Garment processing information identification method and equipment based on data engineering operation system

Abstract

The embodiment of the invention discloses a garment processing information identification method and equipment based on a data engineering operating system, and the method comprises the steps: carrying out the preprocessing of an original garment image uploaded by a client, obtaining a non-deformation garment image, and extracting the image features of the non-deformation garment image; encoding an inspiration label uploaded by a client into a semantic vector, and screening candidate attribute subsets based on the similarity between the semantic vector and each label vector of a preset multi-latitude attribute database; aligning the image features with the label semantic vector based on a cross-modal attention mechanism, generating a joint embedded representation vector, and inputting the joint embedded representation vector and the candidate attribute subset into a preset recognition model to recognize and output initial garment processing information; and performing process completion on the initial garment processing information according to a preset process constraint specification library, detecting the completed initial garment processing information based on a preset manufacturable detection strategy, obtaining the garment processing information passing the detection, and synchronizing the garment processing information to a client.

Description

Technical Field

[0001] This specification relates to the field of automated production technology, and in particular to a method and device for identifying garment processing information based on a data engineering operating system. Background Art

[0002] With the continuous advancement of technology, the apparel industry is transforming towards digitalization and intelligence. Traditional methods for identifying and processing garment processing information often rely on manual labor. However, due to differences in manual experience, technical solutions for the same design vary significantly, leading to significant design misunderstandings. This results in an inefficient and error-prone process for identifying and transmitting garment processing information, making it difficult to meet rapidly changing market demands. Companies urgently need an efficient and accurate method for processing garment processing information to improve production efficiency, reduce costs, enhance product quality, and strengthen their competitiveness in the market.

[0003] Current methods for acquiring garment processing information generally rely on image processing software and CAD software. These simple image analysis tools, along with pre-defined basic rules, extract simple visual features of garment images, such as color histograms and texture features. These features are then matched against a pre-established simple feature library to identify garment information. While this approach reduces reliance on manual labor, relying solely on simple features like color and texture fails to fully and accurately describe complex garment information, making it difficult to effectively identify more in-depth information such as style and craftsmanship. Furthermore, it lacks semantic understanding of garment information, hindering the inherent relationships and semantic meaning between garment style and craftsmanship, making it difficult to perform more detailed information analysis and processing. Furthermore, it lacks adaptive and intelligent learning capabilities, making it difficult to accurately identify the pattern and craftsmanship of innovative garments. Summary of the Invention

[0004] In order to solve the above technical problems, one or more embodiments of this specification provide a method and device for identifying garment processing information based on a data engineering operating system.

[0005] One or more embodiments of this specification adopt the following technical solutions:

[0006] One or more embodiments of this specification provide a method for identifying garment processing information based on a data engineering operating system. The method is performed based on the data engineering operating system and includes:

[0007] Preprocessing the original clothing image uploaded by the client to obtain an invisible clothing image, so as to extract image features of the invisible clothing image;

[0008] Encode the inspiration tag uploaded by the client into a semantic vector, and filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector in the preset multi-dimensional attribute database; wherein the attributes include: style, craftsmanship, fabric, and layout;

[0009] Aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector, and inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model to identify and output initial clothing processing information;

[0010] The initial garment processing information is completed according to a preset process constraint specification library, and the completed initial garment processing information is tested based on a preset manufacturability test strategy, and the garment processing information that passes the test is obtained and synchronized to the client.

[0011] Optionally, in one or more embodiments of the present specification, preprocessing the original clothing image uploaded by the client to obtain the invisible clothing image specifically includes:

[0012] Determining, based on the pixel grayscale values ​​of the original clothing image, texture distribution features corresponding to the original clothing image, and determining, based on the texture distribution features, first image region distribution information and a plurality of second image distribution information of the original clothing image; wherein the texture of the first image region is lower than the texture of the second image region;

[0013] Determining a current clothing type based on the original clothing image uploaded by the client, and obtaining a preset standard template based on the current clothing type; wherein the current clothing type includes: tops, pants, and skirts;

[0014] Extracting clothing key points of the original clothing image based on a convolutional neural network architecture, and constructing a homography matrix corresponding to the original clothing image according to a correspondence between the clothing key points and the preset standard template;

[0015] Performing perspective transformation on the original clothing image according to the homography matrix to obtain an initial non-deformed clothing image;

[0016] A second image region corresponding to the second image distribution information in the initial non-deformable clothing image is extracted through an image mask, and the second image region is smoothed and filtered according to a preset bilateral filter to obtain a filtered second image region. The initial non-deformable clothing image is corrected according to the filtered second image region to obtain a non-deformable clothing image.

[0017] Optionally, in one or more embodiments of this specification, the inspiration tag uploaded by the client is encoded into a semantic vector, so as to screen the candidate attribute subset based on the similarity between the semantic vector and each tag vector of the preset multi-dimensional attribute database, specifically including:

[0018] Based on a preset synonym table, the inspiration tags are replaced to obtain standard inspiration tags, and the standard inspiration tags are encoded to obtain semantic vectors corresponding to the standard inspiration tags;

[0019] Reading the vector representation of each tag in the preset multi-dimensional attribute database to sequentially calculate the similarity between the semantic vector and the vector of each tag in each dimensional attribute database to determine the similarity matrix between the semantic vector and each dimensional attribute database;

[0020] The similarity relationship between the standard inspiration tag and each dimensional attribute database is determined according to the similarity matrix, and each dimensional attribute database is screened according to the similarity relationship to obtain a candidate attribute subset.

[0021] Optionally, in one or more embodiments of this specification, aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector specifically includes:

[0022] Obtaining a feature vector corresponding to the image feature, and mapping the feature vector and the label semantic vector to a recognition dimension corresponding to the preset recognition model based on a linear change;

[0023] Determining the degree of association between each element in the feature vector and each element in the label semantic vector based on a dot product value between each element in the feature vector and each element in the label semantic vector;

[0024] Determining a cross-modal attention weight matrix based on the degree of association between each element in the feature vector and each element in the label semantic vector;

[0025] Adjusting the feature vector and the label semantic vector based on the cross-modal attention weight matrix to obtain an adjusted feature vector and an adjusted label semantic vector;

[0026] The adjusted feature vector and the adjusted label semantic vector are weightedly summed to obtain a joint embedding representation vector.

[0027] Optionally, in one or more embodiments of this specification, before inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model, the method further includes:

[0028] Calling a historical clothing database corresponding to the data engineering operating system to use attribute-annotated clothing images in the historical database as a training data set;

[0029] Performing random cropping of different scales on the historical clothing images in the training data set to obtain partial historical clothing images, and using the partial historical clothing images as new training data to augment the training data set to obtain a first training data set;

[0030] performing a gamma transform on each historical costume image in the expanded training data set, so as to perform a secondary expansion on the first training data set according to the transformed historical costume images to obtain a second training data set;

[0031] Creating a corresponding fabric texture image according to the noise parameters corresponding to each fabric texture, and adding pixel values ​​of the fabric texture image to pixel values ​​of each historical clothing image in the second training data set to obtain an expanded third training data set;

[0032] The visual neural network model is trained based on the third training data to obtain a visual neural network model that meets the requirements as a preset recognition model.

[0033] Optionally, in one or more embodiments of the present specification, the process of completing the initial garment processing information according to the preset process constraint specification library specifically includes:

[0034] Extracting entity information of the initial garment processing information to obtain compliance features corresponding to the initial garment processing information in the preset process constraint specification library based on an entity association relationship between the entity information and the preset process constraint specification library;

[0035] Traversing a preset clothing process knowledge graph based on the compliance features to obtain process nodes associated with each compliance feature and generate a candidate process chain;

[0036] The missing processes of the initial garment processing information are determined according to the screened candidate process chains, so as to complete the process of the initial garment processing information based on the process nodes corresponding to the missing processes.

[0037] Optionally, in one or more embodiments of the present specification, the detecting of the completed initial garment processing information based on a preset manufacturable detection strategy and synchronizing the obtained garment processing information that passes the detection to the client specifically includes:

[0038] Processing the garment processing information based on an intelligent grading engine to obtain structural parameters corresponding to the garment processing information; wherein the structural parameters are BOM data;

[0039] Converting the structural parameters based on corresponding geometric constraint mapping to obtain CAD constraint conditions corresponding to the garment processing information;

[0040] Calling a constraint solver to verify the CAD constraint conditions and obtain a test result; wherein the verification includes: process feasibility verification, process conflict verification and garment piece closure verification;

[0041] If the detection result is that the detection is passed, the processing information of the garment that has passed the detection is synchronized to the client.

[0042] Optionally, in one or more embodiments of this specification, after synchronizing the obtained garment processing information that has passed the inspection to the client, the method further includes:

[0043] Obtaining the garment piece areas corresponding to the garment processing information, and performing descending processing on the garment piece areas based on a greedy strategy to obtain an initial layout of the nesting;

[0044] Calculating the ratio of the total area of ​​all garment pieces to the total area occupied by the initial layout of the nesting material according to the initial layout of the nesting material to determine the fabric utilization rate, using the fabric utilization rate as a fitness function, and optimizing the initial layout of the nesting material according to the fitness function to obtain an optimal nesting material layout;

[0045] Comparing the fabric utilization rate corresponding to the optimal nesting layout with a preset utilization rate threshold, and if the utilization rate is less than the preset utilization rate threshold, triggering a simulated annealing strategy to perform local optimization on the optimal nesting layout to obtain the current nesting layout;

[0046] Receiving order specification information uploaded by the client, comparing the order specification information with the BOM data corresponding to the garment processing information to determine whether there are updated parameters;

[0047] If so, iteratively update the current nesting layout to obtain the nesting layout to be produced;

[0048] The layout of the material to be produced, the order specification information, the clothing processing information and the BOM data are encapsulated to obtain a data packet to be transmitted, so as to send the data packet to be transmitted to the factory end for docking production.

[0049] One or more embodiments of this specification provide a device for identifying garment processing information based on a data engineering operating system, the device comprising:

[0050] at least one processor; and,

[0051] a memory communicatively connected to the at least one processor; wherein,

[0052] 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:

[0053] Preprocessing the original clothing image uploaded by the client to obtain an invisible clothing image, so as to extract image features of the invisible clothing image;

[0054] Encode the inspiration tag uploaded by the client into a semantic vector, and filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector in the preset multi-dimensional attribute database; wherein the attributes include: style, craftsmanship, fabric, and layout;

[0055] Aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector, and inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model to identify and output initial clothing processing information;

[0056] The initial garment processing information is completed according to a preset process constraint specification library, and the completed initial garment processing information is tested based on a preset manufacturability test strategy, and the garment processing information that passes the test is obtained and synchronized to the client.

[0057] Optionally, in one or more embodiments of this specification, the inspiration tag uploaded by the client is encoded into a semantic vector, so as to screen the candidate attribute subset based on the similarity between the semantic vector and each tag vector of the preset multi-dimensional attribute database, specifically including:

[0058] Based on a preset synonym table, the inspiration tags are replaced to obtain standard inspiration tags, and the standard inspiration tags are encoded to obtain semantic vectors corresponding to the standard inspiration tags;

[0059] Reading the vector representation of each tag in the preset multi-dimensional attribute database to sequentially calculate the similarity between the semantic vector and the vector of each tag in each dimensional attribute database to determine the similarity matrix between the semantic vector and each dimensional attribute database;

[0060] The similarity relationship between the standard inspiration tag and each dimensional attribute database is determined according to the similarity matrix, and each dimensional attribute database is screened according to the similarity relationship to obtain a candidate attribute subset.

[0061] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0062] Preprocessing the original garment image to obtain an undistorted garment image eliminates the interference of image deformation on feature extraction, making the extracted image features more realistically reflect the garment itself, thereby providing a more reliable foundation for subsequent information recognition. Encoding inspiration labels into semantic vectors and filtering candidate attribute subsets enables precise matching of relevant candidate attribute subsets based on the customer's personalized inspiration and needs, thereby narrowing the query scope for subsequent recognition and matching. Aligning image features with label semantic vectors based on a cross-modal attention mechanism can capture the complex semantic associations between image and text. The generated joint embedding representation vector integrates multimodal information, obtaining garment-related information from multiple perspectives, no longer limited to a single modality, and can more comprehensively describe the characteristics and attributes of the garment. By completing the process based on a preset process constraint specification library, the gap in process information is filled, making garment processing information more complete and resolving the problem that traditional garment processing information processing methods cannot quickly and accurately respond to customers' personalized customization needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some of the embodiments described in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings:

[0064] Figure 1 A flowchart of a method for identifying garment processing information based on a data engineering operating system provided in an embodiment of this specification;

[0065] Figure 2 A schematic diagram of an upload interface of a client provided in an embodiment of this specification;

[0066] Figure 3 This is a schematic diagram of a clothing processing information interface synchronously obtained by a client provided in an embodiment of this specification;

[0067] Figure 4 A schematic diagram of an interface for order specification information uploaded by a client according to an embodiment of this specification;

[0068] Figure 5 This is a structural diagram of a clothing processing information recognition device based on a data engineering operating system provided in an embodiment of this specification. DETAILED DESCRIPTION

[0069] The embodiments of this specification provide a method and device for identifying garment processing information based on a data engineering operating system.

[0070] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this specification without creative work should fall within the scope of protection of this specification.

[0071] like Figure 1 As shown, in one or more embodiments of this specification, a flow chart of a method for identifying garment processing information based on a data engineering operating system is provided. Figure 1 It can be seen that in one or more embodiments of this specification, a method for identifying garment processing information based on a data engineering operating system includes the following steps:

[0072] S101: Preprocessing the original clothing image uploaded by the client to obtain an invisible clothing image, so as to extract image features of the invisible clothing image.

[0073] The original clothing image may be deformed due to the shooting angle, the state of clothing being worn, or other factors, which may cause the patterns, lines, and other features to be distorted. In order to avoid the influence of inaccurate feature extraction caused by deformation on the judgment of the true information of clothing. In the embodiment of this specification, after obtaining the original clothing image uploaded by the client, it will be preprocessed to eliminate the deformation in the image through preprocessing to obtain an undistorted clothing image, so that the image features extracted based on the undistorted clothing image can more accurately reflect the true attributes of the clothing. Through this process, various clothing images uploaded by different clients can have unified specifications at the feature extraction level, which is convenient for subsequent fusion processing with other modal information, thereby improving the accuracy of clothing processing information recognition.

[0074] Specifically, in one or more embodiments of this specification, preprocessing the original clothing image uploaded by the client to obtain the invisible clothing image specifically includes:

[0075] During the shooting process, the original clothing image will often have different degrees of deformation due to factors such as the shooting angle, the wearing state of the clothing, wrinkles, etc., and the texture in the original clothing image will also interfere with the subsequent analysis and processing due to the shape. Therefore, in order to clarify the image area distribution information of different textures, in the embodiment of this specification, the pixel grayscale value of the original clothing image will be used to determine the texture distribution characteristics corresponding to the original clothing image based on the pixel grayscale value, and the first image area distribution information and multiple second image distribution information of the original clothing image will be determined based on the texture distribution characteristics; wherein the texture of the first image area is lower than the texture of the second image area.

[0076] The current clothing type is then determined based on the original clothing image uploaded by the client, and a preset standard template is obtained based on the current clothing type. It should be noted that the current clothing types include tops, pants, and skirts. Obtaining the corresponding preset standard template based on this allows for the adoption of appropriate processing methods for different clothing types, better addressing the differences between clothing types and improving processing accuracy and adaptability. Subsequently, the original clothing image's key points are extracted using a convolutional neural network architecture. Based on the correspondence between the key points and the preset standard template, a homography matrix corresponding to the original clothing image is constructed. The original clothing image is perspective transformed based on the homography matrix to obtain an initial, undistorted clothing image. This extraction of clothing key points using a convolutional neural network architecture and the construction of a homography matrix based on their correspondence with the preset standard template accurately describe the geometric transformation relationship between the original clothing image and the standard template, providing a reliable basis for perspective transformation and effectively correcting image deformation. To address the interference of multiple textured regions and improve subsequent recognition accuracy, an image mask is used to extract a second image region within the initial non-deformable clothing image that corresponds to the second image distribution information. This second image region is then smoothed using a preset bilateral filter to obtain a filtered second image region. The initial non-deformable clothing image is then corrected based on the filtered second image region to obtain a non-deformable clothing image. This process first uses perspective transformation to obtain the initial non-deformable clothing image, and then smoothing and correcting specific second image regions. This step-by-step approach eliminates deformation while further optimizing image quality, reducing noise and irregular textures in the image, and resulting in a final non-deformable clothing image that more clearly and accurately reflects the clothing's true form.

[0077] S102: Encode the inspiration tag uploaded by the client into a semantic vector to filter a subset of candidate attributes based on the similarity between the semantic vector and each tag vector in a preset multi-dimensional attribute database; wherein the attributes include: style, craftsmanship, fabric, and layout.

[0078] In the embodiment of this specification, after obtaining the image features based on the above steps, since searching in a database with a large amount of data will consume a lot of analysis costs, in order to narrow the search scope and improve the accuracy, the client-uploaded image features will be obtained at the same time. Figure 2The inspiration tag encoding shown in the figure is then encoded into semantic vectors uploaded by the client. By converting the originally complex and diverse text tags that are difficult to process directly into digital vectors that computers can understand and process, the data is quantified and standardized, facilitating subsequent calculations and analysis. Furthermore, by calculating the similarity between the semantic vectors and the label vectors in the preset multi-dimensional attribute database, the association between the inspiration tag and attribute tags such as style, craft, fabric, and layout in the database can be discovered at a semantic level. In this way, based on the semantic connotation of the inspiration tag, a subset of candidate attributes with high relevance can be screened, providing targeted information for further identification and improving the efficiency of database utilization.

[0079] Specifically, in one or more embodiments of this specification, the inspiration tag uploaded by the client is encoded into a semantic vector, and the candidate attribute subset is screened based on the similarity between the semantic vector and each tag vector of the preset multi-dimensional attribute database, specifically including:

[0080] First, based on the preset synonym table, the inspiration tags uploaded by the client are replaced. The purpose is to unify inspiration tags with different expressions but the same or similar semantics into a standard expression form to reduce the interference caused by the diversity of language expressions. For example, if "long coat" and "long coat" are synonyms, they can be unified into "long coat" through the replacement of the synonym table to obtain a standard inspiration tag. Then, the standard inspiration tag is encoded. Here, the word vector technology in natural language processing may be used to convert each standard inspiration tag into a corresponding semantic vector. It can be understood that the semantic vector is a numerical representation that can reflect the semantic features and semantic relationships of the label in the vector space, which is convenient for subsequent similarity calculations.

[0081] Next, the vector representations of each label in the pre-set multi-dimensional attribute database are read to sequentially calculate the similarity between the semantic vector of each standard inspiration label and the label vectors in each dimensional attribute database. Various similarity calculation methods exist, such as cosine similarity and Euclidean distance, which will not be discussed further here. Through these calculations, a similarity matrix is ​​determined between the semantic vector and each dimensional attribute database. It should be noted that the rows of this matrix can represent standard inspiration labels, and the columns can represent labels in each dimensional attribute database. The matrix elements represent the similarity scores between the corresponding standard inspiration labels and the attribute database labels, thus visually presenting the similarity relationships between them in the form of a matrix. Based on the resulting similarity matrix, the similarity relationships between the standard inspiration labels and each dimensional attribute database are then analyzed and determined. For example, a similarity threshold can be set; when the similarity score exceeds this threshold, the standard inspiration label is considered to have a strong similarity with the corresponding attribute database label. Based on the similarity relationships, each dimensional attribute database is screened, retaining attribute labels with a high degree of similarity to the standard inspiration label and forming a candidate attribute subset. This candidate attribute subset includes multi-dimensional attribute information such as style, craftsmanship, fabric, layout, etc. that are semantically related to the inspiration tag.

[0082] S103: Aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector, and inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model to identify and output initial clothing processing information.

[0083] In the above steps S101-S102, the original clothing image uploaded by the client has been preprocessed and the image features of the invisible clothing image have been extracted. The inspiration tag uploaded by the client has been encoded to obtain a semantic vector. Since images and text belong to different modalities and there are differences in the information between them, in order to automatically learn the association between image features and label semantic vectors, the problem of low recognition accuracy and difficulty in recognizing complex clothing when the existing recognition is based only on a single image feature is avoided. In the embodiment of this specification, the association between image features and label semantic vectors is automatically learned through a cross-modal attention mechanism, and then the image features and label semantic vectors are fused to generate a joint embedding representation vector. This vector integrates the information of the image and text, and can more comprehensively and accurately represent the characteristics and semantics of the clothing. In addition, by calculating the similarity between the inspiration label semantic vector and the label vectors of the preset multi-dimensional attribute database, a candidate attribute subset is screened. This subset contains attribute information such as style, craftsmanship, fabric, and layout related to the inspiration tag. Therefore, the joint embedding representation vector and the candidate attribute subset are provided as input to the preset recognition model. The model will analyze and process this input information, use its internal parameters and structure to learn the mapping relationship between the input information and clothing processing information, and then identify and output the initial clothing processing information. This information may include specific clothing style details, required processing technology, applicable fabric types, pattern sizes, etc., providing a basis and guidance for subsequent clothing processing and production.

[0084] Specifically, in one or more embodiments of this specification, the image features and the label semantic vectors are aligned based on the cross-modal attention mechanism to generate a joint embedding representation vector, which can be specifically achieved through the following process:

[0085] First, the feature vectors corresponding to the clothing image features obtained through preprocessing and feature extraction are obtained. The previously encoded label semantic vector is also present. Since the image feature vector and label semantic vector may have different dimensions, and subsequent operations must be performed in the same dimensions, they are mapped to the recognition dimensions corresponding to the preset recognition model using a linear transformation. This allows the two different modal vectors to be compared and integrated in the same space, laying the foundation for subsequent calculations and processing. For example, the image feature vector is originally 512-dimensional, and the label semantic vector is 768-dimensional. A linear transformation converts both to the 1024-dimensional dimension required by the recognition model. The degree of correlation between each element in the feature vector and each element in the label semantic vector is then determined based on the dot product between the elements in the feature vector and the label semantic vector (that is, by calculating the sum of the products of the corresponding elements of the two vectors). Based on this correlation, a cross-modal attention weight matrix is ​​then determined. The feature vector and label semantic vector are then adjusted based on the cross-modal attention weight matrix. Specifically, each element of the feature vector is multiplied by the corresponding weight value in the attention weight matrix to obtain an adjusted feature vector; similarly, the same operation is performed on the label semantic vector to obtain an adjusted label semantic vector. In this way, during the fusion process, elements that are more closely associated with other modalities are given greater weights, thereby highlighting this important information. The adjusted feature vector is weighted and summed with the adjusted label semantic vector to obtain the final joint embedding representation vector. This joint embedding representation vector fuses the information of image features and label semantic vectors, and through the attention mechanism, it highlights the correlation between the two. It can more effectively represent the comprehensive information of clothing and provide higher-quality input data for subsequent input into the preset recognition model to accurately identify initial clothing processing information.

[0086] Furthermore, in one or more embodiments of this specification, in order to improve the accuracy of process feature recognition, before inputting the joint embedding representation vector and the candidate attribute subset into the preset recognition model, the method further includes the following process:

[0087] The historical clothing database corresponding to the data engineering operating system is called, and attribute-labeled clothing images are extracted from the database as the initial training data set. For example, in a certain scenario, clothing images are labeled with attribute labels including style, silhouette, craftsmanship, fabric color, and material. Then, in order to enable the model to learn the characteristics of clothing images in different local situations and improve the model's adaptability to changes in various clothing images, random cropping operations of different scales are performed on the historical clothing images in the training data set. In this way, various local historical clothing images can be obtained, and these local images contain different parts and features of the original images. These local historical clothing images are added as new training data to the original training data set to form the first training data set.

[0088] Then, gamma transforms are performed on each historical clothing image in the expanded first training dataset, allowing the first training dataset to be expanded again based on the transformed historical clothing images to generate a second training dataset. This further enriches the diversity of the training data, enabling the model to learn the characteristics of clothing images with different contrast and brightness levels, and enhancing the model's robustness to conditions such as image illumination variations. Fabric texture images are also created based on the noise parameters corresponding to each fabric texture. These noise parameters describe the characteristics and distribution of different fabric textures. Fabric texture images generated using these parameters have a texture similar to that of actual fabrics. The pixel values ​​of the fabric texture images are then added to the pixel values ​​of each historical clothing image in the second training dataset to generate an expanded third training dataset. This operation simulates the fabric texture interference that may occur in actual clothing production and photography, allowing the model to learn how to handle images with fabric texture noise during training, improving the model's recognition capabilities in complex real-world scenarios. The visual neural network model is trained with the expanded third training data set. During the training process, the model continuously learns the image features and corresponding attribute annotation information in the third training data set, adjusts its own parameters, and enables the model to accurately map the image features to the corresponding clothing attributes. After multiple iterative training, when the performance of the model reaches a certain standard, a visual neural network model that meets the requirements can be obtained as a preset recognition model.

[0089] S104: The initial garment processing information is completed according to a preset process constraint specification library, and the completed initial garment processing information is tested based on a preset manufacturability test strategy, and the garment processing information that passes the test is obtained and synchronized to the client.

[0090] The initial garment processing information is obtained by processing and identifying multiple sources of information such as garment image features and label semantic vectors in the early stage. However, this information may be incomplete and cannot be directly used to guide actual production. Therefore, the initial garment processing information is supplemented according to the preset process constraint specification library, and then the supplemented initial garment processing information is tested based on the preset manufacturability detection strategy to determine its feasibility. If the test is passed, the garment processing information that has passed the test is synchronized to the client. Figure 3 It should be noted that the preset process constraint specification library is a database containing a large amount of clothing process knowledge and standard rules, and stores physical / aesthetic constraint relationships between styles, fabrics and processes, such as the constraint relationship of "lock stitch is prohibited on silk fabrics".

[0091] Specifically, in one or more embodiments of this specification, the process of completing the initial garment processing information according to the preset process constraint specification library includes the following steps:

[0092] The initial garment processing information contains various types of entity information, such as clothing style (e.g., shirt, dress), fabric (e.g., pure cotton, silk), and cut (e.g., slim fit, loose fit). This entity information forms the basis for subsequent process completion, so key entity content can be extracted from the initial garment processing information. Each entity is then associated with a series of specific process characteristics and requirements. For example, "pure cotton fabric" may be associated with "washing process" and "softener treatment process." Therefore, based on the entity association relationship between the entity information and the pre-set process constraint specification library, the compliance features corresponding to the initial garment processing information within the pre-set process constraint specification library can be retrieved. Based on the compliance features, the pre-set garment process knowledge graph is traversed to obtain the process nodes associated with each compliance feature, generating a candidate process chain. The selected candidate process chains are then used to identify the missing processes in the initial garment processing information. The initial garment processing information is then process-completed based on the process nodes corresponding to the missing processes. For example, if the initial garment processing information only mentions cutting and sewing, but the candidate process chain also includes ironing and quality inspection, then ironing and quality inspection are the missing processes. Based on the process nodes corresponding to the missing processes, the missing process information is supplemented into the initial garment processing information. The supplementary content includes detailed information such as the specific process requirements, operation steps, and required equipment, making the initial garment processing information more complete and providing comprehensive guidance for actual garment production.

[0093] Specifically, in one or more embodiments of the present specification, the completed initial garment processing information is tested based on a preset manufacturable testing strategy, and the garment processing information that passes the test is obtained and synchronized to the client, specifically including the following process:

[0094] First, the garment processing information is processed using the intelligent grading engine to obtain the structured parameters corresponding to the garment processing information. For example, in certain application scenarios, the intelligent grading engine automatically generates specification data and a bill of materials (BOM) for structured data output. These structured parameters are then converted based on the corresponding geometric constraint mapping to obtain CAD constraints corresponding to the garment processing information. For example, the process "\text{Hem area} = \text{Cuff edge inward offset 30mm}\quad\Rightarrow\quad\text{CAD constraint: Offset(sleeve_edge,-30mm)}" can be used to convert this into a CAD constraint in the form of a CAD geometric constraint equation. The constraint solver is then invoked to verify the CAD constraints and obtain test results. This verification includes process feasibility, process conflict, and garment closure verification. Specifically, the constraint solver can be invoked to verify geometric feasibility, outputting a conflict report (e.g., "Hem exceeds garment boundary") and recommending parameter adjustment ranges. As for the verification of garment piece closure, the Euler characteristic of the garment piece contour edge can be calculated based on the set topology. If the formula is satisfied: χ = V-E + F = 1 (closed contour standard value), where V (number of vertices), E (number of edges), and F (number of faces), it is determined to be closed; otherwise, contour repair based on the moving least squares method (MLS) is triggered; or based on boundary loop traversal, that is, the connectivity of the boundary is detected through depth-first search. If there is an unclosed loop, the gap position is marked and a repair path is generated. If the test result is passed, the garment processing information that has passed the test is synchronized to the client.

[0095] Furthermore, in one or more embodiments of this specification, after obtaining and synchronizing the garment processing information that has passed the inspection to the client, the method further includes the following process:

[0096] After obtaining the garment processing information that has passed the inspection and synchronizing it to the client, first obtain the area of ​​each garment piece corresponding to the garment processing information. Then, based on the greedy strategy, the garment piece areas are processed in descending order, that is, the garment pieces are arranged in order from large to small. According to the initial layout of the nesting, calculate the ratio of the total area of ​​all garment pieces to the total area occupied by the initial layout area of ​​the nesting. This ratio is the fabric utilization rate. Use the fabric utilization rate as the fitness function, and optimize the initial layout of the nesting according to the fitness function to obtain the optimal nesting layout. Then, compare the fabric utilization rate corresponding to the optimal nesting layout with the preset utilization rate threshold. If it is less than the preset utilization rate threshold, then trigger the simulated annealing strategy to perform local optimization on the optimal nesting layout to obtain the current nesting layout. Then, receive the information uploaded by the client such as Figure 4The order specifications are compared with the BOM data corresponding to the garment processing information to determine if any updated parameters exist. For example, if a user's waist size needs to be increased, the corresponding BOM data needs to be updated. This means that the BOM data, CAD patterns, and markers will be recalculated when the customer selects a fixed-volume custom order. If this comparison confirms that updated parameters exist, the current marker layout is iteratively updated. This involves readjusting the position and rotation angle of the garment pieces based on the updated parameters to obtain a marker layout that meets the new specifications. This iterative update ensures that the marker layout always matches the customer's order specifications. The marker layout, order specifications, garment processing information, and BOM data are then packaged into a data packet to be transmitted. This data packet is then sent to the factory for production integration. Upon receipt of the data packet, the factory can produce the garments based on the information contained therein, ensuring that the garments meet the customer's requirements and quality standards. This process receives the order specifications uploaded by the client, compares them with the BOM data to determine updated parameters, and then recalculates the BOM data, CAD patterns, and markers when the customer selects a fixed-volume custom order. This approach allows for rapid response to individual customer needs, ensuring that garments produced meet their specific requirements and improving customer satisfaction. Furthermore, by iteratively updating the current nesting layout and readjusting the position and rotation angle of garment pieces, the nesting layout consistently matches the customer's order specifications. This dynamic adjustment approach accommodates potential changes during the production process, ensuring continuous and accurate production. The nesting layout to be produced, order specifications, garment processing information, and BOM data are then packaged into a data packet to be transmitted to the factory. The factory can then produce based on the detailed information contained in the data packet, avoiding errors or omissions in information transmission, ensuring that garments meet customer requirements and quality standards, and reducing production errors and rework costs.

[0097] like Figure 5 As shown, one or more embodiments of this specification provide a structural diagram of a clothing processing information recognition device based on a data engineering operating system. Figure 5 As shown in one or more embodiments of this specification, a clothing processing information recognition device based on a data engineering operating system includes:

[0098] at least one processor; and,

[0099] a memory communicatively connected to the at least one processor; wherein,

[0100] 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:

[0101] Preprocessing the original clothing image uploaded by the client to obtain an invisible clothing image, so as to extract image features of the invisible clothing image;

[0102] Encode the inspiration tag uploaded by the client into a semantic vector, and filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector in the preset multi-dimensional attribute database; wherein the attributes include: style, craftsmanship, fabric, and layout;

[0103] Aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector, and inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model to identify and output initial clothing processing information;

[0104] The initial garment processing information is completed according to a preset process constraint specification library, and the completed initial garment processing information is tested based on a preset manufacturability test strategy, and the garment processing information that passes the test is obtained and synchronized to the client.

[0105] Furthermore, in one or more embodiments of this specification, the inspiration tag uploaded by the client is encoded into a semantic vector, so as to screen the candidate attribute subset based on the similarity between the semantic vector and each tag vector of the preset multi-dimensional attribute database, specifically including:

[0106] Based on a preset synonym table, the inspiration tags are replaced to obtain standard inspiration tags, and the standard inspiration tags are encoded to obtain semantic vectors corresponding to the standard inspiration tags;

[0107] Reading the vector representation of each tag in the preset multi-dimensional attribute database to sequentially calculate the similarity between the semantic vector and the vector of each tag in each dimensional attribute database to determine the similarity matrix between the semantic vector and each dimensional attribute database;

[0108] The similarity relationship between the standard inspiration tag and each dimensional attribute database is determined according to the similarity matrix, and each dimensional attribute database is screened according to the similarity relationship to obtain a candidate attribute subset.

[0109] The various embodiments in this specification are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simplified. For relevant details, refer to the descriptions of the method embodiments.

[0110] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] The foregoing description is merely one or more embodiments of this specification and is not intended to limit this specification. It will be apparent to those skilled in the art that various modifications and variations may be made to one or more embodiments of this specification. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of one or more embodiments of this specification are intended to be within the scope of the claims of this specification.

Claims

1. A method for identifying garment processing information based on a data engineering operating system, characterized in that: The method is performed based on a data engineering operating system, and includes: Preprocessing the original clothing image uploaded by the client to obtain an invisible clothing image, so as to extract image features of the invisible clothing image; Encode the inspiration tag uploaded by the client into a semantic vector, and filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector in the preset multi-dimensional attribute database; wherein the attributes include: style, craftsmanship, fabric, and layout; Aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector, and inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model to identify and output initial clothing processing information; The initial garment processing information is completed according to a preset process constraint specification library, and the completed initial garment processing information is tested based on a preset manufacturability test strategy, and the garment processing information that passes the test is obtained and synchronized to the client.

2. The method for identifying garment processing information based on a data engineering operating system according to claim 1, characterized in that: Preprocess the original clothing image uploaded by the client to obtain the invisible clothing image, specifically including: Determining, based on the pixel grayscale values ​​of the original clothing image, texture distribution features corresponding to the original clothing image, and determining, based on the texture distribution features, first image region distribution information and a plurality of second image distribution information of the original clothing image; wherein the texture of the first image region is lower than the texture of the second image region; Determining a current clothing type based on the original clothing image uploaded by the client, and obtaining a preset standard template based on the current clothing type; wherein the current clothing type includes: tops, pants, and skirts; Extracting clothing key points of the original clothing image based on a convolutional neural network architecture, and constructing a homography matrix corresponding to the original clothing image according to a correspondence between the clothing key points and the preset standard template; Performing perspective transformation on the original clothing image according to the homography matrix to obtain an initial non-deformed clothing image; A second image region corresponding to the second image distribution information in the initial non-deformable clothing image is extracted through an image mask, and the second image region is smoothed and filtered according to a preset bilateral filter to obtain a filtered second image region. The initial non-deformable clothing image is corrected according to the filtered second image region to obtain a non-deformable clothing image.

3. The method for identifying garment processing information based on a data engineering operating system according to claim 1, characterized in that: The encoding of the inspiration tag uploaded by the client into a semantic vector to filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector of the preset multi-dimensional attribute database specifically includes: Based on a preset synonym table, the inspiration tags are replaced to obtain standard inspiration tags, and the standard inspiration tags are encoded to obtain semantic vectors corresponding to the standard inspiration tags; Reading the vector representation of each tag in the preset multi-dimensional attribute database to sequentially calculate the similarity between the semantic vector and the vector of each tag in each dimensional attribute database to determine the similarity matrix between the semantic vector and each dimensional attribute database; The similarity relationship between the standard inspiration tag and each dimensional attribute database is determined according to the similarity matrix, and each dimensional attribute database is screened according to the similarity relationship to obtain a candidate attribute subset.

4. The method for identifying garment processing information based on a data engineering operating system according to claim 1, characterized in that: The method of aligning image features and label semantic vectors based on the cross-modal attention mechanism to generate a joint embedding representation vector specifically includes: Obtaining a feature vector corresponding to the image feature, and mapping the feature vector and the label semantic vector to a recognition dimension corresponding to the preset recognition model based on a linear change; Determining the degree of association between each element in the feature vector and each element in the label semantic vector based on a dot product value between each element in the feature vector and each element in the label semantic vector; Determining a cross-modal attention weight matrix based on the degree of association between each element in the feature vector and each element in the label semantic vector; Adjusting the feature vector and the label semantic vector based on the cross-modal attention weight matrix to obtain an adjusted feature vector and an adjusted label semantic vector; The adjusted feature vector and the adjusted label semantic vector are weightedly summed to obtain a joint embedding representation vector.

5. The method for identifying garment processing information based on a data engineering operating system according to claim 1, characterized in that: Before inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model, the method further includes: Calling a historical clothing database corresponding to the data engineering operating system to use attribute-annotated clothing images in the historical database as a training data set; Performing random cropping of different scales on the historical clothing images in the training data set to obtain partial historical clothing images, and using the partial historical clothing images as new training data to augment the training data set to obtain a first training data set; performing a gamma transform on each historical costume image in the expanded training data set, so as to perform a secondary expansion on the first training data set according to the transformed historical costume images to obtain a second training data set; Creating a corresponding fabric texture image according to the noise parameters corresponding to each fabric texture, and adding pixel values ​​of the fabric texture image to pixel values ​​of each historical clothing image in the second training data set to obtain an expanded third training data set; The visual neural network model is trained based on the third training data to obtain a visual neural network model that meets the requirements as a preset recognition model.

6. The method for identifying garment processing information based on a data engineering operating system according to claim 1, characterized in that: The process of completing the initial garment processing information according to the preset process constraint specification library specifically includes: Extracting entity information of the initial garment processing information to obtain compliance features corresponding to the initial garment processing information in the preset process constraint specification library based on an entity association relationship between the entity information and the preset process constraint specification library; Traversing a preset clothing process knowledge graph based on the compliance features to obtain process nodes associated with each compliance feature and generate a candidate process chain; The missing processes of the initial garment processing information are determined according to the screened candidate process chains, so as to complete the process of the initial garment processing information based on the process nodes corresponding to the missing processes.

7. The method for identifying garment processing information based on a data engineering operating system according to claim 1, characterized in that: The method of testing the completed initial garment processing information based on the preset manufacturability detection strategy and synchronizing the obtained garment processing information that passes the detection to the client specifically includes: Processing the garment processing information based on an intelligent grading engine to obtain structural parameters corresponding to the garment processing information; wherein the structural parameters are BOM data; Converting the structural parameters based on corresponding geometric constraint mapping to obtain CAD constraint conditions corresponding to the garment processing information; Calling a constraint solver to verify the CAD constraint conditions and obtain a test result; wherein the verification includes: process feasibility verification, process conflict verification and garment piece closure verification; If the detection result is that the detection is passed, the processing information of the clothing that has passed the detection is synchronized to the client.

8. The method for identifying garment processing information based on a data engineering operating system according to claim 7, characterized in that: After the obtained garment processing information that has passed the inspection is synchronized to the client, the method further includes: Obtaining the garment piece areas corresponding to the garment processing information, and performing descending processing on the garment piece areas based on a greedy strategy to obtain an initial layout of the nesting; Calculating the ratio of the total area of ​​all garment pieces to the total area occupied by the initial layout of the nesting material according to the initial layout of the nesting material to determine the fabric utilization rate, using the fabric utilization rate as a fitness function, and optimizing the initial layout of the nesting material according to the fitness function to obtain an optimal nesting material layout; Comparing the fabric utilization rate corresponding to the optimal nesting layout with a preset utilization rate threshold, and if the utilization rate is less than the preset utilization rate threshold, triggering a simulated annealing strategy to perform local optimization on the optimal nesting layout to obtain the current nesting layout; Receiving order specification information uploaded by the client, comparing the order specification information with the BOM data corresponding to the garment processing information to determine whether there are updated parameters; If so, iteratively update the current nesting layout to obtain the nesting layout to be produced; The layout of the material to be produced, the order specification information, the clothing processing information and the BOM data are encapsulated to obtain a data packet to be transmitted, so as to send the data packet to be transmitted to the factory end for docking production.

9. A clothing processing information recognition device based on a data engineering operating system, characterized in that: The device comprises: 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: Preprocessing the original clothing image uploaded by the client to obtain an invisible clothing image, so as to extract image features of the invisible clothing image; Encode the inspiration tag uploaded by the client into a semantic vector, and filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector in the preset multi-dimensional attribute database; wherein the attributes include: style, craftsmanship, fabric, and layout; Aligning image features and label semantic vectors based on a cross-modal attention mechanism to generate a joint embedding representation vector, and inputting the joint embedding representation vector and the candidate attribute subset into a preset recognition model to identify and output initial clothing processing information; The initial garment processing information is completed according to a preset process constraint specification library, and the completed initial garment processing information is tested based on a preset manufacturability test strategy, and the garment processing information that passes the test is obtained and synchronized to the client.

10. The clothing processing information recognition device based on a data engineering operating system according to claim 9, characterized in that: The encoding of the inspiration tag uploaded by the client into a semantic vector to filter the candidate attribute subset based on the similarity between the semantic vector and each tag vector of the preset multi-dimensional attribute database specifically includes: Based on a preset synonym table, the inspiration tags are replaced to obtain standard inspiration tags, and the standard inspiration tags are encoded to obtain semantic vectors corresponding to the standard inspiration tags; Reading the vector representation of each tag in the preset multi-dimensional attribute database to sequentially calculate the similarity between the semantic vector and the vector of each tag in each dimensional attribute database to determine the similarity matrix between the semantic vector and each dimensional attribute database; The similarity relationship between the standard inspiration tag and each dimensional attribute database is determined according to the similarity matrix, and each dimensional attribute database is screened according to the similarity relationship to obtain a candidate attribute subset.

Images