according to the process flow of plate-making with different materials, and drive the corresponding production line model to execute the simulated production task according to the model drive parameters to perform plate-making simulation on the obtained simulated shoe model and display the simulation control effect. It is also used to store and update the management of the plate-making data of different shoe products. Specifically, the plate-making simulation module includes a decomposition unit, a printing parameter management unit, and a plate-making simulation module; the decomposition unit is used to obtain three-dimensional shoe product data and corresponding material properties, identify geometric features of the shoe through MeshCNN, detect components such as the upper, sole, and decorative strip, and obtain the generated cutting boundary generated by the Blender geometric node system, resulting in a two-dimensional cutting vector diagram and a sewing seam relationship matrix. It is also used to output an optimized process flow according to the optimization model, objective function, and constraint conditions; the printing parameter management unit is used to select the printing materials for the upper and sole and the printing equipment according to the user interaction panel, and output the printing layer thickness parameters and filling rate of the upper and sole; the plate-making simulation module is used to form a simulated production line according to the equipment model in the plate-making model database, adjust the model drive parameters according to the process parameters and process flow, and drive the corresponding production line model to execute a simulated production task according to the model drive parameters to perform plate-making simulation on the obtained simulated shoe model, and display the simulation control data. It is also used to store and update the management of the plate-making data of different shoe products.

[0033] In this embodiment, the input of the decomposition model is a shoe product 3D model (STL format) and material properties (thickness, elastic modulus). The model includes a geometric feature recognition part: detecting components such as the upper / sole / decorative strip based on MeshCNN, a manufacturability constraint part: minimum hemming width (≥3mm), stitch position spacing (5±0.5mm), and an automatic disassembly part: generating the cutting boundary (parameterized control) using the Blender geometric node system. The output is a two-dimensional cutting vector diagram (DXF format) and a sewing seam relationship matrix (adjacency list structure). When planning the sewing sequence, the optimization model uses a mixed integer programming (MIP) combined with a genetic algorithm, with the objective function: minimizing the number of thread changes and the number of sewing tool switches. The constraint condition: the lining is sewn before the outer layer and the decorative parts are installed last. Solver: Gurobi Optimizer, visual output: a Gantt chart showing the process flow (accurate to 0.5 minutes).

[0034] In this embodiment, the printing material for the upper is selected as PA12 material, and the printing material for the sole is selected as EPU 41 elastomer. Layer thickness: 0.1mm (upper) / 0.3mm (sole).

[0035] The simulation equipment model includes a laser cutting machine with a power of 2000W and a positioning accuracy of ±0.1mm; an automatic sewing machine with a needle speed of 5000rpm, supporting 6-axis stitching; a hot pressing machine with a temperature control range of 50-200°C ±1°C.

[0036] The parameter optimization module is used to obtain the parameters of the sensor and the user survey form through the data interface, obtain the local pressure peak value and foot sliding displacement of the shoe product, and perform parameter optimization through the backpropagation algorithm: , where E is the comfort error function, i.e., the mean square error of pressure distribution, η is the learning rate, η = 0.01, α is the momentum term, α = 0.9, and the outputs are the adjustment of the last width and the correction of the midsole hardness gradient.

[0037] In this embodiment, the sensors are used to collect data by means of a pressure-sensitive film (Tekscan F-Scan, sampling at 100Hz) and an IMU sensor (Xsens MTw Awinda, capturing gait data) arranged inside the shoe, and the metrics are the local pressure peak and the foot sliding displacement.

[0038] The artificial intelligence technology empowers the information interaction in the design process, enabling designers to accurately and quickly obtain inspirations and opinions from multiple parties, and improving the work efficiency of designers.

[0039] In some cases, components of a device and / or system may be configured to perform functions such that the components are physically configured and constructed (using hardware and / or software) to enable such performance. In other examples, components of a device and / or system may be arranged to be suitable for, capable of, or adapted to perform functions when operating in a particular manner. A method may include one or more operations, functions, or actions illustrated by one or more of the boxes. Although the boxes are illustrated in sequence, these boxes may also be performed in parallel and / or in a different order than described herein. Additionally, the various boxes may be combined into fewer boxes, divided into additional boxes, and / or removed based on the desired implementation.

[0040] The above are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the embodiments of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. An AI-based interactive footwear design and pattern-making optimization system, characterized in that It includes: An auxiliary recommendation module, an AI interaction design module, a plate-making simulation module, and a parameter optimization module; The auxiliary recommendation module is used to obtain data related to footwear design, and use machine learning methods to perform feature extraction and feature screening on the footwear design-related data according to the trigger instruction signal of the user, obtain recommended footwear design parameters, and send the footwear design parameters to the AI interaction design module; The AI interaction design module is used to generate a design reference diagram of the shoe using deep learning methods, perform visual display and contour extraction on the design reference diagram, and obtain and store the design sketch data and three-dimensional shoe model data uploaded by the designer through the human-computer interaction interface. It is also used to generate a designed shoe product corresponding to the shoe size according to the three-dimensional shoe model data and shoe size data, and perform local parameter modification according to the trial-wearing feedback data sent by the parameter optimization module to generate a new designed shoe product; The plate-making simulation module is connected to the AI interaction design module. The plate-making simulation module is used to disassemble the designed shoe product into pieces, determine the sewing process, output 3D printing parameters for trial-wearing printing, and perform plate-making simulation according to the plate-making parameters after trial-wearing modification. It is also used to form a simulated production line according to the equipment model in the plate-making model database, adjust the model drive parameters according to the process flow of plate-making with different materials, and drive the corresponding production line model to execute the simulated production task according to the model drive parameters to perform plate-making simulation on the obtained simulated footwear model and display the simulation control effect. It is also used to store and update the management of the plate-making data of different shoe products.

2. The AI-based interactive footwear design and pattern-making optimization system according to claim 1, wherein The auxiliary recommendation module includes a data preprocessing unit and a feature extraction unit; The data preprocessing unit is used to clean and label the structured and unstructured data collected through the data interface, store the cleaned data in slices according to the data type, and establish a time series index and a spatial index; The feature extraction unit is used to construct a multi-modal fusion feature extraction model. The multi-modal fusion feature extraction model includes a visual feature extraction module, a semantic feature extraction module, a time series feature extraction module, and a prediction factor output module. The visual feature extraction module is used to extract the visual features of the footwear design image data using a visual feature extraction method based on ResNet-152 and a channel attention mechanism. The semantic feature extraction module uses the BERT model and the TF-IDF weighting method to process the text data related to the footwear description to construct semantic features. The time series feature extraction module uses the Prophet decomposition method to extract the time series features of the footwear, including the trend term, the seasonal term, and the holiday effect. The prediction factor output module is connected to the visual feature extraction module, the semantic feature extraction module, and the time series feature extraction module. The prediction factor output module is used to output the footwear fashion factor features.

3. The AI-based interactive footwear design and pattern-making optimization system according to claim 1, wherein The AI interaction design module includes an image generation unit. The image generation unit is used to generate a design reference diagram of the shoe using an image generation model. The image generation model includes a sampling module, a StyleGAN3 generator module, a conditional control module, and a constraint module; The sampling module is used to sample a 512-dimensional vector z from a random Gaussian distribution and input the vector z into the StyleGAN3 generator module. The StyleGAN3 generator module obtains a 16 * 512-dimensional latent vector w through the mapping network. The StyleGAN3 generator module is also used to receive the vector w* fed back by the conditional control module and input the vector w* into the generator G, then the edited image M is obtained. A multi-attribute classifier is used to score the edited image M on the semantic attributes to be edited, so as to guide the model training based on the backpropagation algorithm, and the FootNet discriminator is introduced to detect the symmetry of shoelace holes and the rationality of the sole ground contact area, and the physical constraint Diffusion model adds mass conservation; The conditional control module is used to input the given semantic attribute vector c and the latent vector w into the latent vector editing network constructed by a neural network to obtain the vector w*, and the vector w* contains the semantic information that needs to be changed; The constraint module is used to add the mass conservation constraint of the shoe weight by using the Diffusion model.

4. The AI-based interactive footwear design and pattern-making optimization system according to claim 3, wherein The AI interaction design module includes an image contour extraction unit and an optimization design unit; The image contour extraction module is used to obtain the design reference diagram, perform non-uniform illumination correction on the design reference diagram, then use FFT filtering to extract high-frequency textures and low-frequency structures, perform artifact detection, and use the canny edge detection method to detect the edge contour of the selected design reference diagram, strengthen the lines at specific angles through a directionally controllable filter, and output a binary edge map and a structural line classification map; The optimization design unit is used to retrieve the design sketch data and three-dimensional shoe model data uploaded by the designer, and automatically adjust the three-dimensional shoe model data according to the preset foot size information corresponding to each shoe size, generate shoe product data corresponding to different shoe sizes, and modify the local parameters according to the trial wear feedback data to generate new designed shoe product data.

5. The AI-based interactive footwear design and pattern-making optimization system according to claim 4, wherein The AI interaction design module also includes a visualization display module, and the visualization display module is used to visually display the design sketch data, three-dimensional shoe model data, and interaction parameters.

6. The AI-based interactive footwear design and pattern-making optimization system according to claim 4, wherein The plate-making simulation module includes a decomposition unit, a printing parameter management unit, and a plate-making simulation module; The decomposition unit is used to obtain the three-dimensional shoe product data and the corresponding material attributes, identify the geometric features of the shoe through MeshCNN, detect the upper, sole, and decorative strip components, and obtain the generated cut piece boundaries generated by the Blender geometric node system, to obtain a two-dimensional cut piece vector diagram and a sewing seam relationship matrix, and is also used to output an optimized process flow according to the optimization model, objective function, and constraint conditions; The printing parameter management unit is used to select the printing materials for the upper and sole and the printing equipment according to the user interaction panel, and output the printing layer thickness parameters and filling rate of the upper and sole; The plate-making simulation module is used to form a simulated production line according to the equipment models in the plate-making model database, adjust the model driving parameters according to the process parameters and process flow, drive the corresponding production line model to execute the simulated production task based on the model driving parameters to perform plate-making simulation on the obtained simulated footwear model, and display the simulation control data. It is also used to store, update and manage the plate-making data of different shoe products.

7. The AI-based interactive footwear design and pattern-making optimization system according to claim 1, wherein The parameter optimization module is used to obtain the parameters of the sensor and the user survey form through the data interface, obtain the local pressure peak value and foot sliding displacement of the shoe product, and optimize the parameters through the backpropagation algorithm: , where E is the comfort error function, η is the learning rate η = 0.01, α is the momentum term, α = 0.9, and the outputs are the adjustment of the shoe last width and the correction of the midsole hardness gradient.

Images