Fabric parameter generation method, fabric generation method and fabric parameter generation device

Abstract

The invention relates to a fabric parameter generation method, a fabric generation method and a fabric parameter generation device.The fabric parameter generation method comprises the steps that user demand information is obtained, and the user demand information comprises text description and / or image data of fabric; feature extraction is carried out on the user demand information to obtain demand features, and the demand features comprise a first feature; target fabric parameters are obtained based on the demand characteristics, and the target fabric parameters comprise target fabric construction parameters, target fabric raw material parameters and target fabric color parameters, so that the problem of high cost for realizing innovation of fabric design is solved, and the cost for realizing innovation of fabric design is reduced.

Description

Technical Field

[0001] This application relates to the technical field of textiles, and in particular to a method for generating fabric parameters, a method for generating fabric, and a device for generating fabric parameters. Background Art

[0002] In the field of knitting design, fabric design usually adopts two main methods: forward design and reverse design. Forward design is concept-based, following established design processes and methods to gradually achieve fabric design goals, covering aspects such as requirement analysis, material selection, organizational structure design, and process parameter determination. However, forward design often relies on designers' profound professional knowledge and rich experience, with extremely high requirements for design and innovation capabilities. The entire process from concept to finished product requires a large amount of time and resources, resulting in a long design cycle and high R & D costs. Reverse design analyzes and studies existing fabric products, deduces their design principles and process parameters, and then realizes similar or improved product designs, mainly involving steps such as product analysis, disassembly and measurement, organizational structure analysis, and design improvement. However, reverse design is often just an imitation and improvement of existing products, and it is difficult to achieve true innovation. Since it is a replication of existing products, it may inherit the defects of the original products and is difficult to reach a higher quality level.

[0003] Currently, there is also a method of obtaining three-dimensional data of a fabric, reconstructing the obtained three-dimensional data into a three-dimensional model, and performing innovative design on the basis of the reconstructed three-dimensional model to improve the innovation degree of fabric design. However, this method has a high dependence on the three-dimensional model, and the cost of obtaining the three-dimensional data of the fabric is high, and the cost of realizing innovation is high. The above methods all have limitations in innovation and cannot directly obtain a fabric design that meets requirements based on users' subjective needs.

[0004] Regarding the problem of high innovation cost in fabric design in related technologies, no effective solution has been proposed yet. Summary of the Invention

[0005] In this embodiment, a method for generating fabric parameters, a method for generating fabric, and a device for generating fabric parameters are provided to solve the problem of high innovation cost in fabric design in related technologies.

[0006] In a first aspect, a method for generating fabric parameters is provided in this embodiment, including: Obtaining user requirement information, where the user requirement information includes text descriptions and / or image data of the fabric; Performing feature extraction on the user requirement information to obtain requirement features, where the requirement features include first features; Obtain target fabric parameters based on the demand characteristics, where the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters.

[0007] In some embodiments, the obtaining of the target fabric parameters based on the demand characteristics includes: Input the first feature into a pre-trained fabric parameter model for fabric parameter extraction to obtain first target fabric parameters, where the first feature includes the visual feature of the fabric. Use the first target fabric parameters as the target fabric parameters, where the first target fabric parameters include first fabric structure parameters, first fabric raw material parameters, and first fabric color parameters.

[0008] In some embodiments, after inputting the first feature into a pre-trained fabric parameter model for fabric parameter extraction to obtain first target fabric parameters, the demand characteristics include a second feature, and the obtaining of the target fabric parameters based on the demand characteristics includes: Obtain second target fabric parameters according to the second feature; Combine the first target fabric parameters and the second target fabric parameters to obtain the target fabric parameters.

[0009] In some embodiments, when the second feature includes a fabric price feature, the target fabric parameters further include target fabric price parameters; when the second feature includes a fabric size feature, the target fabric parameters further include target fabric size parameters.

[0010] In some embodiments, the training process of the fabric parameter model includes: Construct a training data set, where the training data set includes the feature information and fabric parameters of the training fabric, where the feature information includes the visual feature of the training fabric; the fabric parameters include the fabric structure parameters, fabric raw material parameters, and fabric color parameters of the training fabric. Train an initial fabric parameter model based on the feature information of the training fabric and the fabric parameters of the training fabric to obtain the trained fabric parameter model.

[0011] In some embodiments, the training of the initial fabric parameter model based on the feature information of the training fabric and the fabric parameters of the training fabric to obtain the trained fabric parameter model includes: Train a first initial fabric parameter model based on the feature information of the training fabric and the fabric structure parameters to obtain a first fabric parameter model; Train the second initial fabric parameter model based on the feature information of the training fabric, the fabric raw material parameters, and the fabric color parameters to obtain the second fabric parameter model; Combine the first fabric parameter model and the second fabric parameter model to obtain the trained fabric parameter model.

[0012] In some embodiments, the training dataset further includes the fabric tissue diagram of the training fabric.

[0013] In some embodiments, the training of the first initial fabric parameter model based on the feature information of the training fabric and the fabric structure parameters to obtain the first fabric parameter model includes: Train the first model based on the feature information of the training fabric and the fabric tissue diagram to obtain the trained first model; Train the second model based on the fabric tissue diagram and the fabric structure parameters to obtain the trained second model; Combine the trained first model and the trained second model to obtain the first fabric parameter model.

[0014] In a second aspect, a fabric generation method is provided in this embodiment. Applying the fabric parameter generation method described in the first aspect above, it includes: According to the target fabric parameters, determine whether there is a corresponding recommended fabric in the preset fabric library; If so, obtain the corresponding recommended fabric according to the target fabric parameters; otherwise, generate a virtual fabric according to the target fabric parameters and store the virtual fabric in the virtual fabric library.

[0015] In a third aspect, a fabric parameter generation device is provided in this embodiment, including: an acquisition module, an extraction module, and a fabric parameter generation module, where, The acquisition module is used to acquire user requirement information, where the user requirement information includes the text description and / or image data of the fabric; The extraction module is used to extract the features of the user requirement information to obtain requirement features, where the requirement features include first features; The fabric parameter generation module is used to obtain target fabric parameters based on the requirement features, where the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters.

[0016] Fourthly, in this embodiment, an electronic device is provided, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the fabric parameter generation method described in the first aspect above is implemented.

[0017] Fifthly, in this embodiment, a storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the fabric parameter generation method described in the first aspect above is implemented.

[0018] Compared with the related art, in the fabric parameter generation method provided in this embodiment, by obtaining user demand information, where the user demand information includes text descriptions and / or image data of the fabric; extracting features from the user demand information to obtain demand features, where the demand features include first features; and obtaining target fabric parameters based on the demand features, where the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters, the problem of high innovation cost in fabric design is solved, and the innovation cost of fabric design is reduced.

[0019] Details of one or more embodiments of the present application are set forth in the following drawings and description, so that other features, objects, and advantages of the present application will become more comprehensible. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments and descriptions thereof of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings: Figure 1 is a hardware structure block diagram of a terminal of the fabric parameter generation method in this embodiment.

[0021] Figure 2 is a flowchart of the fabric parameter generation method in this embodiment.

[0022] Figure 3 is a flowchart of fabric parameter model training of the fabric parameter generation method in this embodiment.

[0023] Figure 4 is a flowchart of the first fabric parameter model training of the fabric parameter generation method in this embodiment.

[0024] Figure 5 is a flowchart of the fabric generation method in this embodiment.

[0025] Figure 6 is a preferred flowchart of the fabric generation method in this embodiment.

[0026] Figure 7It is a structural block diagram of the fabric parameter generation device of this embodiment. Detailed implementation manners

[0027] To understand the purpose, technical solution and advantages of this application more clearly, the following describes and explains this application in combination with the accompanying drawings and embodiments.

[0028] Unless otherwise defined, the technical terms or scientific terms involved in this application shall have the general meaning understood by those with ordinary skills in the technical field to which this application belongs. In this application, words such as "a", "one", "a kind of", "the", "these" and the like do not represent a limitation in quantity, and they can be singular or plural. The terms "including", "comprising", "having" and any variants thereof involved in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product or device including a series of steps or modules (units) is not limited to the listed steps or modules (units), but may include unlisted steps or modules (units), or may include other steps or modules (units) inherent in these processes, methods, products or devices. The terms "connected", "coupled" and the like involved in this application do not limit to physical or mechanical connections, but may include electrical connections, whether directly or indirectly. The "multiple" involved in this application means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, and B exists alone. Usually, the character " / " represents an "or" relationship between the associated objects before and after. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific sorting for the objects.

[0029] The method embodiment provided in this embodiment can be executed on a terminal, a computer or a similar computing device. For example, it runs on a terminal. Figure 1 It is a hardware structural block diagram of the terminal of the fabric parameter generation method of this embodiment. As Figure 1 shown, the terminal may include one or more ( Figure 1 only one is shown in Figure 1 the figure) processors 102 and a memory 104 for storing data. Among them, the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA. The above terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown in Figure 1The different configurations shown.

[0030] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the fabric parameter generation method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, implements the above method. The memory 104 may include a high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories can be connected to the terminal through a network. Examples of the above network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.

[0031] The transmission device 106 is used to receive or send data via a network. The above network includes a wireless network provided by a communication provider of the terminal. In one instance, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station and thus can communicate with the Internet. In one instance, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0032] In this embodiment, a fabric parameter generation method is provided. Figure 2 It is a flowchart of the fabric parameter generation method in this embodiment, as Figure 2 shown, and this process includes the following steps: Step S201, obtain user demand information, where the user demand information includes text descriptions and / or image data of the fabric.

[0033] Specifically, traditional fabric design highly depends on the professional experience of designers, and it is difficult for ordinary users to accurately convert subjective demands (such as "having a shiny feeling") into fabric parameters that can participate in production. How to enable ordinary users to directly obtain corresponding fabric parameters based on their demands so as to reduce the cost of fabric design innovation is an urgent problem to be solved.

[0034] In some embodiments, the user can directly input text descriptions. For example, a fabric that is soft and skin-friendly and suitable for making a retro-style dress, or a brightly colored fabric suitable for use on an indoor sofa.

[0035] In some embodiments, the user can upload image data as user requirement information. For example, by providing a picture of a banquet hall, fabric parameters suitable for use in the banquet hall can be obtained, or by providing a picture of a star's fashion, fabric parameters of a fabric similar to the fashion worn by the star can be obtained.

[0036] In some embodiments, the user can provide both a text description and image data as user requirement information.

[0037] Step S202: Extract features from the user requirement information to obtain requirement features, where the requirement features include first features.

[0038] Specifically, the first feature is a subjective attribute of the fabric directly perceived by the human sensory system, with non - quantifiable and descriptive characteristics. The first feature includes the visual features of the fabric, such as style features, including elegance, liveliness, fashion, etc.; scene features, including outdoor occasions, home occasions, formal occasions, etc.; appearance features, including drapability, glossiness, etc., where the glossiness can include mercerized luster, high gloss. In some embodiments, the first feature also includes the tactile features of the fabric, such as softness. Among them, the first feature cannot be directly mapped to fabric parameters.

[0039] In some embodiments, the requirement information further includes a second feature. Specifically, the second feature is an objectively measurable physical and chemical attribute of the fabric or other objective attributes, such as fabric physical features, fabric chemical features, fabric size features, fabric price features, etc. There is an explicit / implicit association between the second feature and the fabric parameters.

[0040] Specifically, the description of certain fabric features can be either subjective and non - quantifiable or objective and quantifiable. That is, some fabric features have a physical measurement basis and can be described either quantitatively or non - quantitatively. Depending on the description method, fabric features can be classified into different feature categories. For example, for the drapability of the fabric, it can be measured by a drapability tester. If the extracted requirement feature is "has drapability", the requirement feature is determined as the first feature. If the extracted requirement feature is "draping coefficient 20%", the requirement feature is determined as the second feature.

[0041] In some embodiments, the user requirement information can be feature - extracted by constructing a multimodal feature extraction engine. The multimodal feature extraction engine can include a text feature extraction subsystem and an image feature parsing subsystem. The text feature extraction subsystem extracts features from the text description in the user requirement information, and the image feature parsing subsystem extracts features from the image data in the user requirement information.

[0042] Among them, the text feature parsing subsystem analyzes the text description input by the user to extract the features of the fabric. Specifically, the text feature parsing subsystem can construct a domain adapter based on the RoBERTa model, combined with the textile domain experience map. At the same time, a parameter extraction rule engine is developed to support regular expressions. Among them, the textile domain experience map contains the standardized description and extended corpus of the textile domain. Through the text feature parsing subsystem, the text description can be converted into feature information in the textile domain and the feature category can be judged. For example, if the text description is a fabric that slightly reflects light in the sun, through the text feature parsing subsystem, the required feature can be obtained as "mercerized". Further, the text feature parsing subsystem can judge "mercerized" as the first feature. The image feature extraction subsystem identifies the feature information in the image data through image recognition technology, including style, texture, color, etc. Specifically, the image feature extraction subsystem can adopt an improved ResNet-50 architecture, combined with the textile domain experience map, fine-tune on the basis of the ImageNet pre-trained model, add a texture attention module (Texture Attention Module), and replace the output layer with a feature space projection layer. Through the image feature extraction subsystem, the image data can be converted into feature information in the textile domain and the feature category can be judged. For example, if the image data is a photo of a certain female star wearing a red long dress at a dinner party, through the image feature extraction subsystem, the required features can be obtained as "long dress", "red", and "banquet occasion". Further, the image feature extraction subsystem can judge "banquet occasion" as the first feature and "red" and "long dress" as the second features.

[0043] Step S203, obtaining target fabric parameters based on the required features, where the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters.

[0044] Specifically, the target fabric parameters are fabric parameters that can guide fabric production. The target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters. Among them, the target fabric structure parameters mainly describe the manufacturing process and structural characteristics of the fabric, reflecting the physical form and performance of the fabric. Common target fabric structure parameters include: target fabric process parameters, such as woven fabric, knitted fabric, non-woven fabric; target fabric texture parameters, such as plain weave, twill, satin; target fabric density parameters, for example, the density of woven fabric is expressed by warp density (warp yarn density) and weft density (weft yarn density), the density of knitted fabric is expressed by transverse density and longitudinal density, and the density of non-woven fabric is indirectly expressed by unit area mass (gram weight). Common target fabric raw material parameters include: target fiber type parameters, such as natural fibers like cotton, hemp, silk, wool, etc.; target yarn parameters, such as the thickness, strength, and twist of the yarn; target blend ratio, such as cotton-polyester blend. Common target fabric color parameters include: target fabric color intensity parameter, target fabric saturation parameter, and target fabric color difference parameter.

[0045] In some embodiments, obtaining target fabric parameters based on demand characteristics includes: Inputting the first feature into a pre-trained fabric parameter model for fabric parameter extraction to obtain the first target fabric parameters, where the first feature includes the visual feature of the fabric; Taking the first target fabric parameters as the target fabric parameters, where the first target fabric parameters include the first fabric structure parameters, the first fabric raw material parameters, and the first fabric color parameters.

[0046] In some embodiments, the first feature further includes a tactile feature.

[0047] Specifically, the pre-trained fabric parameter model can obtain the corresponding first target fabric parameters through the first feature. The first target fabric parameters include: the first fabric structure parameters, the first fabric raw material parameters, and the first fabric color parameters.

[0048] Specifically, the first target fabric parameters can be taken as the target fabric parameters, that is, the first fabric structure parameters, the first fabric raw material parameters, and the first fabric color parameters output by the pre-trained fabric parameter model are taken as the target fabric parameters.

[0049] In some embodiments, after inputting the first feature into the pre-trained fabric parameter model for fabric parameter extraction to obtain the first target fabric parameters, the demand characteristics further include a second feature. Obtaining target fabric parameters based on the demand characteristics further includes: Obtaining the second target fabric parameters according to the second feature; Combining the first target fabric parameters and the second target fabric parameters to obtain the target fabric parameters.

[0050] Specifically, the second feature at least includes one of the following features: fabric physical features, fabric chemical features, fabric size features, and fabric price features. According to the second feature, second target fabric parameters can be obtained. The second target fabric parameters include at least one of the following parameters: second fabric structure parameters, second fabric raw material parameters, second fabric color parameters, fabric size parameters, and fabric price parameters. Specifically, one or more of the second fabric structure parameters, second fabric raw material parameters, and second fabric color parameters can be obtained through fabric physical features and / or fabric chemical features; fabric size parameters can be obtained through fabric size features; and fabric price parameters can be obtained through fabric price features. The second feature can have an explicit association with the second target fabric parameters or an implicit association with the second target fabric parameters.

[0051] In some embodiments, when the second feature has an explicit association with the fabric parameters, the second target fabric parameters can be directly obtained through the second feature. For example, the fabric size feature can be a direct size feature, which has an explicit association with the fabric parameters. Through the direct size feature, the fabric size parameters can be directly obtained; the fabric price feature can be a direct price feature, which has an explicit association with the fabric parameters. Through the direct price feature, the fabric price parameters can be directly obtained.

[0052] In some embodiments, when the second feature has an implicit association with the fabric parameters, the second target fabric parameters can be indirectly obtained through the second feature based on a structured experience knowledge base. For example, the fabric size feature can be an indirect size feature used to characterize the user's demand for the fabric finished product, which has an implicit association with the fabric parameters. For example, for "a men's shirt", through the structured experience knowledge base, the fabric size parameters corresponding to "a men's shirt" can be obtained as a narrow width of 2.0 m - 3.5 m and a wide width of 1.5 m - 2.5 m; the fabric price feature can be an indirect price feature used to characterize the user's budget for fabric production, which has an implicit association with the fabric parameters. Based on the structured experience knowledge base, the fabric price parameters can be indirectly obtained through the indirect price feature. Among them, the structured experience knowledge base can be built through the enterprise historical work order database.

[0053] In some embodiments, when the second feature includes the fabric price feature, the target fabric parameters further include target fabric price parameters. Specifically, the fabric price parameters obtained through the fabric price feature are used as the target fabric price parameters; when the second feature includes the fabric size feature, the target fabric parameters further include target fabric size parameters. Specifically, the fabric size parameters obtained through the fabric size feature are used as the target fabric size parameters.

[0054] Specifically, when the first target fabric parameter and the second target fabric parameter overlap in the fabric parameter dimension, the first target fabric parameter and the second target fabric parameter in this dimension can be combined to obtain the target fabric parameter in this dimension. In the dimension without overlap, the first target fabric parameter or the second target fabric parameter existing in this dimension is used as the target fabric parameter, that is, if only the first target fabric parameter exists in this dimension, the first target fabric parameter in this dimension is used as the target fabric parameter in this dimension; similarly, if only the second target fabric parameter exists in this dimension, the second target fabric parameter in this dimension is used as the target fabric parameter in this dimension.

[0055] In some embodiments, the first target fabric parameter and the second target fabric parameter in the same dimension can be combined by the weighted assignment fusion method. Specifically, dynamic weights can be assigned to the first target fabric parameter and the second target fabric parameter, and the combination is achieved through weighted calculation. This solution is simple and intuitive and is suitable for scenarios with clear demand priorities. For example, if the user wants to set the priorities by themselves, different weights can be set for different target fabric parameters to achieve this.

[0056] In some embodiments, the first target fabric parameter and the second target fabric parameter in the same dimension can be combined by the constrained optimization solution method. Specifically, an optimization model can be constructed through an objective function and constraint conditions, and the optimal parameter combination is solved under the satisfaction of production constraints. For example, both the first target fabric parameter and the second target fabric parameter include the fabric thickness parameter. In the first target fabric parameter, the fabric thickness parameter is 0.12 mm - 0.25 mm, and in the second target fabric parameter, the fabric thickness parameter is 0.15 mm - 0.35 mm. Through the optimization model, the fabric thickness parameter in the target fabric parameter is obtained as 0.15 mm - 0.25 mm. This solution strictly satisfies the production hard constraints and is suitable for complex engineering scenarios.

[0057] Through the above steps S201 to S203, user demand information is obtained, where the user demand information includes the text description and / or image data of the fabric; the features of the user demand information are extracted to obtain demand features, where the demand features include the first feature; the target fabric parameters are obtained based on the demand features, where the target fabric parameters include the target fabric structure parameters, the target fabric raw material parameters, and the target fabric color parameters. Compared with the prior art that mainly relies on the profound professional knowledge and rich experience of designers to improve the innovation of fabric design, this embodiment establishes a mapping between user demands and fabric parameters. Through the user demand information, especially through the user's subjective demands (such as "having a shiny feeling"), the fabric parameters of the fabric required by the user are obtained, reducing the dependence on professional capabilities for fabric design innovation and reducing the cost of fabric design innovation.

[0058] In this embodiment, a method for training a fabric parameter model is also provided. Figure 3 It is the flowchart of training the fabric parameter model of the fabric parameter generation method in this embodiment. As Figure 3 shown, the training process of the fabric parameter model includes: Step S301, construct a training data set. Among them, the training data set includes the feature information and fabric parameters of the training fabric. Among them, the feature information includes the visual features of the training fabric; the fabric parameters include the fabric structure parameters, fabric raw material parameters, and fabric color parameters of the training fabric. Specifically, the feature information of the training fabric is the subjective attribute of the fabric directly perceived by the human sensory system, with non-quantitative and descriptive features. The feature information of the training fabric includes visual features. For example, style features include elegance, liveliness, fashion, etc.; scene features include outdoor occasions, home occasions, formal occasions, etc.; appearance features include drape, gloss, etc. Among them, the gloss can include mercerized and high gloss.

[0059] In some embodiments, the first feature further includes the tactile feature of the fabric, such as the soft feature.

[0060] In some embodiments, the feature information of the training fabric can be collected through a multi-modal data acquisition mechanism. Specifically, the multi-modal data acquisition mechanism includes an expert evaluation mechanism and a parameterized evaluation mechanism. Among them, the expert evaluation mechanism relies on textile experts to conduct multi-dimensional sensory evaluations on the training fabric, and realizes the standardized annotation of feature information through a standardized evaluation scale. This mechanism is especially suitable for the collection of non-quantifiable feature information such as style features and scene features. For features with a physical measurement basis, the parameterized evaluation mechanism adopts a parameterized modeling strategy to realize data conversion by establishing a mapping relationship between features and parameters. For example, gloss feature quantification: based on the ratio parameter of specular reflectance (SR) and diffuse reflectance (DR), a gloss parameterization model is constructed. When SR / DR≥0.85, the training fabric is labeled with the "high gloss" feature. When SR / DR is in the range of [0.6, 0.85), the training fabric is labeled with the "mercerized" feature.

[0061] Specifically, the fabric parameters of the training fabric can be obtained through methods such as digital image processing, automatic counting with an optical density ruler, and near-infrared spectroscopy. Among them, the fabric parameters of the training fabric include the fabric structure parameters, fabric raw material parameters, and fabric color parameters of the training fabric.

[0062] Step S302, train the initial fabric parameter model based on the feature information of the training fabric and the fabric parameters of the training fabric to obtain a trained fabric parameter model.

[0063] Specifically, using the feature information of the training fabric as the input and the fabric parameters of the training fabric as the output, the fabric parameter model is trained to obtain a trained fabric parameter model.

[0064] In some embodiments, the fabric parameter model includes an independent binary encoder that converts the feature information of the training fabric into a feature vector. Specifically, the feature information is converted into a feature vector by the presence or absence of features in a certain dimension.

[0065] In some embodiments, the feature information of the input training fabric is randomly masked to force the model to learn incomplete feature reasoning. Specifically, intelligent masking can be performed based on the experience map in the textile field, retaining at least one key feature; or, new training data combinations are generated based on the feature information and fabric parameters of the existing training fabric. Specifically, physical constraint enhancement is performed during the process of generating new training data, and the enhancement method includes generating new training samples that meet the preset textile constraints through a conditional generation model. Through the above solutions, the fabric parameter model can still maintain effective reasoning for unseen but reasonable feature combinations and simultaneously identify abnormal inputs.

[0066] In some embodiments, step S302 includes: Step S3021, training the first initial fabric parameter model based on the feature information of the training fabric and the fabric structure parameters to obtain a first fabric parameter model.

[0067] Specifically, using the feature information of the training fabric as the input and the fabric structure parameters of the training fabric as the output, the first initial fabric parameter model is trained to obtain a first fabric parameter model.

[0068] In some embodiments, a deep residual network can be selected as the basic architecture of the first initial model, mapping the feature information to the fabric structure parameter space through a three-layer fully connected network; an improved hybrid loss function is used for loss calculation; and the AdamW optimizer is used for backpropagation optimization.

[0069] Step S3022, training the second initial fabric parameter model based on the feature information of the training fabric, the fabric raw material parameters, and the fabric color parameters to obtain a second fabric parameter model; Specifically, using the feature information of the training fabric as the input and the fabric raw material parameters and the fabric color parameters of the training fabric as the output, the second initial fabric parameter model is trained to obtain a second fabric parameter model.

[0070] In some embodiments, a Transformer-CNN hybrid model can be selected as the infrastructure of the second initial fabric parameter model to build a multi-task learning framework; differential training can be implemented according to parameter importance to calculate the dynamic weighted loss; and the DeepSpeed Zero-3 optimization strategy can be used for distributed training.

[0071] Step S3023: Combine the first fabric parameter model and the second fabric parameter model to obtain a trained fabric parameter model.

[0072] Combine the first fabric parameter model trained in step S3021 and the second fabric parameter model trained in step S3022 to obtain a trained fabric parameter model. The combination method can be model fusion, model concatenation, or model parallel connection, which can be specifically selected according to actual requirements and is not specifically limited in this embodiment.

[0073] In some embodiments, the first fabric parameter model and the second fabric parameter model can be combined through dynamic weight fusion. Specifically, by statistically analyzing the prediction confidence of the first fabric parameter model and the second fabric parameter model on overlapping parameters (i.e., the feature information of the training fabric), the fusion weights are dynamically allocated, and the non-overlapping parameters are directly output to achieve parameter integration with minimized errors, which is applicable to medium-complexity scenarios that require balancing the advantages of the two models.

[0074] In one of the embodiments, the training dataset further includes the fabric organizational structure diagram of the training fabric, and the fabric organizational structure diagram is a three-dimensional representation diagram of the unit volume of the training fabric.

[0075] Figure 4 This is the training flowchart of the first fabric parameter model of the fabric parameter generation method in this embodiment, where the training process of the first fabric parameter model includes: Step S401: Train the first model based on the feature information of the training fabric and the fabric organizational structure diagram to obtain a trained first model.

[0076] Specifically, when constructing the training dataset, it also includes obtaining the organizational structure diagram of the training fabric, which is also used as training data. The fabric organizational structure diagram is a three-dimensional representation diagram of the fabric in the unit volume. Through its orthogonal projection cross-section and height calibration, a three-dimensional voxel matrix is formed to represent the process, organization, density, thickness, and jacquard of the fabric. The unit volume can be determined according to requirements. For example, 1 unit volume = 1 cm 3 , which is not specifically limited in this embodiment.

[0077] In some embodiments, the fabric organizational structure diagram can be generated through manual annotation, scanning, or simulation.

[0078] In some embodiments, a fabric texture corpus is constituted by the collected fabric texture diagrams.

[0079] Specifically, taking a woven fabric as an example, in the orthogonal projection section of its fabric texture diagram, longitudinal grids are used to represent warp yarns, and transverse grids are used to represent weft yarns. The texture of the fabric is characterized by the topological connection relationship matrix of the warp and weft yarns; the warp density and weft density are characterized by the number of longitudinal grids and the number of transverse grids in the section, that is, the yarn density is characterized; the thickness of the fabric is characterized by calibrating the height of the fabric texture diagram.

[0080] In some embodiments, when the woven fabric contains a jacquard process, the heights of the jacquard part and the non-jacquard part can be calibrated separately. Specifically, multi-scale height calibration can be implemented for the jacquard raised area.

[0081] Specifically, the first model takes the feature information of the training fabric as input, and learns to generate the corresponding fabric texture diagram through a generative network architecture (for example, a conditional generative adversarial network). The feature information is mapped into a latent space representation by an encoder, and the decoder converts it into a structure diagram in the format of 3D voxels or point clouds. The pixel-level reconstruction loss and the adversarial loss are jointly optimized to ensure that the generated structure is consistent with the physical characteristics, and the trained first model is obtained.

[0082] Step S402: Train the second model based on the fabric texture diagram and the fabric structure parameters to obtain the trained second model.

[0083] Specifically, the second model takes the fabric texture diagram as input, extracts multi-scale structure features through a 3D convolutional neural network, maps the output through a multi-layer perceptron to obtain the fabric structure parameters, optimizes the parameter prediction accuracy using the smooth L1 loss, and adds a constraint loss based on physical simulation (for example, mechanical property deviation). The generalization ability of the model is improved through pre-training of the image encoder and data augmentation, and the trained second model is obtained.

[0084] Step S403: Combine the trained first model and the trained second model to obtain the first fabric parameter model.

[0085] Specifically, the trained first model and the trained second model can be combined through cascade training, and the parsing compatibility of the fabric texture diagram is optimized using backpropagation. A physical simulation verification loss is introduced to feedback the process constraints, forming a first fabric parameter model with the fabric texture diagram as an intermediate parameter.

[0086] Through the above steps S401 to S403, the first fabric parameter model can obtain the corresponding fabric structure parameter information according to the user demand information, so as to improve the richness of fabric parameter extraction; taking the fabric organizational structure diagram as the intermediate training feature, the risk of overfitting in the model training process can be reduced through task decomposition, the robustness can be enhanced, and the accuracy can be improved.

[0087] In another embodiment, after obtaining the trained fabric parameter model, it further includes: Construct a fabric feature test set, where the fabric feature test set includes the feature information and fabric parameters of the test fabric; Based on the fabric feature test set, verify the trained fabric parameter model.

[0088] Specifically, using the constructed fabric feature test set to verify the trained fabric parameter model, cross-validation, leave-one-out validation, etc. can be adopted. According to the prediction results of the model and the actual labeled data, calculate the performance indicators of the model, such as accuracy, recall rate, F1 score, etc. For regression tasks, the mean squared error (MSE) or mean absolute error (MAE) can also be calculated.

[0089] Analyze the performance of the model on different features and parameters, and find out the strengths and weaknesses of the model. For the parts with poor performance, the model structure can be further adjusted or the training strategy can be optimized.

[0090] Through the above steps, the fabric parameter model can have high accuracy and reliability in actual applications, providing strong support for the design, production and quality control of fabrics.

[0091] In this embodiment, a fabric generation method is also provided, and this fabric generation method applies the fabric parameter generation method described in any one of the above embodiments. Figure 5 It is the flowchart of the fabric generation method of this embodiment, as Figure 5 shown, and this process includes the following steps: Step S501, according to the target fabric parameters, judge whether there is a corresponding recommended fabric in the preset fabric library. If so, execute step S502; otherwise, execute step S503; Specifically, the preset fabric library contains various fabrics and their detailed information, and the information of each fabric should cover its structure parameters, raw material parameters, color parameters, pricing, inventory, etc.

[0092] In some embodiments, a multi-dimensional parameter matching mechanism can be used to judge whether there is a corresponding recommended fabric in the preset fabric library. Specifically, a hybrid similarity algorithm can be adopted to retrieve in the preset fabric library according to the target fabric parameters, and judge whether there is a fabric that meets the target fabric parameters in the preset fabric library.

[0093] Step S502: Obtain the corresponding recommended fabric according to the target fabric parameters; Specifically, if there is a fabric in the preset fabric library that meets the target fabric parameters, directly recommend this fabric.

[0094] In some embodiments, if there are multiple fabrics in the preset fabric library that meet the target fabric parameters, the recommended fabrics can be sorted based on the principle of parameter similarity priority or the principle of cost - benefit optimization.

[0095] Step S503: Generate a virtual fabric according to the target fabric parameters and store the virtual fabric in the virtual fabric library.

[0096] Specifically, if there is no fabric in the preset fabric library that meets the target fabric parameters, high - precision physical simulation technology can be used to generate a virtual fabric according to the target fabric parameters and store the virtual fabric in the virtual fabric library.

[0097] In some embodiments, a virtual fabric can be generated by calling an AI physical engine (such as HeartDub Pro). Specifically, a 3D voxel model of fiber arrangement and yarn interweaving can be generated based on the target parameters, and mechanical properties such as stretching and draping can be calculated through finite - element analysis to ensure that the deviation between the virtual fabric and the physical object is less than the preset error threshold. When storing, a producibility label (such as "requires a double - sided loom") can be automatically marked, and it can be linked with the supplier database to recommend a suitable manufacturer. Further, the virtual fabric library can be docked with an e - commerce platform so that C - end users can directly preview the fabric effect.

[0098] Through the above - mentioned steps S501 to S503, it can be effectively determined whether there is a recommended fabric in the fabric library that matches the target fabric parameters, and a virtual fabric can be generated when there is no matching fabric. Among them, the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters. Compared with the prior art's reliance on physical fabric samples, this solution improves the fabric development efficiency, reduces the consumption of physical sample fabrics, and at the same time establishes a virtual fabric library that can be directly docked with suppliers and e - commerce platforms, providing more comprehensive support for fabric selection and design.

[0099] Figure 6 is the preferred flowchart of the fabric generation method of this embodiment. As Figure 6 shown, the fabric generation method includes the following steps: Step S601: Obtain user demand information, where the user demand information includes text descriptions and / or image data of the fabric; Step S602: Extract features from the user demand information to obtain demand features, where the demand features include first features; Step S603: Input the first feature into a pre-trained fabric parameter model to extract fabric parameters, obtaining first target fabric parameters. Here, the first feature includes the visual features of the fabric, and the first target fabric parameters include first fabric structure parameters, first fabric raw material parameters, and first fabric color parameters; Step S604: Determine whether the demand feature contains a second feature. If so, execute Step S605; otherwise, execute Step S606; Step S605: Obtain second target fabric parameters based on the second feature; combine the first target fabric parameters and the second target fabric parameters to obtain target fabric parameters; Step S606: Take the first target fabric parameters as the target fabric parameters; Step S607: Based on the target fabric parameters, determine whether there is a corresponding recommended fabric in the preset fabric library. If so, execute Step S608; otherwise, execute Step S609; Step S608: Obtain the corresponding recommended fabric according to the target fabric parameters; Step S609: Generate a virtual fabric according to the target fabric parameters, and store the virtual fabric in the virtual fabric library.

[0100] Through the above steps S601 to S609, according to the demand information of the fabric input by the user, extract the demand feature from the user demand information, input the first feature into a pre-trained fabric parameter model to extract fabric parameters, obtaining first target fabric parameters, so as to convert the subjective demand information of the user into fabric parameters that can guide production. If the demand feature also includes a second feature, extract second target fabric parameters based on the second feature, and combine the first target fabric parameters and the second target fabric parameters to obtain target fabric parameters. Here, the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters; then recommend fabrics to the user according to the target fabric parameters. If there is a fabric with corresponding parameters in the existing fabric library, directly recommend the relevant fabric in the fabric library to the user; if there is no fabric with corresponding parameters in the existing fabric library, generate a virtual fabric according to the target fabric parameters and store the virtual fabric in the virtual fabric library. This fabric generation method can be applied not only to the clothing industry but also to any industry that uses fabrics, such as home furnishing and medical. Through this fabric generation method, it is realized to generate corresponding fabric parameters according to the subjective demand information of the user, and further generate corresponding recommended fabrics, reducing the dependence on professional capabilities for fabric acquisition and reducing the fabric design and innovation cost.

[0101] In this embodiment, a fabric parameter generation device is further provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated here. The following terms such as "module", "unit", "sub-unit", etc. can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0102] Figure 7 is the structural block diagram of the fabric parameter generation device in this embodiment. As Figure 7 shown, the device 70 includes: an acquisition module 71, an extraction module 72, and a fabric parameter generation module 73. Among them, The acquisition module 71 is used to acquire user requirement information, where the user requirement information includes the text description and / or image data of the fabric; The extraction module 72 is used to extract the features of the user requirement information to obtain requirement features, where the requirement features include first features; The fabric parameter generation module 73 is used to obtain target fabric parameters based on the requirement features, where the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters.

[0103] It should be noted that the above-mentioned each module can be a functional module or a program module, and can be implemented either by software or by hardware. For the modules implemented by hardware, the above-mentioned each module can be located in the same processor; or the above-mentioned each module can also be located in different processors in any combination form.

[0104] In this embodiment, an electronic device is further provided, including a memory and a processor. A computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0105] Optionally, the above-mentioned electronic device may further include a transmission device and input / output devices. Among them, the transmission device is connected to the above-mentioned processor, and the input / output devices are connected to the above-mentioned processor.

[0106] Optionally, in this embodiment, the above-mentioned processor can be configured to execute the following steps through a computer program: S1. Acquire user requirement information, where the user requirement information includes the text description and / or image data of the fabric; S2. Extract the features of the user requirement information to obtain requirement features, where the requirement features include first features; S3. Obtain target fabric parameters based on demand characteristics, where the target fabric parameters include target fabric structure parameters, target fabric raw material parameters, and target fabric color parameters.

[0107] It should be noted that specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementation manners, and will not be elaborated herein.

[0108] In addition, in combination with the fabric parameter generation method provided in the above embodiments, a storage medium can also be provided in this embodiment to implement it. A computer program is stored on the storage medium; when the computer program is executed by a processor, any one of the fabric parameter generation methods in the above embodiments is implemented.

[0109] It should be understood that the specific embodiments described here are only used to explain this application, rather than to limit it. According to the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of this application.

[0110] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data that have been authorized by the user or fully authorized by all parties.

[0111] Obviously, the drawings are only some examples or embodiments of this application. For those of ordinary skill in the art, this application can also be applied to other similar situations based on these drawings without creative work. In addition, it can be understood that although the work done during this development process may be complex and time-consuming, for those of ordinary skill in the art, certain design, manufacturing, or production changes based on the technical content disclosed in this application are only conventional technical means and should not be regarded as insufficient disclosure of this application.

[0112] The term "embodiment" in this application means that the specific features, structures, or characteristics described in combination with the embodiment may be included in at least one embodiment of this application. The phrase appears in various positions in the specification does not necessarily mean the same embodiment, nor does it mean being independent or alternative to other embodiments and mutually exclusive. Those of ordinary skill in the art can clearly or implicitly understand that the embodiments described in this application can be combined with other embodiments without conflict.

[0113] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0114] The above embodiments merely represent several implementation manners of the present application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of patent protection. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for generating fabric parameters, characterized in that: include: Acquiring user demand information, wherein the user demand information includes a text description and / or image data of the fabric; Extracting features from the user demand information to obtain demand features, wherein the demand features include a first feature; Target fabric parameters are acquired based on the demand characteristics, wherein the target fabric parameters include target fabric construction parameters, target fabric raw material parameters and target fabric color parameters.

2. The method for generating fabric parameters according to claim 1, characterized in that: The step of obtaining target fabric parameters based on the demand characteristics includes: Inputting the first feature into a pre-trained fabric parameter model to extract fabric parameters to obtain first target fabric parameters, wherein the first feature includes a visual feature of the fabric; The first target fabric parameters are used as the target fabric parameters, wherein the first target fabric parameters include first fabric structure parameters, first fabric material parameters and first fabric color parameters.

3. The method for generating fabric parameters according to claim 2, characterized in that: After the first feature is input into a pre-trained fabric parameter model to extract fabric parameters and obtain the first target fabric parameter, the demand feature further includes a second feature, and the step of obtaining the target fabric parameter based on the demand feature further includes: obtaining a second target fabric parameter according to the second feature; The target fabric parameters are obtained by combining the first target fabric parameters and the second target fabric parameters.

4. The method for generating fabric parameters according to claim 3, characterized in that: When the second feature includes a fabric price feature, the target fabric parameter further includes a target fabric price parameter; when the second feature includes a fabric size feature, the target fabric parameter further includes a target fabric size parameter.

5. The method for generating fabric parameters according to claim 2, characterized in that: The fabric parameter model training process includes: Constructing a training data set, wherein the training data set includes feature information and fabric parameters of training fabrics, wherein the feature information includes visual features of the training fabrics; the fabric parameters include fabric construction parameters, fabric raw material parameters, and fabric color parameters of the training fabrics; The initial fabric parameter model is trained based on the feature information of the training fabric and the fabric parameters of the training fabric to obtain the trained fabric parameter model.

6. The method for generating fabric parameters according to claim 5, characterized in that: The training of the initial fabric parameter model based on the feature information of the training fabric and the fabric parameters of the training fabric to obtain the trained fabric parameter model includes: Training a first initial fabric parameter model based on the feature information of the training fabric and the fabric construction parameters to obtain a first fabric parameter model; Training a second initial fabric parameter model based on the feature information of the training fabric, the fabric raw material parameters, and the fabric color parameters to obtain a second fabric parameter model; The trained fabric parameter model is obtained by combining the first fabric parameter model and the second fabric parameter model.

7. The method for generating fabric parameters according to claim 6, characterized in that: The training data set also includes a fabric structure diagram of the training fabric.

8. The method for generating fabric parameters according to claim 7, characterized in that: The step of training the first initial fabric parameter model based on the feature information and fabric construction parameters of the training fabric to obtain the first fabric parameter model comprises: Training the first model based on the characteristic information of the training fabric and the fabric tissue structure diagram to obtain a trained first model; Training the second model based on the fabric structure diagram and the fabric construction parameters to obtain a trained second model; The first fabric parameter model is obtained by combining the trained first model and the trained second model.

9. A fabric generation method, using the fabric parameter generation method according to any one of claims 1 to 8, characterized in that: include: According to the target fabric parameters, determining whether there is a corresponding recommended fabric in a preset fabric library; If yes, then obtain the corresponding recommended fabric according to the target fabric parameters; otherwise, generate a virtual fabric according to the target fabric parameters, and store the virtual fabric in a virtual fabric library.

10. A fabric parameter generating device, characterized in that: include: Acquisition module, extraction module and fabric parameter generation module, wherein, The acquisition module is used to acquire user demand information, wherein the user demand information includes a text description and / or image data of the fabric; The extraction module is used to extract the features of the user demand information to obtain demand features, wherein the demand features include a first feature; The fabric parameter generation module is used to obtain target fabric parameters based on the demand characteristics, wherein the target fabric parameters include target fabric construction parameters, target fabric raw material parameters and target fabric color parameters.

11. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to run the computer program to execute the fabric parameter generating method according to any one of claims 1 to 8.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the fabric parameter generating method according to any one of claims 1 to 8 are implemented.

Images