Clothing design parameter recommendation method, device, equipment and computer program product

Abstract

The application is suitable for the technical field of garment design, and particularly relates to a garment design parameter recommendation method, device, equipment and computer program product, which comprises the following steps: obtaining user body shape data and candidate garment pattern features of a target user, and establishing an adaptation difference matrix and a style correlation matrix; determining initial anomaly detection parameters based on the adaptation difference matrix and the style correlation matrix, which reflect the adaptation deviation of different pattern parameters under the individualized demand of the user; then, the initial adaptation scores are screened and removed in combination with the anomaly detection parameters, so as to further optimize the precision of pattern matching; and finally, a final garment pattern parameter recommendation scheme is generated according to the retained adaptation scores, so as to realize personalized garment design recommendation for the target user, improve the precision and individualization level of garment pattern recommendation, reduce the blindness of the recommendation process and the dependence on the experience of designers, and comprehensively and systematically improve the efficiency of garment design parameter recommendation and user satisfaction.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of clothing design, and particularly relates to a clothing design parameter recommendation method, device, equipment and computer program product. BACKGROUND

[0002] Clothing design, as an important link to meet individual aesthetic, comfort and functional needs, plays an important role in many fields such as clothing manufacturing, e-commerce, personalized customization, etc. The reasonable design of clothing pattern parameters directly affects the wearing comfort, appearance effect and user shopping experience of clothing. Therefore, scientifically recommending clothing pattern design parameters is an important way to improve clothing design efficiency, reduce design cost and enhance user satisfaction.

[0003] Currently, the recommendation of clothing pattern design mainly relies on traditional experience or manual adjustment. Designers usually match clothing patterns for different user groups based on accumulated experience from previous designs, or through methods such as questionnaire survey, size statistics, single factor analysis, etc. These methods can generate clothing patterns suitable for most users to some extent, but have obvious limitations. Traditional methods usually ignore the dynamic influence of individual differences of users on clothing pattern adaptability. Differences in body data, dressing style preferences and perception of pattern details of different users can lead to difficulties in meeting individual needs through single experience or static recommendation, thereby affecting the accuracy of clothing design and user satisfaction. SUMMARY

[0004] The embodiments of the present application provide a clothing design parameter recommendation method, device, equipment and computer program product, which can solve the problem that traditional clothing design ignores the dynamic influence of individual differences of users on clothing pattern adaptability, leading to difficulties in meeting individual needs through single experience or static recommendation, thereby affecting the accuracy of clothing design and user satisfaction.

[0005] In a first aspect, the embodiments of the present application provide a clothing design parameter recommendation method, comprising:

[0006] determining an initial adaptation score between user body data of a target user and a plurality of candidate clothing patterns, and establishing an adaptation difference matrix of a plurality of the initial adaptation scores; wherein the initial adaptation score corresponds to the candidate clothing pattern one by one;

[0007] obtaining dressing style preferences of the target user, and constructing a style correlation matrix between the initial adaptation scores based on the dressing style preferences of the target user and style features of a plurality of the candidate clothing patterns;

[0008] calculate an anomaly detection parameter of the initial fitting scores based on the fitting difference matrix, the style correlation matrix and the initial fitting scores; wherein the anomaly detection parameter is used to determine whether there is an abnormal score in the initial fitting scores;

[0009] when the anomaly detection parameter indicates that there is an abnormal score in the initial fitting scores, eliminate the abnormal score;

[0010] generate a final garment pattern parameter recommendation scheme of the target user according to all the retained initial fitting scores.

[0011] The technical solutions provided in the embodiments of the present application have at least the following technical effects:

[0012] The garment design parameter recommendation method provided in the embodiments of the present application obtains user body shape data, dressing style preference and candidate garment pattern features of a target user, and establishes a fitting difference matrix and a style correlation matrix, thereby starting the entire garment pattern recommendation process. The initial anomaly detection parameter is determined based on the fitting difference matrix and the style correlation matrix, which can reflect the fitting deviation of different pattern parameters under the individualized needs of the user. In combination with the anomaly detection parameter, the initial fitting scores are screened and eliminated, thereby further optimizing the accuracy of pattern matching. The final garment pattern parameter recommendation scheme is generated according to the retained fitting scores, thereby realizing personalized garment design recommendation for the target user. Not only a single pattern parameter is recommended, but also the entire pattern system can be synergistically optimized based on the correlation analysis of the body shape data, style preference and pattern parameters, thereby improving the accuracy and personalization level of garment pattern recommendation, reducing the blindness of the recommendation process and the dependence on the experience of designers, and comprehensively and systematically improving the efficiency and user satisfaction of garment design parameter recommendation.

[0013] In a second aspect, the embodiments of the present application provide a garment design parameter recommendation device, comprising:

[0014] An initial unit is configured to determine initial fitting scores between user body shape data of a target user and a plurality of candidate garment patterns, and establish a fitting difference matrix of the initial fitting scores; wherein the initial fitting scores correspond to the candidate garment patterns one by one.

[0015] A construction unit is configured to obtain dressing style preference of the target user, and construct a style correlation matrix between the initial fitting scores based on the dressing style preference of the target user and style features of the plurality of candidate garment patterns.

[0016] The detection unit is configured to calculate an anomaly detection parameter of the initial fitting scores based on the adaptation difference matrix, the style correlation matrix, and the initial fitting scores, wherein the anomaly detection parameter is used to determine whether there is an abnormal score in the initial fitting scores.

[0017] The screening unit is configured to eliminate the abnormal score when the anomaly detection parameter indicates that there is an abnormal score in the initial fitting scores.

[0018] The generating unit is configured to generate a final garment pattern parameter recommendation scheme of the target user according to all the retained initial fitting scores.

[0019] In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method according to any one of the above aspects when executing the computer program.

[0020] In a fourth aspect, a computer program product is provided, which, when executed on an electronic device, causes the electronic device to perform the method according to any one of the above aspects.

[0021] It can be understood that the beneficial effects of the above-mentioned second aspect to fourth aspect can be referred to the related description in the above aspects, which will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0023] Figure 1 is a flowchart of a garment design parameter recommendation method provided by an embodiment of the present application;

[0024] Figure 2 is a part of the interface diagram of the garment design parameter recommendation method provided by an embodiment of the present application;

[0025] Figure 3 is a part of the interface diagram of the garment design parameter recommendation method provided by an embodiment of the present application;

[0026] Figure 4 is a principle diagram of step S400 of the garment design parameter recommendation method provided by an embodiment of the present application;

[0027] Figure 5is a structural schematic diagram of a garment design parameter recommendation device provided by an embodiment of the present application;

[0028] Figure 6 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0029] In the following description, for the purposes of explanation and not limitation, specific details are set forth, such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the application. However, it will be apparent to those skilled in the art that the application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods, devices, circuits, and

[0030] It should be understood that the term “includes” when used in the specification and the appended claims herein, specifies the presence of stated features, integers, steps, operations, elements, and / or components but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0031] It should also be understood that the term “and / or” when used in the specification and the appended claims herein, means any one or more of the associated listed items or a combination thereof.

[0032] As used in the description of the application and the appended claims herein, the term “if’ can be interpreted as meaning “when” or “upon” or “in response to a determination” or “in response to a detection” depending on the context. Similarly, the phrase “if it is determined” or “if a described condition or event is detected” can be interpreted as meaning “upon a determination” or “in response to a determination” or “upon a detection of a described condition or event” or “in response to a detection of a described condition or event” depending on the context.

[0033] In addition, in the description of the application and the appended claims herein, the terms “first”, “second”, “third”, etc. are used only to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

[0034] Reference within the specification of this application to "one embodiment" or "some embodiments" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrase "in one embodiment" or "in some embodiments" in various places within specified descriptions in this specification are not necessarily all referring to the same embodiment, however, can mean one or more but not all embodiments. The terms "including," "comprising," "featuring," and variations thereof are meant to encompass the items listed thereafter and equivalents thereof as well as additional items.

[0035] Currently, the recommendation of garment pattern design mainly relies on traditional experience or manual adjustment. Designers usually match the garment patterns of different user groups based on the accumulated experience of previous designs, or through questionnaire investigation, size statistics, single factor analysis, etc. These methods can generate garment patterns suitable for most users to a certain extent, but have obvious limitations. Traditional methods usually ignore the dynamic influence of individual differences of users on the adaptability of garment patterns. The body shape data, dressing style preference and perception of pattern details of different users will lead to the difficulty of meeting individual needs by single experience or static recommendation, thereby affecting the accuracy of garment design and user satisfaction.

[0036] To solve the above problems, the embodiment of the application provides a garment design parameter recommendation method, device, equipment and computer program product. In the method, the user body shape data, dressing style preference and candidate garment pattern features of a target user are obtained, and an adaptation difference matrix and a style correlation matrix are established to start the whole garment pattern recommendation process. The initial anomaly detection parameters are determined based on the adaptation difference matrix and the style correlation matrix, which can reflect the adaptation deviation of different pattern parameters under the individual needs of users. Then, the initial adaptation scores are screened and removed in combination with the anomaly detection parameters, and the accuracy of pattern matching is further optimized. The final garment pattern parameter recommendation scheme is generated according to the retained adaptation scores, realizing the personalized garment design recommendation for the target user. Not only the recommendation of single pattern parameter is realized, but also the whole pattern system can be optimized in combination with the correlation analysis of body shape data, style preference and pattern parameters, so as to improve the accuracy and individualization level of garment pattern recommendation, reduce the blindness of the recommendation process and the dependence on the experience of designers, and comprehensively and systematically improve the efficiency and user satisfaction of garment design parameter recommendation.

[0037] The garment design parameter recommendation method provided by the embodiment of the application can be applied to an electronic device, and the electronic device is the execution subject of the garment design parameter recommendation method provided by the embodiment of the application. The specific type of the electronic device is not limited in the embodiment of the application.

[0038] For example, the electronic device can be an ultra-mobile personal computer (UMPC), a netbook, a desktop computer, a computer, a laptop computer, a communication device, a computing device, a satellite wireless device, etc.

[0039] In order to better understand the garment design parameter recommendation method provided by the embodiments of the present application, the specific implementation process of the garment design parameter recommendation method provided by the embodiments of the present application is exemplarily introduced below.

[0040] Figure 1 The schematic flowchart of the garment design parameter recommendation method provided by the embodiments of the present application is shown, and the garment design parameter recommendation method comprises:

[0041] S100, determining initial fitting scores between user body shape data of a target user and a plurality of candidate garment patterns, and establishing a fitting difference matrix of the plurality of initial fitting scores; wherein the initial fitting scores correspond to the candidate garment patterns one by one.

[0042] It can be understood that the user body shape data refers to structural data capable of representing the body features of the target user, which is obtained by measurement or scanning, such as shoulder width, bust, waist, hip, height, weight, and limb proportion, etc. Please refer to Figure 2 The user body shape data can be obtained from manual measurement, three-dimensional human body scanner, or numerical values input by the user on the mobile terminal. The candidate garment pattern refers to a plurality of standardized pattern schemes stored in a preset garment database, each pattern contains a series of structural parameters to define the size and proportion of the garment at different parts. The initial fitting score refers to a value obtained by matching and calculating the user body shape data and a certain candidate garment pattern in the parameter dimension, which is used to measure the matching degree of the pattern and the user body shape. Each candidate pattern will correspond to a fitting score, and finally a plurality of scores will be formed.

[0043] The user body shape data can be compared item by item (such as bust difference, shoulder width difference, waist-hip ratio difference, etc.) to calculate the fitting score of each candidate garment pattern. In order to quantify the difference between these scores, a fitting difference matrix can be further constructed, and the elements in the matrix represent the difference between the fitting scores of any two candidate patterns. The role of this matrix is to reveal the fitting difference relationship between different patterns, which is convenient for subsequent steps of anomaly detection and optimization.

[0044] The reason for such design is that the body shape features of the user are complex and diverse, and a single score is difficult to reflect the overall law, and by constructing the difference matrix, the relative difference between the candidate patterns can be intuitively presented, thereby providing a data basis for subsequent anomaly elimination and recommendation optimization. This method avoids the limitation of simply relying on the highest score, making the whole recommendation process more robust.

[0045] In a possible implementation, in step S100, an initial fitting score between the user body shape data of the target user and a plurality of candidate garment patterns is determined, including:

[0046] S110, obtaining an original fitting score between the user body shape data and a plurality of basic garment patterns.

[0047] It can be understood that the original fitting score refers to an initial calculation result obtained by directly comparing the user body shape data with the preset basic garment patterns in the database. The basic garment pattern is a standard structure pattern without style or individualization modification, and is generally obtained by industry standard or large-scale population data statistics, for example, the reference model of typical S, M, L, XL and the like. The calculation method of the original fitting score is generally based on parameter difference or proportional matching, for example, by the formula Score = 1-|(user parameter-pattern parameter) / pattern parameter|, wherein Score is the original fitting score, and each dimension is standardized and scored, and then the overall original fitting score is obtained by weighted summation.

[0048] The user body shape data can be compared with the parameters of the basic garment pattern one by one, and the matching degree result is generated by an algorithm. For example, if the user's bust is 92 cm and the basic pattern bust is 90 cm, the difference of this dimension is small, and the corresponding score is high; on the contrary, if the difference is large, the score is reduced.

[0049] The reason for this is that the original fitting score, as an initial evaluation index, can quickly reflect the matching of the user body shape and each pattern. Although the original score may be affected by local anomalies or extreme values, it provides basic data for subsequent evaluation value calculation and anomaly value filtering.

[0050] S120, calculating a plurality of pattern evaluation values based on the original fitting scores; wherein each original fitting score corresponds to a pattern evaluation value.

[0051] It can be understood that the pattern evaluation value is the result of further processing of the original fitting score, and can be combined with statistical methods or normalization methods to make the comparison between different patterns more objective. Since the original fitting score may have problems such as uneven data distribution and unbalanced parameter weight, direct use may cause errors. Therefore, by converting each original fitting score into an evaluation value, the data deviation can be corrected to a certain extent, and the stability of the fitting calculation is improved.

[0052] All original fit scores can be mapped to a unified interval through standardization (e.g., z-score standardization or min-max normalization) to avoid the impact of different user body type data distribution differences on the results. A weight factor can be assigned to each original fit score, which can be based on user needs through a set front-end page or based on a default weight (e.g., more emphasis on shoulder width matching or waist circumference matching) to generate a weighted evaluation value. Finally, each version will correspond to an evaluation value to reflect its overall fit degree.

[0053] The reason for this is that the matching between user body type and garment version cannot rely solely on the original score, but needs to be adjusted through statistics and weights to form a more comparable index. By calculating the evaluation value, the robustness of the scoring system can be enhanced, and the interference of abnormal scores on the results can be reduced.

[0054] S130, according to the median and dispersion of the plurality of version evaluation values, filtering the abnormal values from the original fit scores to obtain the effective initial fit scores.

[0055] It can be understood that the median refers to the middle value of all version evaluation values arranged in order of size, used to measure the central tendency of the overall data. The dispersion refers to the dispersion of the data distribution, which can be measured by standard deviation, variance or interquartile range. The process of filtering abnormal values is to compare the deviation of the original fit score from the overall distribution, eliminate abnormal points that differ greatly from most data, and thus obtain more reliable effective initial fit scores.

[0056] The median of all version evaluation values can be calculated to determine the central level of normal scores. Then, based on the dispersion, the deviation between each score and the median is calculated. If the deviation of a certain score exceeds a preset threshold (such as more than twice the standard deviation), the score will be marked as abnormal and removed from the original score set. After this step, the remaining scores are the effective initial fit scores.

[0057] The reason for this design is that there are often abnormal scores in actual data due to measurement errors, extreme body types or algorithm inaccuracies. If not handled, these abnormal values will affect the accuracy of the overall recommendation. By filtering based on the median and dispersion, the influence of abnormal points can be effectively eliminated, making the final recommendation result more robust.

[0058] Illustratively, S130, according to the median and dispersion of the plurality of version evaluation values, filtering the abnormal values from the original fit scores to obtain the effective initial fit scores, comprising:

[0059] S131, for each version evaluation value, calculating an evaluation difference between the version evaluation value and the median.

[0060] It can be understood that the evaluation difference refers to the numerical difference between the individual version evaluation value and the median of the overall evaluation value distribution. The purpose of calculating the difference is to measure the deviation of each version score relative to the overall central tendency. The median is a robust statistical quantity that is less susceptible to extreme values than the mean, so it is often chosen as the baseline in anomaly detection. By calculating the evaluation difference, scores that deviate significantly from the overall trend can be identified.

[0061] All evaluation values can be sorted first to obtain the median value. Then, for each version evaluation value, the difference between it and the median is calculated, for example, Δi = |E i -Median(E)|, where E i represents the ith evaluation value, and Median(E) is the median. The larger the difference, the more obvious the deviation of the evaluation value from the overall trend, and the more likely it is to become an outlier. After calculation, a set of difference vectors is obtained for subsequent dispersion comparison and anomaly identification.

[0062] The reason for designing this step is that there may be individual extreme matching cases in user body data and version fitting calculation, such as extremely high or low scores caused by a particular measurement dimension. If these data are not analyzed, it may mislead the final recommendation results. Through median-based difference calculation, a simple and effective measurement standard can be constructed for subsequent outlier removal.

[0063] S132, according to the evaluation difference and the dispersion, mark the obviously abnormal scores from all original fitting scores, and remove the obviously abnormal scores to obtain the effective initial fitting scores.

[0064] It can be understood that the evaluation difference and the overall dispersion can be used to jointly determine which original fitting scores are obviously abnormal. The dispersion refers to the divergence of the overall data distribution, which can be measured by standard deviation or interquartile range (IQR). When the difference of a score is too large relative to the overall dispersion, the score is marked as an abnormal score.

[0065] A threshold standard can be determined first, for example, when the difference between a score and the median is greater than 2 or 3 times the standard deviation, the score is determined to be abnormal. The marked scores will be removed and will not be involved in subsequent calculation and recommendation. Finally, the score set after cleaning is the effective initial fitting score.

[0066] The reason for this is that outliers often seriously affect the stability and fairness of the fitting results. For example, a certain version may have a score much higher than other versions due to measurement errors, which may result in inappropriate clothing for the user if used directly for recommendations. By using statistical-based outlier detection methods, these extreme data can be effectively excluded, ensuring the scientificity and reliability of the recommendation results.

[0067] Illustratively, S132, according to the evaluation difference and the dispersion degree, mark the obviously abnormal scores from all original fitting scores, and remove the obviously abnormal scores to obtain effective initial fitting scores, including:

[0068] S1321, according to the evaluation difference and the dispersion degree, calculate the proportion degree, determine the dispersion degree ratio between the dispersion degree and the evaluation difference of each original fitting score.

[0069] It can be understood that the proportion degree calculation refers to the ratio operation between the evaluation difference of each score and the overall dispersion degree, so as to quantify the degree of deviation of a single score from the overall distribution. The ratio can be defined as R i =Δi / D, where R i The dispersion degree ratio, Δi represents the difference between a certain score and the median, and D represents the overall dispersion (such as standard deviation or interquartile range). The larger the ratio obtained by calculation, the more serious the deviation of the score from the overall trend.

[0070] All original fitting scores can be calculated one by one to generate a ratio vector {R1, R2, …, R n}. These ratios will be used as the core indicators for subsequent anomaly identification. In this way, the scale problem that may be caused by directly using the absolute difference value can be avoided, because the score interval of different users may have large differences, and the use of ratio calculation can achieve standardization, making the anomaly identification more robust.

[0071] By constructing the dispersion degree ratio, the abnormal strength of each score relative to the overall distribution can be quantified, thereby providing a basis for subsequent threshold judgment. By borrowing the principles of z-score and IQR anomaly detection in statistics, the problem of anomaly identification under different data scales can be effectively addressed.

[0072] S1322, compare each dispersion degree ratio with a preset threshold, and when the dispersion degree ratio is greater than the preset threshold, mark the original fitting score corresponding to the dispersion degree ratio as an obviously abnormal score.

[0073] It can be understood that each score can be marked as abnormal through a threshold determination method. The so-called preset threshold refers to the discrimination boundary set according to experience or statistical laws during design. For example, the threshold can be set to 2 or 3, similar to the range of the 95% or 99% confidence interval in the standard normal distribution. When the ratio of the dispersion degree of a certain score exceeds this threshold, it indicates that its gap from the overall trend is too large and should be determined as abnormal.

[0074] Each value in the ratio vector can be checked one by one, and the scores corresponding to the elements that exceed the threshold are marked as significantly abnormal scores. These abnormal scores will be temporarily stored in the abnormal list and wait for subsequent elimination processing.

[0075] The reason for such a design is that the threshold method is an intuitive and efficient abnormal detection means, which can quickly screen out abnormal data without complex modeling. By comparing the ratio of the dispersion degree with the threshold, the reliability and stability of detection can be improved while ensuring the simplicity of calculation.

[0076] S1323, clear all significantly abnormal scores to obtain effective initial adaptation scores.

[0077] It can be understood that all the abnormal scores marked in the previous step can be eliminated from the original score set to form a new score set. This set is the effective initial adaptation score. The elimination process is not just simply deleting data, but may also include recording and archiving the abnormal scores for use in subsequent analysis or model optimization.

[0078] An effective score set {S1, S2,..., S m} can be generated, where m < n, indicating that the number of original scores decreases and the quality of the remaining data is higher. This effective score set will be used as the core input for subsequent style matrix construction and abnormal detection parameter calculation.

[0079] The reason for doing this is that once abnormal scores enter subsequent calculations, they may cause serious interference to the structure of the style matrix and error evaluation. By clearing the outliers at this stage, it can ensure that subsequent analysis and recommendations are based on purer and more reliable data. This link is similar to outlier cleaning in data preprocessing and is a key step to ensure the accuracy and robustness of the overall algorithm.

[0080] S200, obtain the dressing style preference of the target user, and based on the dressing style preference of the target user and the style characteristics of multiple candidate clothing patterns, construct a style association matrix between the initial adaptation scores.

[0081] It can be understood that the style correlation matrix refers to a matrix expression form for describing the relationship between the user dressing style preference and the style features of multiple candidate garment patterns. The user dressing style preference here is the tendency of the target user to the garment style in daily dressing, such as preference for loose, body-fitting, casual, formal, sporty or trendy style information. Please refer to Figure 3 These dressing style preferences can be obtained through the user's historical purchase records, daily detection results of smart wear devices, online interactive questionnaires or garment tag identification on social platforms. The pattern style features are the style labeling processing for different candidate garment patterns, for example, the same pattern may have both body-fitting + formal features, or loose + casual combination.

[0082] The style preference parameters of the target user can be extracted and converted into quantifiable numerical vectors; for example, casual is assigned a value of 1, formal is assigned a value of 0.2, and so on. The style features of each candidate garment pattern are vectorized. Finally, the style correlation matrix can be generated by calculating the similarity (such as cosine similarity or weighted Euclidean distance) between the user preference vector and the pattern style vector. Each element in the matrix represents the degree of association between a pattern and the user's style preference, and the larger the numerical value, the more consistent it is with the user's style habits.

[0083] The reason for this is that even if two patterns are very close in body fit, if their style is very different from the user's actual preference, it is difficult for the user to accept. Therefore, by incorporating user preferences into the scoring system and expressing them uniformly through the style correlation matrix, a connection between body fit and style fit can be established, thus achieving a more comprehensive pattern recommendation. This matrix not only provides a reference for style anomaly detection, but also ensures that the final recommendation result meets both the body fit and the user's personalized dressing style.

[0084] S300, based on the adaptation difference matrix, the style correlation matrix and the plurality of initial adaptation scores, calculating an anomaly detection parameter of the plurality of initial adaptation scores; wherein the anomaly detection parameter is used to judge whether there is an abnormal score in the plurality of initial adaptation scores.

[0085] It can be understood that the anomaly detection parameter refers to an index for comprehensively judging whether there is an abnormal score in the plurality of initial adaptation scores, which is calculated based on the statistical characteristics of the adaptation difference matrix, the style correlation matrix and the plurality of initial adaptation scores. The adaptation difference matrix refers to a matrix structure used to describe the difference between different candidate patterns in the adaptation score, for example, when the adaptation scores of two patterns differ greatly, the matrix element value will be higher. The style correlation matrix provides a reference in the style dimension, so that the anomaly detection not only depends on the numerical difference, but also considers the style consistency. The plurality of initial adaptation scores is the effective score set after the previous anomaly removal.

[0086] The initial adaptation score can be combined with the adaptation difference matrix and the style association matrix to construct a multi-dimensional anomaly detection parameter. For example, if a certain score is greatly different in numerical value from the same type of version, but is highly related to the user in style, it can be determined as an anomaly; for another example, a score is close in numerical value to most versions, but has a very low degree of style association, which can also be determined as an anomaly. The finally generated anomaly detection parameter is usually a numerical value or a probability, indicating the possibility of a score being considered as an abnormal value.

[0087] The reason for this is that a single adaptation score or difference matrix cannot fully reveal the abnormal points, especially in the presence of user style preferences, some surface outliers may be the real needs of the user. By constructing an anomaly detection parameter, a balance between numerical deviation and style consistency can be achieved, avoiding misjudgment and omission, so that the process of removing anomalies is more scientific and reasonable. This step provides a quantitative basis for subsequent removal of unreasonable scores, ensuring the accuracy and personalization of the final recommended scheme

[0088] In one possible implementation, S300, based on the adaptation difference matrix, the style association matrix and the plurality of initial adaptation scores, calculates an anomaly detection parameter of the plurality of initial adaptation scores, comprising:

[0089] S310, based on the style association matrix, the adaptation difference matrix and the plurality of initial adaptation scores, obtains a score deviation matrix of the plurality of initial adaptation scores.

[0090] It can be understood that the score deviation matrix refers to a matrix structure used to describe the relative deviation between scores, which is constructed based on the style association matrix, the adaptation difference matrix and the plurality of initial adaptation scores. Here, the deviation is not a single numerical difference, but a modified expression of the difference after considering the user preferences and version characteristics. For example, if two versions have a numerical difference in adaptation scores, but they have high similarity in style characteristics, the deviation value in the score deviation matrix will be weakened; on the contrary, if the style difference is large, even if the numerical difference is small, it may be amplified.

[0091] Each initial adaptation score can be compared with other scores one by one to calculate the numerical difference, and then normalized in combination with the results of the adaptation difference matrix to avoid high deviation caused by too large or too small absolute value of the score. Subsequently, the normalized difference is combined with the style association matrix to calculate the final deviation value. The result of this processing is that each element of the score deviation matrix reflects the dual information of numerical difference and style difference, thereby describing the relationship between scores in multiple dimensions.

[0092] The reason for generating the score deviation matrix is that a single initial adaptation score cannot reflect the global relationship. Only through matrix deviation analysis can it be revealed whether some scores are inconsistent with the overall trend. Especially in the context of multiple version comparison, abnormal scores often deviate from the consistency of most scores, and the score deviation matrix can provide mathematical basis for such outliers. Finally, the score deviation matrix will be used as the core input for the subsequent calculation of comprehensive error value and abnormal detection parameter, so as to make the abnormal detection more global and robust.

[0093] S320, based on the score deviation matrix and the adaptation difference matrix, calculating the comprehensive error value of the initial adaptation score; wherein the comprehensive error value is used to reflect the overall error situation of all initial adaptation scores.

[0094] It can be understood that the comprehensive error value refers to a global index calculated based on the score deviation matrix and the adaptation difference matrix, which is used to reflect the deviation and inconsistency of all initial adaptation scores in the global range. Unlike a single score difference value, the comprehensive error value not only considers the numerical difference between each score and other scores, but also integrates the inherent difference between versions expressed by the adaptation difference matrix. For example, when two versions themselves have great differences in design structure, their score difference may be reasonable and should not be directly considered as abnormal, and the comprehensive error value calculation will correct such cases.

[0095] The score deviation matrix can be scanned as a whole, the deviation degree between each score and the remaining scores is counted, and the deviations are weighted by the adaptation difference matrix. All weighted results are normalized or aggregated to obtain a value reflecting the overall error situation. The higher the comprehensive error value, the more inconsistent the entire score system is; the lower the comprehensive error value, the more coordinated each score is, and the higher the overall reliability is.

[0096] The reason for this is that the detection of abnormal scores not only needs to focus on the comparison of individual scores with standard values, but also must combine the stability of the overall score system for judgment. If the overall error level is too high, there is obvious abnormality in the data, which must be screened out through further analysis. The introduction of the comprehensive error value can provide a global reference to avoid one-sided judgment based on local differences. This step provides a solid statistical basis for the calculation of abnormal detection parameters in the next stage.

[0097] S330, according to the comprehensive error value, the score deviation matrix and the adaptation difference matrix, calculating the abnormal detection parameter of the initial adaptation score.

[0098] It can be understood that the anomaly detection parameter is a judgment index obtained by comprehensively using the comprehensive error value, the score deviation matrix and the adaptation difference matrix, and the purpose is to determine which initial adaptation scores may belong to abnormal values. The anomaly detection parameter here is usually expressed in numerical or probability form, indicating the possibility of a score becoming abnormal. For example, when the deviation value of a certain score is continuously high compared to most scores, and it is still in an unreasonable interval after correction of adaptation difference and style association, the corresponding anomaly detection parameter will be significantly higher than the threshold.

[0099] The elements in the score deviation matrix can be analyzed one by one to extract the average deviation value related to each score, and combined with the comprehensive error value for proportional calculation. Further correction is made using the adaptation difference matrix to ensure that no misjudgment occurs when the pattern itself difference is reasonable. Finally, each score will get an anomaly detection parameter, and the larger the parameter, the more uncoordinated the score is with the overall system, and the more likely it is to be considered abnormal.

[0100] The reason for such design is that abnormal scores are often not single-point absolute errors, but are inconsistent with overall trends, user styles and pattern differences. A single-dimensional judgment is easy to misjudge, while an anomaly detection parameter formed by multi-dimensional fusion can more accurately distinguish between normal deviation and true anomaly. The anomaly detection parameter will directly guide the subsequent anomaly removal operation to ensure that the final recommended garment pattern is both scientific and reasonable and meets user needs.

[0101] S400, when the anomaly detection parameter indicates that there are abnormal scores in the plurality of initial adaptation scores, removing the abnormal scores.

[0102] It can be understood that abnormal score removal refers to removing scores that are unreasonable or deviate from the overall rule from the final score set when the anomaly detection parameter indicates that there are such scores in the plurality of initial adaptation scores. The abnormal score here is not simply an extreme score, but a score that still shows unreasonable after multi-dimensional cross verification of the aforementioned comprehensive error value, score deviation matrix and style association matrix. For example, although some scores may seem reasonable in a single dimension, they still show high inconsistency with other scores in the overall framework of style and adaptation difference, and such scores should be judged as abnormal.

[0103] The anomaly detection parameter corresponding to each score can be threshold judged, and when the detection parameter exceeds the preset judgment threshold, the score will be marked as an abnormal candidate. All candidate abnormal scores are confirmed again, combined with the user's preference weight and pattern characteristics, and finally the truly abnormal scores that need to be removed are screened out. The result after removal will form a more pure and consistent score set, laying a foundation for subsequent generation of recommended solutions.

[0104] The reason for this design is that any automated scoring process can be affected by data noise, acquisition errors or user subjective differences, resulting in outliers deviating from the overall trend. If these abnormal scores are not cleaned up, it may directly affect the accuracy of the final recommendation, or even cause the recommended result to completely deviate from the user's demand. Therefore, the role of S400 is to improve the overall stability and credibility of the score set through a scientific outlier removal mechanism, and to ensure that the final recommended result is more in line with the user's real needs.

[0105] In one possible implementation, S400, when the abnormal detection parameter indicates that there are abnormal scores in the plurality of initial fitting scores, the abnormal scores are removed, comprising:

[0106] S410, when the abnormal detection parameter indicates that there are abnormal scores in the plurality of initial fitting scores, the fitting index value corresponding to each initial fitting score is extracted.

[0107] It can be understood that the fitting index value refers to the sub-evaluation dimension parameter corresponding to each initial fitting score. For example, the fitting score of a garment style is not only an overall score, but also may include multiple dimensions such as shoulder width fit, chest circumference fit, waist tightness, pant length or skirt length ratio, etc. These indicators jointly constitute the basis for calculating the initial fitting score and are an important reference for judging its rationality.

[0108] When it is determined that there are abnormal scores, the fitting index value corresponding to each score is extracted first to track the source of the abnormality. For example, a score is determined to be abnormal, it may be found that the difference in shoulder width fit is too large, while other dimensions are normal, which indicates that the abnormality mainly comes from a single dimension, rather than the overall evaluation. In this way, not only can the abnormal score be removed, but also the problem can be located.

[0109] The reason for this is that the initial fitting score is a highly abstract comprehensive value, and direct removal may result in information loss. By extracting the fitting index value, the explanation of the source of the abnormality can be retained while removing the abnormality, making it more transparent and credible. This explanation ability is particularly important for clothing recommendation, because users not only need the recommended result, but also want to know why it is recommended. Therefore, S410 plays a role in linking the previous and the next in the abnormality removal link, making the removal process more scientific and reasonable.

[0110] S420, according to the style association matrix and the fitting difference matrix, a weight distribution matrix of the plurality of initial fitting scores is calculated.

[0111] It can be understood that the weight distribution matrix refers to a weighted matrix generated based on the style correlation matrix and the fitting difference matrix, which is used to allocate the weight proportion of different initial fitting scores in anomaly detection. In other words, this matrix not only considers the difference between the styles in the fitting scores, but also integrates the importance of the user style preference for each score, thereby forming a more user demand-oriented weighting system.

[0112] Exemplarily, please refer to Figure 4 The generation of the weight distribution matrix (W) is based on the style correlation matrix (A) and the fitting difference matrix (B), and the core is to quantitatively integrate the information of the two to allocate weights for each index of the initial fitting score. The specific process is as follows: first, clarify the meaning of the matrix - the element A_jj in A represents the style correlation degree of style i and j (the larger the value, the more similar), and the element B_ij in B represents the fitting score difference degree of style i and j (the smaller the value, the more consistent); then, standardize to convert A into A_ij=A_ij / max(A) to preserve the relative correlation degree, and invert standardize B into B_ij=1-(B_ij / max(B)) to reflect consistency (the larger the value, the more reliable); then calculate the comprehensive influence factor F_ik=α×Ajk+β×Bik (α, β are weight coefficients set by the business, which respectively reflect the importance of style correlation and fitting consistency); then convert F_ik into initial weight W"_ik=F_ik / ΣF_ik (the sum of the weights of the indexes in the same style is 1); finally, through cross-style verification (such as limiting abnormally high weight), fine-tune to obtain the final weight distribution matrix W, where W_ik represents the weight of the kth index of style i in the initial fitting score, which reflects both the style rationality and the fitting score reliability.

[0113] The reason for such design is that whether the score is abnormal cannot simply depend on the numerical value, but must be weighed in combination with the user's personalized needs. Through the weight distribution matrix, it can be more reasonably determined which scores are more worthy of retention and which scores are not significant in the overall recommendation. Ultimately, this weight-based screening method makes the recommendation result more personalized and intelligent, while avoiding the misremoval of styles that truly conform to the user's style due to extreme numerical values.

[0114] S430, based on the weight distribution matrix and the fitting index value corresponding to each initial fitting score, screening the abnormal scores from the plurality of initial fitting scores.

[0115] It can be understood that screening the abnormal scores refers to comparing and weighting analyzing all scores one by one based on the weight distribution matrix and the fitting index value of each initial fitting score, and finally confirming which scores are abnormal and are removed, please refer to Figure 4The key here is to combine the index value with the weight for double judgment, rather than simply relying on the abnormal detection parameter of a certain dimension.

[0116] The weight distribution matrix can be used to weight and adjust each fitting index value, so that it is more consistent with the user's real needs in the final calculation. For example, if a score deviates greatly in the bust circumference index, but the user's style preference does not emphasize the fit of the upper body, the abnormality may be weakened after weighting, thereby avoiding false rejection. After overall analysis of the weighted results, those scores that do not conform to the overall trend in multiple indexes are screened out and determined as abnormal scores, and finally removed from the set.

[0117] The reason for this is that simply relying on numerical differences or style similarity may lead to inaccurate abnormal screening. By combining index values and weight matrices, the essence of abnormal scores can be more accurately grasped, ensuring that unreasonable noise is removed rather than potentially reasonable individualized needs.

[0118] Optionally, S430, based on the weight distribution matrix and the fitting index value corresponding to each initial fitting score, screening abnormal scores from the plurality of initial fitting scores, comprising:

[0119] S431, when the value in the diagonal position of the weight distribution matrix is greater than any other value in the column, selecting an abnormal index value corresponding to the value from all fitting index values.

[0120] It can be understood that the value in the diagonal position refers to the weight value corresponding to each score and its own dimension in the weight distribution matrix. This value usually represents the degree of self-matching or self-importance of the score in the overall scoring system. Compared with the values in other columns of the matrix, if the value in the diagonal position is significantly higher, it means that the score is highly related to the fitting index corresponding to itself, but lacks consistency with other scores, which often reflects that the score is abnormal or deviates excessively.

[0121] The weight distribution matrix can be traversed column by column, and the weight value in the diagonal position is compared with the weight values in the remaining positions of the column. When it is found that the diagonal weight value is greater than any other value in the column, the fitting index value corresponding to the value is extracted as a potential abnormal index value candidate. For example, a fitting score of a garment pattern is highly consistent with its own score in the shoulder width dimension, but shows great inconsistency when compared with other patterns, which indicates that the shoulder width index of the score may deviate from the overall rule and should be paid attention to.

[0122] The reason for this design is that anomalies often manifest as self-reinforcing but lack group consensus. That is, a certain score may seem reasonable on an individual dimension, but lacks correlation with other scores, leading to deviation from the overall trend. By comparing the diagonal values with other values, this deviation can be intuitively revealed, and the problem indicators can be locked in time, avoiding the dependence on the single nature of the global error parameter, making the anomaly detection more hierarchical and targeted, thus ensuring a more accurate and robust screening process.

[0123] S432, determining the initial adaptive score corresponding to the abnormal indicator value as the abnormal score.

[0124] It can be understood that the initial adaptive score corresponding to the abnormal indicator value refers to the overall score associated with the potential abnormal indicator value determined by S431. When a certain indicator value is abnormal in the weight distribution matrix, not only the abnormal dimension needs to be recorded, but also the corresponding overall initial adaptive score needs to be determined as the abnormal score to ensure the integrity of the abnormal elimination.

[0125] The marked abnormal indicator value can be mapped to the original adaptive score. Since each adaptive score is the result of multiple indicator values, the abnormality of a single indicator often affects the rationality of the overall score. Therefore, when one or more indicator values are confirmed to be abnormal, the corresponding initial adaptive score will be marked as abnormal as a whole and eliminated in the final set. For example, when the waistline indicator of a version is determined to be abnormal, and the waistline is a key component of the overall score, the overall score of the version is directly determined as an abnormal score.

[0126] The reason for this is that the core of the scoring system is overall consistency. If only a single indicator is eliminated, while its corresponding overall score is retained, it may lead to the use of unreasonable data in the subsequent recommendation process, thereby affecting the accuracy of the final result. By binding the abnormal indicator value with the overall score, the completeness and scientificity of the elimination mechanism can be ensured. At the same time, this approach also improves stability, avoiding secondary deviation caused by partial anomalies.

[0127] S500, generating a final garment version parameter recommendation scheme for the target user according to all the retained initial adaptive scores.

[0128] It can be understood that the final garment pattern parameter recommendation scheme refers to a comprehensive recommendation result generated based on all valid initial fitting scores retained after abnormality elimination and multi-dimensional analysis. The parameter recommendation scheme here does not only select a single pattern, but includes a set of garment pattern parameters matched with the user's body shape characteristics, style preferences and comfort requirements. These parameters can include structural dimensions such as shoulder width, bust, waist, sleeve length and garment length, and can also be extended to pattern modification styles such as waistline, ease allowance or cutting angle, thereby forming a comprehensive recommendation scheme.

[0129] All retained initial fitting scores can be weighted and fused to ensure that the recommendation result is consistent with the user's body shape characteristics and the user's style preferences. Personalize the parameters of each garment pattern, for example, for users with a tall body, increase the length of the hem appropriately while maintaining the overall proportions; for users who prefer a slim style, make more compact recommendations on parameters such as waist and shoulder width. Finally, output a recommended garment scheme containing a combination of multi-dimensional parameters as a reference data for virtual fitting and online shopping.

[0130] The reason why the final recommendation scheme is generated by retaining the scores, rather than simply selecting the pattern with the highest score, is that the user's needs are diverse and complex. A single score may perform well in one dimension, but may be insufficient in other dimensions, making it difficult to meet overall needs. The recommendation scheme generated through comprehensive analysis can balance between overall adaptability and personalized preferences, making the result more consistent with the actual wearing effect. At the same time, the parameterized output of the recommendation scheme can seamlessly connect with the garment customization module of the intelligent clothing system or e-commerce platform, realizing the application value of intelligence and automation.

[0131] Corresponding to the garment design parameter recommendation method of the above embodiment, the embodiment of the present application also provides a garment design parameter recommendation device. Each unit of the device can implement each step of the garment design parameter recommendation method. Figure 5 The structure block diagram of the garment design parameter recommendation device provided by the embodiment of the present application is shown, and only the part related to the embodiment of the present application is shown for ease of illustration.

[0132] Referring to Figure 5 The garment design parameter recommendation device includes:

[0133] An initial unit is configured to determine initial fitting scores between user body shape data of a target user and a plurality of candidate garment patterns, and to establish a fitting difference matrix of a plurality of the initial fitting scores; wherein the initial fitting scores correspond one-to-one to the candidate garment patterns;

[0134] A construction unit is used to obtain the clothing style preferences of the target user, and construct a style correlation matrix between the initial adaptation scores based on the clothing style preferences of the target user and the style features of multiple candidate clothing patterns;

[0135] The detection unit is used to calculate anomaly detection parameters for the multiple initial adaptation scores based on the adaptation difference matrix, the style correlation matrix, and the multiple initial adaptation scores; wherein, the anomaly detection parameters are used to determine whether there are any anomalous scores among the multiple initial adaptation scores.

[0136] A filtering unit is used to remove the abnormal score when the anomaly detection parameters indicate that there is an abnormal score among the multiple initial adaptation scores.

[0137] The generation unit is used to generate a final clothing pattern parameter recommendation scheme for the target user based on all the retained initial adaptation scores.

[0138] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0139] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit module can exist physically separately, or two or more unit modules can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0140] This application also provides an electronic device. Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 6 As shown, the electronic device 6 of this embodiment includes: at least one processor 60 ( Figure 6 Only one is shown in the image), at least one memory 61 ( Figure 6The computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60, when the processor 60 executes the computer program 62, causes the electronic device 6 to implement the steps in any of the above-mentioned various garment design parameter recommendation method embodiments, or causes the electronic device 6 to implement the functions of the units in the above-mentioned various device embodiments.

[0141] For example, the computer program 62 can be divided into one or more units, which are stored in the memory 61 and executed by the processor 60 to complete the present application. The one or more units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 62 in the garment design parameter recommendation 6.

[0142] The electronic device 6 can be a desktop computer, a notebook, a palm computer, a cloud server, and other computing devices or terminal devices. The electronic device can include, but is not limited to, the processor 60, the memory 61. Those skilled in the art can understand that the electronic device 6 can include more or less components, or combine certain components, or different components, for example, it can also include an input / output device, a network access device, a bus, etc. Figure 6 The electronic device 6 is only an example and does not constitute a limitation on the electronic device 6, and can include more or less components than those shown, or combine certain components, or different components, for example, it can also include an input / output device, a network access device, a bus, etc.

[0143] The processor 60 can be a central processing unit (CPU), and the processor 60 can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0144] The memory 61 can be an internal storage unit of the electronic device 6, such as a hard disk or a memory of the electronic device 6 in some embodiments. The memory 61 can also be an external storage device of the electronic device 6, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device 6 in other embodiments. Further, the memory 61 can include both an internal storage unit and an external storage device of the electronic device 6. The memory 61 is used to store an operating system, an application program, a BootLoader, data, and other programs, such as program codes of the computer program, etc. The memory 61 can also be used to temporarily store data that has been output or is to be output.

[0145] The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps in any of the above method embodiments.

[0146] The computer program product, when running on an electronic device, causes the electronic device to implement the steps in any of the above method embodiments.

[0147] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application implements all or part of the processes in the above embodiments, which can be completed by instructing related hardware through a computer program. The computer program can be stored in a computer readable storage medium, and the computer program, when executed by a processor, can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer readable medium at least includes any entity or device capable of carrying the computer program code to an electronic device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. For example, a U disk, a mobile hard disk, a magnetic disk or an optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium cannot be an electrical carrier signal and a telecommunications signal.

[0148] In the above embodiments, the description of each embodiment is focused on, and the part not described or recorded in a certain embodiment can be referred to the relevant description of other embodiments.

[0149] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0150] In the embodiments provided in the present application, it should be understood that the disclosed garment design parameter recommendation apparatus / electronic device and method can be implemented in other ways. For example, the garment design parameter recommendation apparatus / electronic device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0151] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments according to actual needs.

[0152] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent ones; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for recommending clothing design parameters, characterized in that, include: Determine the initial fit score between the target user's body shape data and multiple candidate clothing patterns, and establish multiple fit difference matrices for the initial fit scores; wherein, the initial fit score corresponds one-to-one with the candidate clothing pattern; the elements in the fit difference matrix represent the difference in fit scores between any two candidate clothing patterns. The clothing style preferences of the target user are obtained, and a style correlation matrix is ​​constructed based on the clothing style preferences of the target user and the style features of multiple candidate clothing styles; wherein, each element in the style correlation matrix represents the degree of correlation between a certain style and the user's style preference; Based on the adaptation difference matrix, the style correlation matrix, and multiple initial adaptation scores, anomaly detection parameters are calculated for the multiple initial adaptation scores; wherein, the anomaly detection parameters are used to determine whether there are any anomalous scores among the multiple initial adaptation scores. When the anomaly detection parameters indicate that there is an abnormal score among the multiple initial adaptation scores, the abnormal score is removed. Based on all the retained initial fit scores, a final clothing pattern parameter recommendation scheme is generated for the target user.

2. The method for recommending clothing design parameters as described in claim 1, characterized in that, Determine the initial fit score between the target user's body shape data and multiple candidate clothing patterns, including: Obtain the original fit score between user body shape data and multiple basic clothing patterns; Based on the original adaptation score, multiple pattern evaluation values ​​are calculated; wherein, each of the original adaptation scores corresponds to one pattern evaluation value; Based on the median and dispersion of multiple pattern evaluation values, outliers are filtered from the original adaptation score to obtain an effective initial adaptation score.

3. The method for recommending clothing design parameters as described in claim 2, characterized in that, The step of filtering outliers from the original adaptation score based on the median and dispersion of multiple version evaluation values ​​to obtain an effective initial adaptation score includes: For each pattern evaluation value, calculate the evaluation difference between the pattern evaluation value and the median; Based on the evaluation difference and the degree of dispersion, obviously abnormal scores are marked from all the original adaptation scores, and the obviously abnormal scores are removed to obtain effective initial adaptation scores.

4. The method for recommending clothing design parameters as described in claim 3, characterized in that, The step of marking obviously abnormal scores from all the original adaptation scores based on the evaluation difference and the degree of dispersion, and removing the obviously abnormal scores to obtain a valid initial adaptation score includes: Based on the evaluation difference and the degree of dispersion, a proportionality calculation is performed to determine the degree of dispersion ratio between the degree of dispersion and the evaluation difference of each original fit score; Each of the dispersion ratios is compared with a preset threshold. When the dispersion ratio is greater than the preset threshold, the original adaptation score corresponding to the dispersion ratio is marked as a significant abnormal score. All obviously abnormal scores are removed to obtain a valid initial fit score.

5. The method for recommending clothing design parameters as described in claim 1, characterized in that, The step of calculating anomaly detection parameters for multiple initial adaptation scores based on the adaptation difference matrix, the style correlation matrix, and multiple initial adaptation scores includes: Based on the style association matrix, the adaptation difference matrix, and the multiple initial adaptation scores, a score deviation matrix of the multiple initial adaptation scores is obtained; Based on the scoring deviation matrix and the adaptation difference matrix, the comprehensive error value of the initial adaptation score is calculated; wherein, the comprehensive error value is used to reflect the overall error of all the initial adaptation scores; Based on the comprehensive error value, the scoring deviation matrix, and the adaptation difference matrix, the anomaly detection parameters of the initial adaptation score are calculated.

6. The method for recommending clothing design parameters as described in claim 5, characterized in that, When the anomaly detection parameters indicate that there is an anomaly score among the multiple initial adaptation scores, the anomaly score is removed, including: When the anomaly detection parameters indicate that there is an anomaly score among the multiple initial adaptation scores, the adaptation index value corresponding to each initial adaptation score is extracted. Based on the style association matrix and the adaptation difference matrix, a weight distribution matrix of multiple initial adaptation scores is calculated; Based on the weight distribution matrix and the adaptation index value corresponding to each initial adaptation score, abnormal scores are filtered from multiple initial adaptation scores, and the abnormal scores are selected from multiple initial adaptation scores.

7. The method for recommending clothing design parameters as described in claim 6, characterized in that, Based on the weight distribution matrix and the adaptation index value corresponding to each initial adaptation score, abnormal scores are filtered from multiple initial adaptation scores, including: When the value at the diagonal position in the weight distribution matrix is ​​greater than any other value in the column containing that value, an abnormal indicator value corresponding to that value is selected from all the adaptation indicator values. The initial adaptation score corresponding to the abnormal indicator value is determined as the abnormal score.

8. A device for recommending clothing design parameters, characterized in that, For implementing the method according to any one of claims 1 to 7, the garment design parameter recommendation device comprises: An initial unit is used to determine the initial fit score between the target user's body shape data and multiple candidate clothing patterns, and to establish a fit difference matrix of multiple initial fit scores; wherein, the initial fit score corresponds one-to-one with the candidate clothing pattern; A construction unit is used to obtain the clothing style preferences of the target user and, based on the clothing style preferences of the target user and the style features of multiple candidate clothing patterns, construct a style correlation matrix between the initial adaptation scores; The detection unit is used to calculate anomaly detection parameters for the multiple initial adaptation scores based on the adaptation difference matrix, the style correlation matrix, and the multiple initial adaptation scores; wherein, the anomaly detection parameters are used to determine whether there are any anomalous scores among the multiple initial adaptation scores. A filtering unit is used to remove the abnormal score when the anomaly detection parameters indicate that there is an abnormal score among the multiple initial adaptation scores. The generation unit is used to generate a final clothing pattern parameter recommendation scheme for the target user based on all the retained initial adaptation scores.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the clothing design parameter recommendation method as described in any one of claims 1 to 7.

10. A computer program product that, when run on an electronic device, causes the electronic device to perform the method as described in any one of claims 1 to 7.

Images