Clothing panel generation method and apparatus, electronic device, and storage medium

Abstract

This application relates to the field of clothing design and discloses a method, apparatus, electronic device, and storage medium for generating clothing patterns. The method for generating clothing patterns includes: determining a clothing design language input by a user; and calling a pattern generation model to generate a target clothing pattern based on the clothing design language, including: converting the clothing design language into a symbolic pattern-making language, and outputting the target clothing pattern according to the symbolic pattern-making language. Compared with existing technologies, this solution can reduce the professional skills required of users in the clothing pattern preparation process and improve pattern-making efficiency.

Description

Technical Field

[0001] This application relates to the field of clothing design, and more specifically, to a method, apparatus, electronic device, and storage medium for generating clothing patterns. Background Technology

[0002] In the apparel industry, pattern making has always been a major challenge. With technological advancements, traditional hand-made methods are gradually revealing their limitations in the context of rapid digitalization and personalized demands. The market's demand for more refined and diverse designs continues to grow, while hand-made pattern making, due to its time-consuming nature, high error rate, and low efficiency, is no longer able to meet the requirements of modern garment production.

[0003] To address this challenge, pattern-making software such as Richpeace employs a symbolic programming-based procedural modeling method. This software transforms the garment pattern-making process into symbolic drawing based on points, lines, and surfaces, offering advantages such as easily understandable design parameters, support for random variations, high output quality, and compact representation.

[0004] Nevertheless, the pattern making process remains complex, requiring not only deep knowledge of pattern making from the maker, but also intricate coupling relationships between steps and parameters. This necessitates strong mathematical reasoning, geometric intuition, and programming skills from the user, limiting its flexibility and universal applicability. Consequently, the preparation of garment patterns demands high levels of expertise while resulting in low efficiency. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus, electronic device and storage medium for generating garment patterns, which can reduce the professional skills required of users in the garment pattern preparation process and improve pattern making efficiency.

[0006] In a first aspect, embodiments of this application provide a method for generating garment patterns, comprising: determining a garment design language input by a user; and calling a pattern generation model to generate a target garment pattern based on the garment design language, comprising: converting the garment design language into a symbolic pattern-making language, and outputting the target garment pattern according to the symbolic pattern-making language.

[0007] Compared with related technologies, the garment pattern generation method provided in this application can directly call the pattern generation model to generate the target garment pattern based on the user's input of garment design language after receiving the input. Therefore, users only need to input garment design language such as natural language, design drafts, or photos when making patterns, without needing to perform complex professional operations such as typesetting or programming. This significantly reduces the professional skills required of users in the garment pattern preparation process. Furthermore, due to the high degree of automation in the entire pattern making process, users can perform a large number of garment pattern making tasks simultaneously, thereby improving overall pattern making efficiency. In addition, this solution breaks down the target garment pattern generation process into a language generation process and a pattern output process through language conversion, reducing the coupling between steps and making the complete generation process of the target garment pattern clearer and more orderly. This allows users to accurately control parameters such as pattern size, dimensions, and shape through symbolic pattern making language, which helps improve the accuracy of the generated patterns.

[0008] Secondly, embodiments of this application provide a training method for a target language parsing model, comprising: acquiring a training dataset and a standard multimodal large model, wherein each training data in the training dataset includes sample clothing design language and sample symbolic pattern making language; and fine-tuning the standard multimodal large model using the training dataset to obtain the target language parsing model.

[0009] Compared with related technologies, the embodiments of this application train a target language parsing model by fine-tuning a standard multimodal large model. It is understood that, on the one hand, the standard multimodal large model itself already possesses powerful analytical and reasoning capabilities for multimodal data; therefore, the target language parsing model obtained by fine-tuning this model naturally also possesses this capability. On the other hand, using the sample symbolic pattern-making language as a label for the sample garment design language to construct samples for fine-tuning the standard multimodal large model ensures that the generated target language parsing model can comprehensively analyze the garment design language and generate relatively accurate symbolic pattern-making language, which helps improve the accuracy and quality of the final generated target garment pattern.

[0010] Thirdly, embodiments of this application provide a training method for a target symbol translation model, comprising: acquiring an initial model and training samples; training the initial model using the training samples to obtain the target symbol translation model, which is used to convert symbolic pattern making language into drawing commands, and / or convert modified symbolic pattern making language into modified drawing commands. Similar to the training process of the aforementioned pattern generation model, the initial model itself already possesses powerful data analysis and reasoning capabilities, so the target symbol translation model trained with this model naturally also possesses this capability; on the other hand, training the initial model with training samples enables the generated target language parsing model to comprehensively analyze clothing design language and generate relatively accurate symbolic pattern making language, which also helps to improve the accuracy and quality of the final generated target clothing pattern.

[0011] Fourthly, embodiments of this application provide a target language parsing model software, which is used to: convert clothing design language into symbolic pattern making language; and / or convert modified clothing design language into modified symbolic pattern making language.

[0012] Fifthly, embodiments of this application provide a target symbol translation model software, which is used to: convert symbolic stencil language into drawing commands, and / or convert modified symbolic stencil language into modified drawing commands.

[0013] Sixthly, embodiments of this application provide a garment pattern generation device, comprising: a design language determination module for determining a garment design language input by a user; and a garment pattern generation module for calling a pattern generation model to generate a target garment pattern based on the garment design language.

[0014] In a seventh aspect, embodiments of this application provide a training apparatus for a target language parsing model, comprising: a standard model acquisition unit, configured to acquire a training dataset and a standard multimodal large model, wherein each training data in the training dataset includes sample clothing design language and sample symbolic pattern making language; and a parsing model training unit, configured to fine-tune the standard multimodal large model using the training dataset to obtain the target language parsing model.

[0015] Eighthly, embodiments of this application provide a training apparatus for a target symbol translation model, comprising: an initial model acquisition unit for acquiring an initial model and training samples; and a translation model training unit for training the initial model using the training samples to obtain the target symbol translation model, wherein the model is used to convert symbolic stencil language into drawing commands, and / or convert modified symbolic stencil language into modified drawing commands.

[0016] The advantages of the above-mentioned devices compared with related technologies can be found in the detailed analysis of the corresponding method embodiments above, and will not be repeated here.

[0017] Ninthly, embodiments of this application provide an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to implement the methods described in the first, second, or third aspects above.

[0018] In a tenth aspect, embodiments of this application provide a computer-readable storage medium storing a computer program, the computer program being implemented by a processor as described in the first, second, or third aspects above.

[0019] In one aspect, embodiments of this application provide a computer program product, the computer program product including computer program code, wherein when the computer program code is executed by a computer device, the computer device implements the method described in the first, second or third aspect above. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings.

[0021] Figure 1 This is a schematic flowchart of a method for generating garment patterns according to an embodiment of this application;

[0022] Figure 2 This is a schematic diagram of the design draft and garment photograph in the garment pattern generation method provided in an embodiment of this application;

[0023] Figure 3 This is a schematic diagram illustrating the display effect of a target garment pattern provided in an embodiment of this application;

[0024] Figure 4 This is a flowchart illustrating the generation of a target garment pattern based on a pattern generation model, provided in one embodiment of this application.

[0025] Figure 5 This is a flowchart illustrating a training method for a target language parsing model provided in an embodiment of this application.

[0026] Figure 6A schematic flowchart illustrating the training method of the target symbol translation model provided in an embodiment of this application;

[0027] Figure 7 This is a schematic diagram of the structure of a garment pattern generating device provided in an embodiment of this application;

[0028] Figure 8 A schematic diagram of the structure of a training device for a target language parsing model provided in an embodiment of this application;

[0029] Figure 9 A schematic diagram of the structure of a training device for a target symbol translation model provided in an embodiment of this application;

[0030] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0032] Therefore, the following detailed description of embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the present application.

[0033] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0034] Furthermore, the terms "first" and "second" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0035] It should be noted that, where there is no conflict, the features in the embodiments of this application can be combined with each other.

[0036] Embodiment 1 of this application provides a method for generating garment patterns, which is applied to garment pattern generating equipment, such as... Figure 1 As shown, the specific methods for generating garment patterns include:

[0037] Step S101: Determine the clothing design language input by the user.

[0038] In step S101, the clothing design language input by the user can be natural language, user-uploaded design drafts, photos, or other image-based languages. For example, the clothing design language can be natural language descriptions entered by the user during pattern making, such as "A black, A-line dress with a fitted bodice and flaring skirt"; or the clothing design language can also be images uploaded by the user to the garment pattern generation device during pattern making. Figure 2 The design sketches or garment photos shown can be obtained according to the actual design needs.

[0039] This system can display a text input interface to the user, allowing them to select and upload manually entered text, copied and pasted text, etc., at a preset location on the interface, and use this text as the user's input clothing design language. Alternatively, a file upload control can be provided to the user. The user can then select and upload design drafts, photos, or other files stored locally on the clothing pattern generation device, or download tables or images from a preset online file platform and specify them as the clothing design language. The clothing design language can be pre-created by the user or temporarily generated using a file generation plugin provided by the clothing pattern generation device (such as the AI-generated clothing image function provided by the target clothing software). This embodiment does not impose any limitations on this.

[0040] Step S102: Call the pattern generation model to generate the target garment pattern based on the garment design language, including: converting the garment design language into a symbolic pattern-making language, and outputting the target garment pattern according to the symbolic pattern-making language.

[0041] Compared with existing technologies, the garment pattern generation method provided in one embodiment of this application can directly call the pattern generation model to generate the target garment pattern based on the garment design language input by the user after receiving the garment design language input by the user. Therefore, when making patterns, users only need to input garment design language such as natural language, design drafts, or photos, without having to perform complex professional operations such as typesetting or programming. This not only significantly reduces the professional ability requirements of users in the garment pattern preparation process, but also, due to the high degree of automation of the entire pattern making process, users can perform a large number of garment pattern making tasks simultaneously, thereby improving the overall pattern making efficiency. In addition, this solution breaks down the target garment pattern generation process into a language generation process and a pattern output process through language conversion, reducing the coupling between steps and making the complete generation process of the target garment pattern clearer and more orderly. This allows users to accurately control parameters such as pattern size, dimensions, and shape through symbolic pattern making language, which helps to improve the accuracy of the generated patterns.

[0042] In step S102, the clothing design language needs to be converted into a symbolic pattern-making language first, and then the target clothing pattern is output according to the symbolic pattern-making language. This method breaks down the generation process of the target clothing pattern into a language generation process and a pattern output process, making the complete generation process of the target clothing pattern clearer and more orderly. The symbolic pattern-making language may include at least one of the following: structural commands, symbolic programs, and program parameters. For example, a symbolic program may be a function such as "make a top (_make_top_bodice)" or "make a skirt (_make_skirt)"; structural commands may be a set of functions that define the combination relationship and logical order between different symbolic programs, such as "stitching (Stitching)" or "paste (Paste)", for example, "stitch the top and skirt" or "paste the collar and hat"; program parameters may be relevant parameters for function execution, such as the size parameters of the top to be made, the size parameters of the skirt to be made, the stitch spacing for stitching, and the width of the overlapping part for pasting.

[0043] In one embodiment, target garment patterns can be output by target pattern-making software according to the symbolic pattern-making language. Pattern-making software in related technologies typically has the function of generating patterns based on symbolic languages. By converting the garment design language into a symbolic pattern-making language that the pattern-making software can recognize, this solution is compatible with pattern-making software in related technologies to quickly generate target garment patterns, thereby reducing the difficulty of implementing this solution and helping to improve pattern generation efficiency.

[0044] Furthermore, to help users more intuitively perceive the generation progress and intermediate processes of the target garment pattern, and to enhance the user experience, at least some symbolic pattern-making language can be displayed to the user. For example, the above-mentioned structural commands can be displayed so that users know the garment manufacturing process and the connection relationships between the patterns; and / or, the above-mentioned program parameters can also be displayed so that users know the garment dimensions and details such as the relative sizes of each pattern.

[0045] It is understood that the symbolic pattern-making language, including structural commands, symbolic programs, and program parameters, is merely a specific example in the embodiments of this application. It is only a specific, general logical abstraction of pattern-making rules, defining the garment objects and operations involved in the pattern-making process through the structure of structural commands, symbolic programs, and program parameters, and is not specific to the syntax of any industrial pattern-making language. Furthermore, the symbolic pattern-making language is independent of the specific syntax of the pattern-making software; therefore, different pattern-making software can use a symbolic pattern-making language that follows the same syntax, which helps improve the compatibility of this solution with different pattern-making software.

[0046] Furthermore, when outputting the target garment pattern according to the symbolic pattern-making language, the symbolic pattern-making language can first be converted into drawing commands, and then the target garment pattern can be output according to the drawing commands. Specifically, the target garment pattern can be generated by the target pattern-making software based on the drawing commands, such as inputting the drawing commands into the target pattern-making software and using the output of the target pattern-making software as the target garment pattern. When the pattern-making software has the function of generating patterns based on drawing commands, by converting the symbolic pattern-making language into drawing commands that the pattern-making software can recognize, this solution can be compatible with various types of pattern-making software to quickly generate target garment patterns, thereby reducing the difficulty of implementing this solution and helping to improve the efficiency of pattern generation.

[0047] The drawing command is a code segment used to generate a board piece, which can be applied to CAD board making software such as ET and Assyst. CAD board making software such as ET and Assyst can generate corresponding boards pieces according to the drawing command.

[0048] Furthermore, before converting the symbolic PCB design language into drawing commands, the target PCB design software can be determined first. This allows the symbolic PCB design language to be converted into the corresponding drawing commands for the target PCB design software, ensuring that the converted drawing commands can be accurately recognized by the target PCB design software used subsequently, thereby improving the quality of the generated PCB. It is understood that this solution can generate drawing commands in different forms (such as different syntax formats, different standards, etc.) by selecting different target PCB design software, thus achieving high compatibility with software formats.

[0049] In the foregoing embodiments, target garment patterns can be generated using target pattern-making software. This target pattern-making software can be pre-set pattern-making software or pattern-making software included in the garment design language, such as "using pattern-making software A to generate patterns for a black A-line dress, corset, and flared skirt," etc., and can be specifically set according to actual needs. For example, the pattern-making software can be the aforementioned CAD pattern-making software such as ET or Assyst. Furthermore, the target pattern-making software can run on the garment pattern generation device or on an external device connected to the garment pattern generation device, and can be used flexibly according to actual needs.

[0050] In one embodiment, after generating the target garment pattern based on the clothing design language in step S102, some or all of the target garment pattern can be displayed to the user so that the user can view its display effect and determine whether the current target garment pattern meets the user's pattern generation requirements. To facilitate a more intuitive and accurate viewing and decision-making process for the user, the patterns can be displayed in an orderly manner according to their relative positional relationships. For example, planar patterns can be displayed according to a planar relationship, or three-dimensional patterns can be displayed in three-dimensional space according to a spatial relationship (such as displaying patterns and their connections around a three-dimensional human body model). Furthermore, the three-dimensional garment wearing effect after pattern synthesis can be displayed in three-dimensional space according to the three-dimensional connection relationships of the patterns, etc., which will not be elaborated further.

[0051] In one embodiment, after the user inputs the clothing design language, the clothing pattern generation device can display the program parameters and target clothing pattern generated in the aforementioned embodiment in the relevant interface of the pattern generation function. For example... Figure 3 As shown, the relevant interface includes a parameter display area 301 and a pattern display area 302. The parameter display area 301 displays multiple program parameters 303 generated by the target language parsing model; the pattern display area 302 displays the target garment pattern 304 finally generated by the target pattern making software.

[0052] It is worth noting that the aforementioned program parameter 303 corresponds to the target garment pattern 304 and can be displayed in conjunction with it. For example, if the user is not satisfied with certain details of the currently displayed target garment pattern 304, they can adjust the corresponding program parameter 303 in the parameter display area 301 on the left, and click the "Reset" button after adjustment to trigger the target pattern-making software to regenerate the target garment pattern based on the adjusted program parameters. Conversely, the user can also directly adjust the size, relative position, and other information of one or more patterns 304 by dragging, pulling, or slugging. In this case, one or more program parameters 303 related to the pattern being adjusted displayed on the left interface will change in real time as the operation progresses, thereby achieving linkage between parameters and pattern display effects. This helps users view the adjustment effect in real time, improving adjustment efficiency and user experience.

[0053] Of course, if the user is not satisfied with the above program parameters 303 and / or target garment pattern 304, they can also click the "Re-enter" button and re-enter a new garment design language in the subsequently displayed interface, and trigger the device to regenerate the corresponding target garment pattern according to the language, which will not be elaborated further.

[0054] After adjusting in the aforementioned way, if the user is satisfied with the effect of the target garment pattern, they can click the "Save" button to save the relevant data, such as directly entering a standard format pattern file, etc., which will not be elaborated further.

[0055] Users may not be satisfied with the target garment pattern generated through the aforementioned method. Therefore, this solution allows users to further modify the target garment pattern. Specifically, users can input a modified garment design language. The garment pattern generation device can then determine the user's input modified garment design language, convert it into a modified symbolic pattern-making language, and then modify the target garment pattern according to the modified symbolic pattern-making language. After generating the target garment pattern, this solution can continue to receive user input of modified garment design languages ​​and enable convenient modifications to the target garment pattern based on these modified designs, simplifying the user's modification operations and improving modification efficiency.

[0056] Furthermore, when modifying the target garment pattern according to the modified symbolic pattern-making language, the modified symbolic pattern-making language is directly input into the aforementioned target pattern-making software, and the pattern output by the model is used as the modified target garment pattern. Alternatively, the modified symbolic pattern-making language can first be converted into modification drawing commands, and then the target garment pattern can be modified according to the modification drawing commands, such as inputting the modification drawing commands into the aforementioned target pattern-making software, and then using the pattern output by the model as the modified target garment pattern.

[0057] The specific operational processes of determining the user-inputted modified clothing design language, converting the modified clothing design language into a modified symbolic pattern-making language, and modifying the target according to the modified symbolic pattern-making language are not essentially different from the previous steps of determining the user-inputted clothing design language, converting the clothing design language into a symbolic pattern-making language, and generating the target clothing pattern according to the symbolic pattern-making language. Please refer to the previous description, and it will not be repeated here.

[0058] In addition, the language type of the modified clothing design language can be the same as or different from the clothing design language mentioned above. For example, both can be natural language or both can be image language such as design drawings; another example is that the aforementioned clothing design language is image language such as design drawings, while the modified clothing design language in this embodiment is natural language, etc.

[0059] As mentioned above, during the generation of the target garment pattern, it may be necessary to convert the garment design language into a symbolic pattern-making language; and / or, during the modification of the target garment pattern, it may be necessary to convert the modified garment design language into a modified symbolic pattern-making language. In one embodiment, the above conversion can be performed by the aforementioned pattern generation model. For example, if the pattern generation model includes a target language parsing model, this model can be used to convert the garment design language into the symbolic pattern-making language; and / or, to convert the modified garment design language into the modified symbolic pattern-making language.

[0060] Furthermore, during the generation of the target garment pattern, it may be necessary to convert the symbolic pattern-making language into drawing commands; and / or, during the modification of the target garment pattern, it may be necessary to convert the modified symbolic pattern-making language into modified drawing commands. In one embodiment, the above conversion can also be performed by the aforementioned pattern generation model. For example, if the pattern generation model further includes a target symbol translation model, this model can be used to convert the symbolic pattern-making language into the drawing commands; and / or, to convert the modified symbolic pattern-making language into the modified drawing commands.

[0061] As can be seen from the above embodiments, this solution breaks down the process of generating the target garment pattern into three sub-processes: "converting the garment design language into a symbolic pattern-making language by the target language parsing model," "converting the symbolic pattern-making language into drawing commands by the target symbolic translation model," and "generating the target garment pattern by the target pattern-making software based on the drawing commands." It is understandable that because the conversion logic of converting the garment design language into a symbolic pattern-making language differs significantly from the conversion logic of converting the symbolic pattern-making language into drawing commands, this solution uses the symbolic pattern-making language as an intermediary. Through the above breakdown, the two conversion logics are decoupled, making the respective operating logics of the target language parsing model and the target symbolic translation model clearer. Consequently, the training process is simpler, the amount of data required for training is smaller, and the inference process is more efficient.

[0062] In one embodiment, before invoking the target symbology translation model to convert the symbolic pattern-making language into drawing commands and drawing parameters, the target symbology translation model can be determined based on the target pattern-making software. This method ensures that the target symbology translation model used for the conversion matches the target pattern-making software used when generating the target garment pattern; that is, the pattern-making commands generated by the target symbology translation model are more compatible with the target pattern-making software used in subsequent steps, which helps improve the pattern-making quality of the target pattern-making software.

[0063] In determining the target symbol translation model based on the target template-making software, the target symbol translation model can be generated temporarily, or a model matching the target template-making software can be selected from multiple pre-trained symbol translation models. Taking the latter as an example, a preset symbol translation model list can be obtained first. This list records the correspondence between trained sample symbol translation models and sample template-making software. For example, the list includes multiple sample symbol translation models and sample template-making software corresponding to each sample symbol translation model. Based on this, the sample symbol translation model corresponding to the target template-making software recorded in the preset symbol translation model list can be determined as the target symbol translation model. The preset symbol translation model list includes multiple trained sample symbol translation models and their corresponding sample template-making software. The target symbol translation model can be determined directly by looking up the table without temporary model training, which helps improve the efficiency of determining the target symbol translation model.

[0064] Either the target language parsing model or the target symbol translation model mentioned in the foregoing embodiments can be an LLM (Large Language Model) or a VLM (Vision Language Model) or other LMM (Large Multimodal Model). The target language parsing model and / or target symbol translation model trained based on the aforementioned large models naturally possess the efficient analysis and reasoning capabilities inherent in large models for massive amounts of data. Therefore, it ensures that the symbolic pattern-making language generated by the target language parsing model and the drawing commands generated by the target symbol translation model are more accurate, thus helping to improve the quality of the final generated target garment pattern.

[0065] In one embodiment, when using target pattern-making software to determine the target garment pattern based on drawing commands and parameters, the drawing commands can be input into the target pattern-making software first, and the garment pattern generated by the software can be received. Then, a quality inspection model is used to determine whether the garment pattern conforms to preset pattern-making rules: if the garment pattern does not conform to the preset rules, the drawing commands can be adjusted according to the rules, and the adjusted commands can be input into the software to receive the target garment pattern; if the pattern conforms to the rules, it can be used as the target garment pattern. In this way, a pre-trained quality inspection model and preset pattern-making rules can be used to inspect the garment pattern generated by the software, making adjustments and regenerating it if necessary, to ensure that the final target garment pattern conforms to the preset rules, effectively improving the quality of the target garment pattern.

[0066] The quality inspection model is a mathematical model trained according to preset board-making rules. The preset board-making rules can be, for example, standardized board-making rules in the board-making industry, or board-making rules set according to one's own needs. Specifically, they can be flexibly set according to actual application needs.

[0067] In some embodiments of this application, the quality inspection model can also be a multimodal large model. Furthermore, in some embodiments, the target language parsing model and the quality inspection model can be the same multimodal large model, which can reduce the training process and training requirements of the model, improve the efficiency of board making while reducing the cost of board making.

[0068] Corresponding to the foregoing embodiments, such as Figure 4 As shown, the pattern generation model includes a target language parsing model and a target symbol translation model. When generating target garment patterns based on the pattern generation model, various methods can be employed. For example, the user-input garment design language can be determined first, then input into the target language parsing model to be converted into a symbolic pattern-making language; the symbolic pattern-making language is then input into the target symbol translation model for further conversion into drawing commands; finally, the drawing commands (and corresponding drawing parameters) are input into the target pattern-making software, and the target garment pattern output by the software is received. Alternatively, the user-input garment design language can be determined first, then input into the target language parsing model to be converted into a symbolic pattern-making language; the symbolic pattern-making language is then input into the target pattern-making software, and the target garment pattern output by the software is received.

[0069] The following is in conjunction with the appendix Figure 5 and Figure 6 The training process of the target language parsing model and the target symbol translation model is explained in detail.

[0070] like Figure 5 As shown, the training method for the target language parsing model is applied to a model training device, and the method includes steps 501-502. Among them,

[0071] Step 501: Obtain the training dataset and the standard multimodal large model. Each training data in the training dataset includes sample clothing design language and sample symbolic pattern making language.

[0072] Step 502: Fine-tune the standard multimodal large model using the training dataset to obtain the target language parsing model.

[0073] As mentioned earlier, the target language parsing model can be a multimodal large model. The training process for the multimodal large model of the target language parsing model can be based on the constructed standard multimodal large model as the initial language parsing model, and fine-tuning the initial language parsing model of the standard multimodal large model using a pre-collected training dataset.

[0074] For example, a training dataset and a standard multimodal large model can be obtained first. The training dataset can include multiple training data sets, each of which can include, for example, sample clothing design language as data and sample symbolic pattern-making language as labels. For example, for data in the format of "style description-symbol procedure-pattern", the "style description" and "symbol procedure" in this set of data constitute one training data set.

[0075] This approach uses the symbolic pattern-making language of the samples as labels for the sample clothing design language to construct samples that fine-tune the standard multimodal large model. This enables the trained target language parsing model to have powerful analysis and reasoning capabilities for multimodal data. It can output accurate symbolic pattern-making language through comprehensive and detailed analysis of multimodal (i.e., multiple formats) clothing design language, which helps to improve the accuracy and quality of the final generated target clothing patterns.

[0076] Accordingly, embodiments of this application also propose a target language parsing model, which is used for:

[0077] Convert clothing design language into symbolic pattern making language; and / or convert modified clothing design language into modified symbolic pattern making language.

[0078] The specific training process and language conversion process of this model can be found in the previous embodiment, and will not be repeated here.

[0079] like Figure 6 As shown, the training method for the target symbol translation model is applied to a model training device, and the method includes steps 601-602. Among them,

[0080] Step 601: Obtain the initial model and training samples.

[0081] Step 602: Train the initial model using the training samples to obtain the target symbology translation model, which is used to convert symbology language into drawing commands and / or convert modified symbology language into modified drawing commands.

[0082] In one embodiment, when acquiring the initial model and training samples, an initial symbol translation model and the target typography syntax rules of the target typography software can be acquired. Correspondingly, when training the initial model using the training samples, the target typography syntax rules can be used to train the initial symbol translation model. This method of acquiring the target typography syntax rules of the target typography software in real time for model training allows the target symbol translation model to adapt accordingly when the target typography syntax rules change. This further improves the matching between the typography commands generated by the target symbol translation model and the target typography software, thereby enhancing the typography quality of the target typography software.

[0083] In one embodiment, when training the initial symbology model using the target stencil syntax rules, sample stencil languages ​​can be generated according to the target stencil syntax rules, and the initial symbology model can be trained using these sample stencil languages. For example, training the initial symbology model using the target stencil syntax rules can involve generating a large number of sample stencil languages ​​according to the target stencil syntax rules, and then using these generated sample stencil languages ​​to train the constructed initial symbology model. Once the model training is complete, the target symbology model is obtained. This method can improve the accuracy and quality of the trained target symbology model.

[0084] In one embodiment, when acquiring the initial model and training samples, an initial neural network model based on an encoder-decoder (Encoder-Decoder) conversion architecture can be acquired, and data pairs consisting of corresponding sample symbolic stencil languages ​​and sample drawing commands can be acquired. Accordingly, when training the initial model using the training samples, the data pairs can be used as training samples to train the initial neural network model.

[0085] The encoder, for example, can be a neural network encoder, used to convert clothing design language such as natural language, design drafts, and photographs into feature space vector representations. The decoder, for example, can be a neural network decoder, used to parse the feature space vectors into symbolic pattern-making language. The initial neural network model based on the encoder-decoder conversion architecture can be a neural network model based on a "text-to-text" conversion architecture. The input to the target symbolic translation model is the aforementioned symbolic pattern-making language, and the output is the corresponding drawing command. In the training process of the target symbolic translation model, an initial neural network model can first be built based on the encoder-decoder conversion architecture. Then, a large amount of training data can be used to train the initial neural network model. The training data can be a large number of data pairs consisting of corresponding symbolic pattern-making language and sample drawing commands. By training the initial neural network model based on the training data, the target symbolic translation model can be obtained.

[0086] In this embodiment, when the target pattern-making software runs in the garment pattern generation device, the target pattern-making software used can be different. Since different target pattern-making software typically correspond to different syntax formats for pattern-making commands, the sample drawing commands used for model training during the target symbol translation model training process can also be different. Specifically, the sample drawing commands used to train the target symbol translation model can be pattern-making commands based on the syntax format of the target pattern-making software. In this way, different symbol translation models can be trained for different pattern-making software. Therefore, during the usage phase, the appropriate symbol translation model can be selected as the target symbol translation model according to the target pattern-making software to ensure that the uniformly formatted symbolic pattern-making language is converted into drawing commands that the target software can recognize.

[0087] Accordingly, this specification also proposes a target symbol translation model, which is used for:

[0088] Convert symbolic plotting language into drawing commands; and / or convert modified symbolic plotting language into modified drawing commands.

[0089] The specific training process of the model and the process of performing the above conversion can be found in the description of the previous embodiment, and will not be repeated here.

[0090] Corresponding to the foregoing embodiments, this application provides a target language parsing model software, which is used to: convert clothing design language into symbolic pattern making language; and / or convert modified clothing design language into modified symbolic pattern making language.

[0091] Corresponding to the foregoing embodiments, this application provides a target symbol translation model software, which is used to: convert symbolic slab language into drawing commands, and / or convert modified symbolic slab language into modified drawing commands.

[0092] It should be noted that the aforementioned target language parsing model software is the same as the target language parsing model described above, and the aforementioned target symbol translation model software is the same as the target symbol translation model described above. Both are newly proposed software products in this application, used to implement corresponding software functions, and will not be elaborated further. Furthermore, either the aforementioned target language parsing model software or the target symbol translation model software can be a callable independent functional software (such as clothing design software that can call the software through its open calling interface), or it can be a functional plug-in integrated into other software (such as a functional plug-in integrated into clothing design software, which can depend on the clothing design software to run), and will not be elaborated further.

[0093] This application provides a garment pattern generating apparatus, such as... Figure 7 As shown, it includes:

[0094] The design language determination module 701 is used to determine the clothing design language input by the user.

[0095] The garment pattern generation module 702 is used to call the pattern generation model to generate target garment patterns based on the garment design language, including: converting the garment design language into a symbolic pattern-making language, and outputting the target garment pattern according to the symbolic pattern-making language.

[0096] Optionally, the clothing design language includes at least one of the following: natural language, design sketches, and photographs.

[0097] Optionally, the garment pattern generation module 702 is specifically used to: convert the symbolic pattern-making language into drawing commands; and output the target garment pattern according to the drawing commands.

[0098] Optionally, the garment pattern generation module 702 is specifically used to: convert the symbolic pattern-making language into drawing commands corresponding to the target pattern-making software.

[0099] Optionally, the apparatus further includes a software determination unit 703, used to determine the target plate-making software before converting the symbolic plate-making language into drawing commands corresponding to the target plate-making software.

[0100] Optionally, the symbolic typography language includes at least one of the following: structure commands, symbolic programs, and program parameters.

[0101] Optionally, it also includes a typesetting language display module 704, used to: display at least a portion of the symbolic typesetting language to the user.

[0102] Optionally, the garment pattern generation module 702 is specifically used to: output target garment patterns through target pattern making software according to the symbolic pattern making language.

[0103] Optionally, the garment pattern generation module 702 is specifically used to: output the target garment pattern through the target pattern-making software according to the drawing command.

[0104] Optionally, it also includes a modification module 705, used to: determine the modified clothing design language input by the user; convert the modified clothing design language into a modified symbolic pattern-making language; and modify the target clothing pattern according to the modified symbolic pattern-making language.

[0105] Optionally, the modification module 705 is specifically used to: convert the modification symbolic pattern making language into modification drawing commands; and modify the target garment pattern according to the modification drawing commands.

[0106] Optionally, the pattern generation model includes a target language parsing model, which is used to convert the clothing design language into the symbolic pattern-making language.

[0107] Optionally, the pattern generation model includes a target language parsing model, which is used to: convert the clothing design language into the symbolic pattern-making language; and / or convert the modified clothing design language into the modified symbolic pattern-making language.

[0108] Optionally, the plate generation model further includes a target symbol translation model, which is used to convert the symbolic plate-making language into the drawing commands.

[0109] Optionally, the plate generation model further includes a target symbol translation model, which is used to: convert the symbolic plate-making language into the drawing command; and / or, convert the modified symbolic plate-making language into the modified drawing command.

[0110] Optionally, the target language parsing model is a multimodal large model.

[0111] Optionally, it also includes a pattern display unit 706 for displaying the target garment pattern to the user.

[0112] This application provides a training device for a target language parsing model, such as... Figure 8 As shown, it includes:

[0113] The standard model acquisition unit 801 is used to acquire the training dataset and the standard multimodal large model. Each training data in the training dataset includes sample clothing design language and sample symbolic pattern making language.

[0114] The parsing model training unit 802 is used to fine-tune the standard multimodal large model using the training dataset to obtain the target language parsing model.

[0115] This application provides a training device for a target symbol translation model, such as... Figure 9 As shown, it includes:

[0116] Initial model acquisition unit 901 is used to acquire the initial model and training samples;

[0117] The translation model training unit 902 is used to train the initial model using the training samples to obtain the target symbol translation model, which is used to convert symbolic stencil language into drawing commands, and / or convert modified symbolic stencil language into modified drawing commands.

[0118] Optionally, the initial model acquisition unit 901 is specifically used to: acquire the initial symbol translation model and the target typography rules of the target typography software; the translation model training unit 902 is specifically used to: train the initial symbol translation model using the target typography rules.

[0119] Optionally, the translation model training unit 902 is specifically used to: generate a sample template language according to the target template grammar rules, and train the initial symbol translation model using the sample template language.

[0120] Optionally, the initial model acquisition unit 901 is specifically used to: acquire an initial neural network model based on an encoder-decoder conversion architecture, and acquire data pairs consisting of corresponding sample symbolic stencil languages ​​and sample drawing commands; the translation model training unit 902 is specifically used to: use the data pairs as training samples to train the initial neural network model.

[0121] This application provides an electronic device, such as... Figure 10 As shown, it includes: at least one processor 1001; and a memory 1002 communicatively connected to at least one processor 1001; wherein the memory 1002 stores instructions executable by at least one processor 1001, the instructions being executed by at least one processor 1001 to enable at least one processor 1001 to perform the methods in the above embodiments.

[0122] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0123] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0124] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When these computer program instructions are loaded and executed on a device, they generate, in whole or in part, the processes or functions described in the embodiments of this application. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to the device or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, and magnetic tape), an optical medium (e.g., digital video disk (DVD), etc.), or a semiconductor medium (e.g., solid-state drive).

[0125] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0126] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for generating garment patterns, characterized in that, include: Determine the clothing design language input by the user, wherein the clothing design language includes at least one of the following: natural language, design drafts, and photographs; The process of calling the pattern generation model to generate target garment patterns based on the garment design language includes: converting the garment design language into a symbolic pattern-making language, and outputting the target garment pattern according to the symbolic pattern-making language; the symbolic pattern-making language includes at least two of the following: symbolic programs, structural commands, and program parameters, wherein the symbolic programs include program functions, and the structural commands are a set of functions used to define the combination relationship and logical order between different symbolic programs; The pattern generation model includes a target language parsing model, which is a large language model or a multimodal large model, used to convert the clothing design language into the symbolic pattern-making language.

2. The method according to claim 1, characterized in that, The step of outputting the target garment pattern according to the symbolic pattern-making language includes: Convert the symbolic template language into drawing commands; The target garment pattern is output according to the drawing command.

3. The method according to claim 2, characterized in that, The process of converting the symbolic stencil language into drawing commands includes: The symbolic plate-making language is converted into the corresponding drawing commands of the target plate-making software.

4. The method according to claim 3, characterized in that, Also includes: Before converting the symbolic plate-making language into the drawing commands corresponding to the target plate-making software, the target plate-making software is determined.

5. The method according to claim 1, characterized in that, Also includes: The user is shown at least a portion of the symbolic typesetting language.

6. The method according to claim 1, characterized in that, The step of outputting the target garment pattern according to the symbolic pattern-making language includes: The target garment pattern is output using the target pattern-making software based on the symbolic pattern-making language.

7. The method according to claim 2 or 3, characterized in that, The step of outputting the target garment pattern according to the drawing command includes: The target garment pattern is output by the target pattern-making software according to the drawing commands.

8. The method according to claim 1, characterized in that, Also includes: Determine the user's input to modify the clothing design language; The modified clothing design language is converted into a modified symbolic pattern-making language; The target garment pattern is modified according to the modified symbolic pattern-making language.

9. The method according to claim 8, characterized in that, The modification of the target garment pattern according to the modified symbolic pattern-making language includes: The modified symbolic template language is converted into modified drawing commands; The target garment pattern is modified according to the modified drawing command.

10. The method according to claim 8 or 9, characterized in that, The target language parsing model is also used for: Convert the clothing design language into the symbolic pattern-making language; and / or, The modified clothing design language is converted into the modified symbolic pattern-making language.

11. The method according to any one of claims 2-4, characterized in that, The plate generation model also includes a target symbol translation model, which is used for: The symbolic template language is converted into the drawing commands.

12. The method according to claim 9, characterized in that, The plate generation model also includes a target symbol translation model, which is used for: The modified symbolic template language is converted into the modified drawing command.

13. The method according to any one of claims 1-6, 8-9 and 12, characterized in that, Also includes: The target garment pattern is shown to the user.

14. A training method for a target language parsing model, characterized in that, include: Obtain a training dataset and a standard multimodal large model, wherein each training data in the training dataset includes sample clothing design language and sample symbolic pattern making language; The standard multimodal large model is fine-tuned using the training dataset to obtain the target language parsing model as described in any one of claims 1-13.

15. A training method for a target symbol translation model, characterized in that, include: Obtain the initial model and training samples; The initial model is trained using the training samples to obtain the target symbolic translation model, which is used to convert the symbolic pattern-making language as described in any one of claims 1-13 into drawing commands, the drawing commands being used to output the target garment pattern as described in any one of claims 1-13; and / or, the model is used to convert the modified symbolic pattern-making language as described in any one of claims 8-10 into modified drawing commands, the modified drawing commands being used to modify the target garment pattern.

16. The method according to claim 15, characterized in that, The acquisition of the initial model and training samples includes: acquiring the initial symbol translation model and the target typography rules of the target typography software; The step of training the initial model using the training samples includes: training the initial symbol translation model using the target typography rules.

17. The method according to claim 16, characterized in that, The step of training the initial symbol translation model using the target template grammar rules includes: A sample template language is generated based on the target template syntax rules, and then trained using the initial symbol translation model of the sample template language.

18. The method according to claim 17, characterized in that, The process of obtaining the initial model and training samples includes: obtaining an initial neural network model based on an encoder-decoder conversion architecture, and obtaining data pairs consisting of corresponding sample symbolic stencil language and sample drawing commands; The step of training the initial model using the training samples includes: using the data pairs as training samples to train the initial neural network model.

19. A target language parsing model software, characterized in that, The target language parsing model software is used for: Convert the clothing design language as described in any one of claims 1-13 into a symbolic pattern-making language; and / or, The modified clothing design language as described in any one of claims 8-10 is converted into a modified symbolic pattern-making language.

20. A target symbol translation model software, characterized in that, The target symbol translation model software is used for: Convert the symbolic pattern language as described in any one of claims 1-13 into drawing commands; and / or, The modified symbolic panel language as described in any one of claims 8-10 is converted into a modified drawing command.

21. A garment pattern generating device, characterized in that, include: The design language determination module is used to determine the clothing design language input by the user, wherein the clothing design language includes at least one of the following: natural language, design artwork, and photograph; The garment pattern generation module is used to call the pattern generation model to generate target garment patterns based on the garment design language. Specifically, it is used to convert the garment design language into a symbolic pattern-making language and output the target garment pattern according to the symbolic pattern-making language. The symbolic pattern-making language includes at least two of the following: symbolic programs, structural commands, and program parameters. The symbolic programs include program functions, and the structural commands are a set of functions used to define the combination relationship and logical order between different symbolic programs. The pattern generation model includes a target language parsing model, which is a large language model or a multimodal large model, used to convert the clothing design language into the symbolic pattern-making language.

22. A training device for a target language parsing model, characterized in that, include: The standard model acquisition unit is used to acquire the training dataset and the standard multimodal large model. Each training data in the training dataset includes sample clothing design language and sample symbolic pattern making language. A parsing model training unit is used to fine-tune the standard multimodal large model using the training dataset to obtain a target language parsing model as described in any one of claims 1-13.

23. A training device for a target symbol translation model, characterized in that, include: The initial model acquisition unit is used to acquire the initial model and training samples; A translation model training unit is used to train the initial model using the training samples to obtain the target symbol translation model; wherein, the target garment pattern as described in any one of claims 1-13 is output according to a drawing command, the drawing command being obtained by the trained target symbol translation model from the symbolic pattern making language, and / or the trained target symbol translation model is used to convert the modified symbolic pattern making language as described in any one of claims 8-10 into a modified drawing command.

24. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1 to 18.

25. A computer-readable storage medium storing a computer program, characterized in that, The computer program is executed by a processor to implement the method described in any one of claims 1 to 18.

26. A computer program product, characterized in that, The computer program product includes computer program code, which, when executed by a computer device, performs the method described in any one of claims 1 to 18.

Images