Garment fitness prediction method

Abstract

The invention discloses a clothes fitness prediction method, which belongs to the technical field of clothes performance evaluation, and comprises the following steps: obtaining local clothes fitness scores and overall clothes fitness scores of key parts of a human body, mapping the local clothes fitness scores and the overall clothes fitness scores into fitness grades to construct an output data set, and according to under-clothes air layer thicknesses and clothes area factors of the key parts of the human body, calculating the fitness grades of the key parts of the human body; constructing an input data set by combining the garment size and the physical characteristic parameters of the virtual fabric, constructing a local garment fitness grade prediction model of the key parts of the human body by using a multilayer perceptron algorithm, and training by using the output data set and the input data set; and performing weighted summation on the trained local clothing fitness level prediction model of the key parts of the human body according to the feature weights corresponding to the local clothing fitness levels of the key parts of the human body, and constructing an overall clothing fitness level prediction model to perform overall clothing fitness prediction. The method can improve objectivity and accuracy of garment fitness prediction.

Description

Technical Field

[0001] The invention relates to a method for predicting clothing fit, and belongs to the technical field of clothing performance evaluation. Background Art

[0002] Garment fit is a key indicator of clothing's fit and comfort, serving as a core parameter in clothing design, intelligent recommendation systems, and virtual fitting platforms. Currently, the input parameters of clothing fit prediction models have expanded from basic physical parameters such as key point garment pressure and garment relaxation to more complex parameters such as 3D human models and pattern data. Innovative psychological dimensions such as user satisfaction ratings have also been incorporated. The output objective primarily focuses on predicting fit levels, while also expanding into pattern matching and looseness quantification.

[0003] However, the fit of loose-fitting garments is more significantly influenced by spatial tolerance and freedom of movement, leading to significant limitations in traditional fit prediction models for this type of garment. For one thing, loose-fitting garments typically have a large amount of looseness, which cannot generate effective pressure points in a virtual environment, potentially rendering fit prediction models ineffective. Furthermore, in garment structural design, looseness is primarily designed for two-dimensional dimensions such as bust, waist, and hips. However, garment fit is the human body's perception of the fit of clothing in three-dimensional space. Relying solely on two-dimensional planar data cannot fully represent the complex spatial relationship between clothing and the human body in a three-dimensional wearing state. Summary of the Invention

[0004] The object of the present invention is to provide a method for predicting clothing fit, which can improve the objectivity and accuracy of clothing fit prediction.

[0005] In order to achieve the above object, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for predicting clothing fit, comprising: Through human wearing tests, local clothing fit scores of key parts of the human body and overall clothing fit scores are obtained and mapped into fit grades to construct an output dataset; Based on the 3D human body model and the 3D clothing model, a 3D human clothing model is constructed and divided into local human clothing sub-models corresponding to key parts of the human body. The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-models corresponding to the key parts of the human body. The input data set is constructed by combining the clothing size and the physical property parameters of the virtual fabric. Based on the output data set and the input data set, a local clothing fit grade prediction model for key parts of the human body is constructed and trained using the multi-layer perceptron algorithm to obtain a trained local clothing fit grade prediction model for key parts of the human body; According to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, the feature weights corresponding to the local clothing fit levels of key parts of the human body are set, and the trained local clothing fit level prediction models of key parts of the human body are weighted summed according to the feature weights to construct the overall clothing fit level prediction model; The overall clothing fit grade prediction model is used to predict the overall clothing fit and obtain the overall clothing fit grade.

[0006] Combined with the first aspect, the human wearing test further includes: selecting multiple types of clothing covering different materials, sizes and styles as test samples, and having subjects with different body characteristics randomly wear the test samples in a standard test environment of constant temperature and humidity, and move according to multiple preset working postures, and respectively perform local clothing fit scores and overall clothing fit scores on key parts of the human body and the whole body.

[0007] In combination with the first aspect, further, the method for constructing a three-dimensional human body model includes: performing a full-range three-dimensional scan of a naked human body in a standard standing posture, obtaining point cloud data of the human body surface morphology, and using reverse engineering technology to encapsulate, fill holes, simplify and smooth the point cloud data to generate a three-dimensional human body model.

[0008] In combination with the first aspect, further, the method for constructing a three-dimensional clothing model includes: using three-dimensional modeling software to draw a two-dimensional clothing sample according to the actual size of the real clothing, inputting the three-dimensional human body model and the two-dimensional clothing sample into the three-dimensional virtual fitting software, using the three-dimensional human body model as a virtual fitting model, placing the two-dimensional clothing sample around the virtual model, and completing the three-dimensional forming of the clothing through virtual stitching technology, setting virtual fabric parameters in combination with the fabric properties of the clothing, and generating a three-dimensional clothing model.

[0009] In combination with the first aspect, further, constructing a three-dimensional human clothing model based on the three-dimensional human body model and the three-dimensional clothing model and dividing it into local human clothing sub-models corresponding to key parts of the human body includes: Use reverse engineering modeling software to perform spatial position matching and feature point alignment on the 3D human body model and the 3D clothing model to generate a 3D human clothing model; Taking the human skeletal joint structure as a reference, reverse engineering software is used to divide the three-dimensional human clothing model into regions according to the human anatomical parts, and the redundant parts outside the regions are deleted to generate local human clothing sub-models corresponding to the key parts of the human body.

[0010] In combination with the first aspect, further, the key parts of the human body include shoulders, chest, waist, hips and crotch, thighs, calves, upper arms, forearms, arms, upper body and lower body; The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-model corresponding to the key parts of the human body. The input data set is constructed by combining the clothing size and the physical property parameters of the virtual fabric. According to the local human clothing sub-models corresponding to the shoulders, chest, waist, hip crotch, thighs, calves, upper arms and forearms, the thickness of the air layer under the clothes at the shoulders, chest, waist, hip crotch, thighs, calves, upper arms and forearms are calculated respectively; Calculate the clothing area factors of the arms, upper body and lower body respectively according to the local human clothing sub-models corresponding to the arms, upper body and lower body; The input dataset consists of the air layer thickness under clothing, clothing size and virtual fabric physical property parameters of the shoulder, chest, waist, hip crotch, thigh, calf, upper arm and forearm, as well as the clothing area factor, clothing size and virtual fabric physical property parameters of the arm, upper body and lower body. Among them, the physical property parameters of the virtual fabric include warp tensile modulus, weft tensile modulus, diagonal tensile modulus, warp bending stiffness, weft bending stiffness, diagonal bending stiffness, dynamic friction coefficient, static friction coefficient, fabric weight and fabric thickness.

[0011] In combination with the first aspect, further, the calculation formula of the thickness of the air layer under the clothes is: ; in, Indicates the thickness of the air layer under the clothes. Represents the volume of the 3D human body model corresponding to the local human clothing sub-model, Represents the volume of the 3D clothing model corresponding to the local human clothing sub-model, Indicates the height of the local human clothing sub-model, Indicates fabric thickness; The calculation formula for clothing area factor is: ; in, represents the clothing area factor, represents the side surface area of ​​the 3D human body model corresponding to the local human clothing sub-model, Represents the side surface area of ​​the 3D clothing model corresponding to the local human clothing sub-model.

[0012] In combination with the first aspect, further, based on the output data set and the input data set, a local clothing fit level prediction model for key parts of the human body is constructed and trained using a multi-layer perceptron algorithm. Obtaining the trained local clothing fit level prediction model for key parts of the human body includes: Preprocess and encode the output and input datasets, standardize the feature variables, and divide them into training and test sets in proportion; Define a multilayer perceptron neural network model consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the characteristic dimension of the input dataset, the number of neurons in the output layer is consistent with the characteristic dimension of the output dataset, and the hidden layer adopts a "pyramid" structure with decreasing number of neurons and the activation function is the ReLU function. The cross-entropy loss function, Adam optimizer and L2 regularization term were introduced. Several rounds of training were performed using the mini-batch stochastic gradient descent method. The accuracy, precision, recall and F1 score were evaluated on the test set every 10 rounds. If the accuracy of the test set stopped improving for more than 10 rounds, the training was terminated early. The multi-layer perceptron neural network model with the best overall performance was selected as the trained local clothing fit level prediction model for key parts of the human body.

[0013] Combined with the first aspect, further, the calculation formula for the overall clothing fit grade is: ; in, Indicates the overall garment fit level. Indicates the The degree of fit of local clothing at key parts of the human body, express The corresponding feature weights, Indicates the total number of key parts of the human body.

[0014] In a second aspect, the present invention provides a clothing fit prediction device, comprising: An output dataset construction module is used to obtain local clothing fit scores and overall clothing fit scores of key parts of the human body through human wearing tests and map them into fit grades to construct an output dataset; An input dataset construction module is used to construct a 3D human clothing model based on the 3D human body model and the 3D clothing model, and divide it into local human clothing sub-models corresponding to key parts of the human body. The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-models corresponding to the key parts of the human body, and the input dataset is constructed by combining the clothing size and the physical property parameters of the virtual fabric; A prediction model construction module is used to construct and train a local clothing fit grade prediction model for key parts of the human body based on the output data set and the input data set using a multi-layer perceptron algorithm to obtain a trained local clothing fit grade prediction model for key parts of the human body; and to set feature weights corresponding to the local clothing fit grade of key parts of the human body according to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, and to perform weighted summation of the trained local clothing fit grade prediction models for key parts of the human body according to the feature weights to construct an overall clothing fit grade prediction model; The prediction module is used to predict the overall clothing fit using the overall clothing fit grade prediction model to obtain the overall clothing fit grade.

[0015] In a third aspect, the present invention provides a computer device, comprising: Storage medium for storing computer programs; A processor is used to execute the computer program to implement the clothing fit prediction method described in the first aspect.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the clothing fit prediction method described in the first aspect.

[0017] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the clothing fit prediction method described in the first aspect.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The clothing fit prediction method provided by the present invention no longer relies on traditional loose volume calculations or pressure distribution diagrams, but instead introduces the spatial morphological characteristics of the air layer under the clothing (such as thickness and area factor) as core evaluation indicators. It is particularly suitable for clothing types with large loose volume and non-close-fitting structure. It solves the problem that the existing technology is difficult to accurately evaluate the fit of loose clothing, and improves the objectivity and accuracy of clothing fit prediction.

[0019] The feature system constructed by the present invention simultaneously integrates the clothing size information, virtual fabric physical property parameters, and the thickness of the air layer under the clothing and the clothing area factor extracted through simulation, realizing data fusion in the three dimensions of human body, clothing, and simulation results, significantly improving the generalization ability and prediction accuracy of the model.

[0020] Using the multi-layer perceptron algorithm in deep learning to process high-dimensional feature data can effectively model nonlinear relationships, and the accuracy of predicting the local clothing fit level can reach 92.49%, which is better than traditional machine learning methods such as decision trees and SVMs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of a method for predicting clothing fit provided by an embodiment of the present invention; Figure 2 is a flow chart of a method for constructing an output data set provided by an embodiment of the present invention; Figure 3 is a flow chart of a method for constructing an input data set provided by an embodiment of the present invention; Figure 4 3 is a schematic diagram of the application of the clothing fit prediction method provided by an embodiment of the present invention in an online shopping scenario. DETAILED DESCRIPTION

[0022] The technical solution of the present invention will be further described in detail below in conjunction with specific implementation methods.

[0023] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention and are not to be construed as limiting the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0024] An embodiment of the present invention provides a method for predicting clothing fit, comprising: Through human wearing tests, local clothing fit scores of key parts of the human body and overall clothing fit scores are obtained and mapped into fit grades to construct an output dataset; Based on the 3D human body model and the 3D clothing model, a 3D human clothing model is constructed and divided into local human clothing sub-models corresponding to key parts of the human body. The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-models corresponding to the key parts of the human body. The input data set is constructed by combining the clothing size and the physical property parameters of the virtual fabric. Based on the output data set and the input data set, a local clothing fit grade prediction model for key parts of the human body is constructed and trained using the multi-layer perceptron algorithm to obtain a trained local clothing fit grade prediction model for key parts of the human body; According to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, the feature weights corresponding to the local clothing fit levels of key parts of the human body are set, and the trained local clothing fit level prediction models of key parts of the human body are weighted summed according to the feature weights to construct the overall clothing fit level prediction model; The overall clothing fit grade prediction model is used to predict the overall clothing fit and obtain the overall clothing fit grade.

[0025] The clothing fit prediction method provided by the embodiment of the present invention integrates two spatial morphological parameters, namely the thickness of the air layer under the clothing and the clothing area factor, and combines them with clothing size and virtual fabric physical property parameters to construct a local clothing fit grade prediction model for key parts of the human body and an overall clothing fit grade prediction model, and performs grade prediction on the local clothing fit and the overall clothing fit, which can improve the objectivity and accuracy of clothing fit prediction and is suitable for intelligent recommendation and customized design of loose clothing.

[0026] Figure 1 This is a flow chart of a method for predicting the fit of clothing provided by an embodiment of the present invention. This flow chart only shows the logical sequence of the method of this embodiment. Different methods can be used without conflict. Figure 1 The steps shown or described are accomplished in the order shown.

[0027] The clothing fit prediction method provided in an embodiment of the present invention can be applied to a terminal and can be executed by a clothing fit prediction device, which can be implemented by software and / or hardware and can be integrated into a terminal, for example: any tablet computer or computer device with communication function.

[0028] In one possible embodiment, Figure 2 As shown in the figure, the human wearing test includes: selecting multiple types of clothing covering different materials, sizes and styles as test samples, and having subjects with different body characteristics randomly wear the test samples in a standard test environment of constant temperature and humidity, and move according to multiple preset working postures, and performing local clothing fit scores and overall clothing fit scores on key parts of the human body and the whole body respectively.

[0029] Specifically, to ensure the comprehensiveness of the output data set, a variety of clothing types covering different materials, sizes, and styles were selected as test samples. A representative group of subjects with no less than 21 body shapes were recruited to conduct human wearing tests to cover a wide range of human morphological differences. The subjects were asked to wear different types of clothing and complete a variety of work movements (such as raising hands, bending over, squatting, etc.). During the wearing process, the subjects were required to score the local fit of the clothing for 11 key parts of the human body (shoulders, chest, waist, hips and crotch, thighs, calves, upper arms, forearms, arms, upper body, and lower body) in each movement state. After completing all the work movements, the overall fit of the clothing was scored, and the subjective fit perception was converted into a fit score, which was then mapped to five fit levels.

[0030] First, a five-level quantitative scale was used to subjectively evaluate the fit of clothing on key parts of the human body during various work movements. The scale ranged from -2 to 2, corresponding to different degrees of local clothing fit. Among them, -2 means "very tight or very short", -1 means "tight or short", 0 means "moderate", 1 means "loose or long", and 2 means "very loose or very long".

[0031] Secondly, the local clothing fit scores of key parts of the human body under various work movements are arithmetic averaged to obtain the local clothing fit scores of the key parts of the human body. Based on the preset grade standard, the local clothing fit scores and the overall clothing fit scores are mapped into five fit levels. Among them, level 1 means very tight or very short, level 2 means tight or short, level 3 means moderate, level 4 means loose or long, and level 5 means very loose or very long.

[0032] Finally, the output dataset consists of 11 local clothing fit grade data of key parts of the human body and 1 overall clothing fit grade data.

[0033] In one possible embodiment, Figure 3 As shown, the method for constructing a three-dimensional human body model includes: performing a full-scale three-dimensional scan of a naked human body in a standard standing posture to obtain point cloud data of the human body surface morphology, and using reverse engineering technology to encapsulate, fill holes, simplify and smooth the point cloud data to generate a three-dimensional human body model.

[0034] Specifically, a 3D scanner is used to perform a full-scale 3D scan of a naked human body in a standard standing posture, and point cloud data of the human body surface morphology is collected. Reverse engineering technology is then used to encapsulate, fill holes, simplify and smooth the point cloud data to generate a high-precision, smooth and complete encapsulated 3D human body model.

[0035] In one possible embodiment, Figure 3 As shown, the method for constructing a three-dimensional clothing model includes: using three-dimensional modeling software to draw a two-dimensional clothing sample according to the actual size of the real clothing, inputting the three-dimensional human body model and the two-dimensional clothing sample into the three-dimensional virtual fitting software, using the three-dimensional human body model as a virtual fitting model, placing the two-dimensional clothing sample around the virtual model, and completing the three-dimensional formation of the clothing through virtual stitching technology, setting virtual fabric parameters in combination with the fabric properties of the clothing, and generating a three-dimensional clothing model.

[0036] Specifically, the actual dimensions of real garments are measured, and professional 3D clothing modeling software is used to accurately draw digital 2D garment patterns that are completely consistent with the actual garment dimensions, ensuring the accuracy and standardization of the 2D garment pattern. The 3D human body model and 2D garment pattern are input into the 3D virtual fitting software, which converts the 3D human body model into a virtual fitting mannequin. The 2D garment pattern is placed around the virtual mannequin, and the garment is formed into three dimensions using virtual stitching technology. Simultaneously, virtual parameters such as the stretch and bend of the virtual fabric are set based on the fabric properties, and the color and texture of the garment are adjusted. Ultimately, a 3D garment model is generated, with each subject wearing different styles of clothing.

[0037] In one possible embodiment, Figure 3 As shown, constructing a 3D human clothing model based on a 3D human body model and a 3D clothing model and dividing it into local human clothing sub-models corresponding to key parts of the human body specifically includes the following steps: Step 1: Use reverse engineering modeling software to perform spatial position matching and feature point alignment on the 3D human body model and the 3D clothing model to generate a 3D human clothing model; Specifically, reverse engineering modeling software is used to match the spatial positions of the three-dimensional human body model and the three-dimensional clothing model. By aligning the feature points, the three-dimensional human body model and the three-dimensional clothing model are accurately aligned, thereby obtaining a three-dimensional human clothing model that fits the actual wearing state.

[0038] Step 2: Using the human skeletal joint structure as a reference, use reverse engineering software to divide the 3D human clothing model into regions according to the human anatomical parts, delete the redundant parts outside the region, and generate local human clothing sub-models corresponding to the key parts of the human body.

[0039] Specifically, reverse engineering modeling software is used to divide the three-dimensional human clothing model into multiple independent local areas according to the human anatomical parts and with the skeletal joint structure as a reference. The redundant parts beyond the preset range are deleted, and finally the local human clothing sub-models corresponding to 11 key parts of the human body, including shoulders, chest, waist, hips and crotch, thighs, calves, upper arms, forearms, arms, upper torso and lower torso, are obtained.

[0040] In one possible embodiment, the following steps are specifically used to calculate the thickness of the air layer under clothing and the clothing area factor at the key parts of the human body based on the local human clothing sub-model corresponding to the key parts of the human body, and to construct an input data set based on the clothing size and the physical property parameters of the virtual fabric: Step 1: Based on the local human clothing sub-models corresponding to the shoulders, chest, waist, hip crotch, thigh, calf, upper arm and forearm, calculate the thickness of the air layer under the clothes at the shoulders, chest, waist, hip crotch, thigh, calf, upper arm and forearm respectively; Specifically, the calculation formula for the thickness of the air layer under the clothes is: ; in, Indicates the thickness of the air layer under the clothes, in mm. Indicates the volume of the 3D human body model corresponding to the local human clothing sub-model, in mm 3 , Indicates the volume of the 3D clothing model corresponding to the local human clothing sub-model, in mm 3 , Indicates the height of the local human clothing sub-model, in mm. Indicates the fabric thickness in mm.

[0041] Step 2: Calculate the clothing area factors of the arms, upper body, and lower body respectively based on the local human clothing sub-models corresponding to the arms, upper body, and lower body; Specifically, the calculation formula of clothing area factor is: ; in, represents the clothing area factor, dimensionless, Indicates the side surface area of ​​the 3D human body model corresponding to the local human clothing sub-model, in mm 2 , Indicates the side surface area of ​​the 3D clothing model corresponding to the local human clothing sub-model, in mm 2 .

[0042] Step 3: The input data set is composed of the air layer thickness under the clothes, clothing size and virtual fabric physical property parameters of the shoulders, chest, waist, hip crotch, thighs, calves, upper arms and forearms, as well as the clothing area factor, clothing size and virtual fabric physical property parameters of the arms, upper body and lower body.

[0043] Specifically, the physical property parameters of the virtual fabric include warp tensile modulus, weft tensile modulus, diagonal tensile modulus, warp bending stiffness, weft bending stiffness, diagonal bending stiffness, dynamic friction coefficient, static friction coefficient, fabric weight and fabric thickness.

[0044] In this embodiment, a three-dimensional human body model is constructed for each subject using three-dimensional scanning technology. A three-dimensional human body clothing model is constructed for each subject wearing each type of clothing, and structured segmentation is performed based on anatomical features to obtain local human clothing sub-models corresponding to 11 key human body parts. The thickness of the air layer under the clothing is calculated for the local human clothing sub-models corresponding to eight key human body parts, namely the shoulder, chest, waist, hip crotch, thigh, calf, upper arm, and forearm. The clothing area factor is calculated for the local human clothing sub-models corresponding to three key human body parts, namely the arm, upper torso, and lower torso. The clothing size, virtual fabric physical property parameters, and the calculated air layer thickness under the clothing or clothing area factor are integrated to construct an input data set, making the input features more comprehensive.

[0045] In one possible embodiment, based on the output data set and the input data set, a local clothing fit level prediction model for key parts of the human body is constructed using a multi-layer perceptron algorithm and trained. Obtaining the trained local clothing fit level prediction model for key parts of the human body specifically includes the following steps: Step 1: Preprocess and encode the output dataset and input dataset, standardize the feature variables, and divide them into training set and test set in proportion; Specifically, the output dataset and input dataset are preprocessed and encoded, and the feature variables are standardized and divided into training set and test set in a ratio of 7:3 to ensure the effectiveness of model training and evaluation.

[0046] Step 2: Define a multi-layer perceptron (MLP) neural network model consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the characteristic dimension of the input dataset, the number of neurons in the output layer is consistent with the characteristic dimension of the output dataset, and the hidden layer adopts a "pyramid" structure with decreasing number of neurons and the activation function is the ReLU function. Specifically, the hidden layer is set to 2 to 3 layers, and a "pyramid" structure is used to decrease the number of neurons. The hidden layer activation function uniformly uses the ReLU function to enhance the nonlinear fitting ability.

[0047] Step 3: Introduce the cross-entropy loss function, Adam optimizer, and L2 regularization term, and train for several rounds using the mini-batch stochastic gradient descent method. Evaluate the accuracy, precision, recall, and F1 score on the test set every 10 rounds. If the accuracy of the test set stops improving for more than 10 rounds, terminate the training early and select the multi-layer perceptron neural network model with the best overall performance as the trained local clothing fit level prediction model for key parts of the human body.

[0048] Specifically, the cross-entropy loss function is used to measure the difference between the predicted and true grades output by the local clothing fit prediction model for key body parts. The Adam optimizer is used, and an L2 regularization term (weight_decay parameter) is introduced to prevent overfitting. During training, mini-batch stochastic gradient descent (e.g., batch size = 64) is used, with a total training round number of 100 to 150 rounds. The training loss and accuracy trends are recorded after each round. After every 10 rounds of training, the accuracy, precision, recall, and F1 score are evaluated on the test set, and the loss function and accuracy curve are plotted to monitor overfitting. If the accuracy on the test set stops improving for more than 10 rounds, training is terminated early. The multi-layer perceptron neural network model with the best overall performance on the test set is selected as the trained local clothing fit prediction model for key body parts.

[0049] Based on the measured data of 21 subjects, an MLP local clothing fit prediction model was constructed for 11 key human body parts to conduct local clothing fit prediction experiments. The experimental results showed an average accuracy of 92.49%, a precision of 89.11%, a recall of 90.96%, and an F1 score of 89.86%, significantly outperforming traditional models such as logistic regression, random forest, and support vector machine.

[0050] In this embodiment, the calculation formula for the overall garment fit grade is: ; in, Indicates the overall garment fit level. Indicates the The degree of fit of local clothing at key parts of the human body, express The corresponding feature weights, Indicates the total number of key parts of the human body.

[0051] In this embodiment, according to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, the feature weights corresponding to the local clothing fit levels of key parts of the human body are set, and the local clothing fit levels of key parts of the human body are weighted and summed according to the feature weights to obtain the overall clothing fit level, thereby realizing structural combination prediction.

[0052] Specifically, the feature weights corresponding to the local clothing fit levels of the shoulder, chest, waist, hip crotch, thigh, calf, upper arm, forearm, arm, upper torso and lower torso are set to 0.069, 0.093, 0.108, 0.106, 0.108, 0.105, 0.084, 0.053, 0.091, 0.097 and 0.087 respectively.

[0053] The clothing fit prediction method provided by the embodiment of the present invention clarifies the structural attribution relationship of 11 key parts of the human body in clothing fit, and assigns different feature weights to each key part of the human body, thereby establishing a combined prediction path of "local fit → overall fit". The prediction structure is clearer and more interpretable, which is conducive to achieving high-precision personalized recommendations in clothing e-commerce, virtual fitting platforms, and industrial clothing design.

[0054] The embodiment of the present invention provides an application of a clothing fit prediction method in an online shopping scenario, which can realize accurate size recommendations for end users, such as Figure 4 As shown, the specific process is as follows: End users select clothing sizes on e-commerce platforms and input body size data (such as height, weight, shoulder width, chest circumference, waist circumference, hip circumference, limb length, etc.) through the platform's interactive interface; after receiving the body size data, the system automatically calls the 3D modeling software to build a 3D body model for the end user; then, the system imports the 3D body model into the virtual fitting software, and simultaneously loads the digital 2D sample of the user's selected clothing and the preset virtual fabric physical property parameters to complete the virtual fitting.

[0055] After the virtual fitting is completed, the system uses reverse engineering technology to perform structured segmentation on the three-dimensional human clothing model, extracts the local human clothing sub-models corresponding to 11 key parts of the human body, and calculates the thickness of the air layer under the clothes in 8 parts of the shoulders, chest, waist, hips and crotch, thighs, calves, upper arms and forearms, as well as the clothing area factors of 3 parts of the arms, upper torso and lower torso; after integrating these air layer morphological characteristic parameters with the clothing size and virtual fabric physical property parameters, they are input into the trained local clothing fit level prediction model of 11 key parts of the human body to obtain the local clothing fit levels of the 11 key parts of the human body; then, the local clothing fit levels of the 11 key parts of the human body are weighted and summed according to the feature weights to obtain the overall clothing fit level and feedback to the end user.

[0056] If the end user is satisfied with the recommendation results, they can complete the purchase directly; if not, the system will automatically recommend adjusted clothing sizes (such as increasing or decreasing the size) based on the current prediction results, repeating the virtual fitting, feature extraction, and fit prediction process until the user obtains a satisfactory fit level, thus realizing a closed-loop recommendation of "try on-prediction-adjustment-confirmation", effectively improving the size selection efficiency and user experience of online shopping.

[0057] In this example, a subject, 179 cm tall and 93 kg, was selected for a one-piece garment of appropriate size. Three to five sizes of the same fabric were selected for virtual fitting (in this example, all sizes of pure cotton fabric, including M, L, XL, 2XL, and 3XL, were virtually fitted). The predicted local fit grades were then weighted and summed to calculate the overall fit grade. The predicted results are shown in Table 1. The overall fit grade for size 2XL is 3.094, which is closest to the "moderate" grade (grade 3). This indicates that size 2XL is more suitable for this subject than other sizes. Therefore, the system recommends a size 2XL one-piece garment.

[0058] Table 1: Prediction results of local clothing fit grade and overall clothing fit grade .

[0059] An embodiment of the present invention provides a clothing fit prediction device, comprising: An output dataset construction module is used to obtain local clothing fit scores and overall clothing fit scores of key parts of the human body through human wearing tests and map them into fit grades to construct an output dataset; An input dataset construction module is used to construct a 3D human clothing model based on the 3D human body model and the 3D clothing model, and divide it into local human clothing sub-models corresponding to key parts of the human body. The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-models corresponding to the key parts of the human body, and the input dataset is constructed by combining the clothing size and the physical property parameters of the virtual fabric; A prediction model construction module is used to construct and train a local clothing fit grade prediction model for key parts of the human body based on the output data set and the input data set using a multi-layer perceptron algorithm to obtain a trained local clothing fit grade prediction model for key parts of the human body; and to set feature weights corresponding to the local clothing fit grade of key parts of the human body according to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, and to perform weighted summation of the trained local clothing fit grade prediction models for key parts of the human body according to the feature weights to construct an overall clothing fit grade prediction model; The prediction module is used to predict the overall clothing fit using the overall clothing fit grade prediction model to obtain the overall clothing fit grade.

[0060] The clothing fit prediction device provided by the embodiment of the present invention can execute the clothing fit prediction method provided by the embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0061] An embodiment of the present invention provides a computer device, including: Storage medium for storing computer programs; The processor is configured to execute a computer program to implement the garment fit prediction method provided by an embodiment of the present invention.

[0062] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method for predicting clothing fit provided by the embodiment of the present invention is implemented.

[0063] This embodiment provides a computer program product, including a computer program. When the computer program is executed by a processor, the method for predicting clothing fit provided by the embodiment of the present invention is implemented.

[0064] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROMs, optical storage, etc.) containing computer-usable program code.

[0065] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0066] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0067] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0068] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting clothing fit, characterized in that: include: Through human wearing tests, local clothing fit scores of key parts of the human body and overall clothing fit scores are obtained and mapped into fit grades to construct an output dataset; Based on the 3D human body model and the 3D clothing model, a 3D human clothing model is constructed and divided into local human clothing sub-models corresponding to key parts of the human body. The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-models corresponding to the key parts of the human body. The input data set is constructed by combining the clothing size and the physical property parameters of the virtual fabric. Based on the output data set and the input data set, a local clothing fit grade prediction model for key parts of the human body is constructed and trained using the multi-layer perceptron algorithm to obtain a trained local clothing fit grade prediction model for key parts of the human body; According to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, the feature weights corresponding to the local clothing fit levels of key parts of the human body are set, and the trained local clothing fit level prediction models of key parts of the human body are weighted summed according to the feature weights to construct the overall clothing fit level prediction model; The overall clothing fit grade prediction model is used to predict the overall clothing fit and obtain the overall clothing fit grade.

2. The clothing fit prediction method according to claim 1, wherein: The human wearing test includes: selecting multiple types of clothing covering different materials, sizes and styles as test samples, and having subjects with different body characteristics randomly wear the test samples in a standard test environment of constant temperature and humidity, and move according to multiple preset working postures. The local clothing fit score and the overall clothing fit score are respectively conducted on the key parts and the whole body.

3. The clothing fit prediction method according to claim 1, wherein: The method for constructing a three-dimensional human body model includes: performing a full-scale three-dimensional scan of a naked human body in a standard standing posture to obtain point cloud data of the human body surface morphology, and using reverse engineering technology to encapsulate, fill holes, simplify and smooth the point cloud data to generate a three-dimensional human body model.

4. The clothing fit prediction method according to claim 1, wherein: The method for constructing a three-dimensional clothing model includes: using three-dimensional modeling software to draw a two-dimensional clothing sample according to the actual size of the real clothing, inputting the three-dimensional human body model and the two-dimensional clothing sample into the three-dimensional virtual fitting software, using the three-dimensional human body model as a virtual fitting model, placing the two-dimensional clothing sample around the virtual model, and completing the three-dimensional formation of the clothing through virtual stitching technology, setting virtual fabric parameters in combination with the fabric properties of the clothing, and generating a three-dimensional clothing model.

5. The clothing fit prediction method according to claim 1, wherein: The 3D human clothing model is constructed based on the 3D human body model and the 3D clothing model and divided into local human clothing sub-models corresponding to key parts of the human body, including: Use reverse engineering modeling software to perform spatial position matching and feature point alignment on the 3D human body model and the 3D clothing model to generate a 3D human clothing model; Taking the human skeletal joint structure as a reference, reverse engineering software is used to divide the three-dimensional human clothing model into regions according to the human anatomical parts, and the redundant parts outside the regions are deleted to generate local human clothing sub-models corresponding to the key parts of the human body.

6. The clothing fit prediction method according to claim 1, wherein: Key parts of the human body include shoulders, chest, waist, hips and crotch, thighs, calves, upper arms, forearms, arms, upper body and lower body; The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-model corresponding to the key parts of the human body. The input data set is constructed by combining the clothing size and the physical property parameters of the virtual fabric. According to the local human clothing sub-models corresponding to the shoulders, chest, waist, hip crotch, thighs, calves, upper arms and forearms, the thickness of the air layer under the clothes at the shoulders, chest, waist, hip crotch, thighs, calves, upper arms and forearms are calculated respectively; Calculate the clothing area factors of the arms, upper body and lower body respectively according to the local human clothing sub-models corresponding to the arms, upper body and lower body; The input dataset consists of the air layer thickness under clothing, clothing size and virtual fabric physical property parameters of the shoulder, chest, waist, hip crotch, thigh, calf, upper arm and forearm, as well as the clothing area factor, clothing size and virtual fabric physical property parameters of the arm, upper body and lower body. Among them, the physical property parameters of the virtual fabric include warp tensile modulus, weft tensile modulus, diagonal tensile modulus, warp bending stiffness, weft bending stiffness, diagonal bending stiffness, dynamic friction coefficient, static friction coefficient, fabric weight and fabric thickness.

7. The clothing fit prediction method according to claim 1, wherein: The calculation formula for the thickness of the air layer under the clothes is: ； in, Indicates the thickness of the air layer under the clothes. Represents the volume of the 3D human body model corresponding to the local human clothing sub-model, Represents the volume of the 3D clothing model corresponding to the local human clothing sub-model, Indicates the height of the local human clothing sub-model, Indicates fabric thickness; The calculation formula of clothing area factor is: ； in, represents the clothing area factor, represents the side surface area of ​​the 3D human body model corresponding to the local human clothing sub-model, Represents the side surface area of ​​the 3D clothing model corresponding to the local human clothing sub-model.

8. The clothing fit prediction method according to claim 1, wherein: Based on the output data set and the input data set, a local clothing fit level prediction model for key parts of the human body is constructed and trained using the multi-layer perceptron algorithm. The trained local clothing fit level prediction model for key parts of the human body includes: Preprocess and encode the output and input datasets, standardize the feature variables, and divide them into training and test sets in proportion; Define a multilayer perceptron neural network model consisting of an input layer, a hidden layer, and an output layer. The number of neurons in the input layer is consistent with the characteristic dimension of the input dataset, the number of neurons in the output layer is consistent with the characteristic dimension of the output dataset, and the hidden layer adopts a "pyramid" structure with decreasing number of neurons and the activation function is the ReLU function. The cross-entropy loss function, Adam optimizer and L2 regularization term were introduced. Several rounds of training were performed using the mini-batch stochastic gradient descent method. The accuracy, precision, recall and F1 score were evaluated on the test set every 10 rounds. If the accuracy of the test set stopped improving for more than 10 rounds, the training was terminated early. The multi-layer perceptron neural network model with the best overall performance was selected as the trained local clothing fit level prediction model for key parts of the human body.

9. The clothing fit prediction method according to claim 1, wherein: The formula for calculating the overall garment fit grade is: ； in, Indicates the overall garment fit level. Indicates the The degree of fit of local clothing at key parts of the human body, express The corresponding feature weights, Indicates the total number of key parts of the human body.

10. A clothing fit prediction device, characterized in that: include: An output dataset construction module is used to obtain local clothing fit scores and overall clothing fit scores of key parts of the human body through human wearing tests and map them into fit grades to construct an output dataset; An input dataset construction module is used to construct a 3D human clothing model based on the 3D human body model and the 3D clothing model, and divide it into local human clothing sub-models corresponding to key parts of the human body. The thickness of the air layer under the clothing and the clothing area factor of the key parts of the human body are calculated based on the local human clothing sub-models corresponding to the key parts of the human body, and the input dataset is constructed by combining the clothing size and the physical property parameters of the virtual fabric; A prediction model construction module is used to construct and train a local clothing fit grade prediction model for key parts of the human body based on the output data set and the input data set using a multi-layer perceptron algorithm to obtain a trained local clothing fit grade prediction model for key parts of the human body; and to set feature weights corresponding to the local clothing fit grade of key parts of the human body according to the contribution of the local clothing fit of key parts of the human body to the overall clothing fit, and to perform weighted summation of the trained local clothing fit grade prediction models for key parts of the human body according to the feature weights to construct an overall clothing fit grade prediction model; The prediction module is used to predict the overall clothing fit using the overall clothing fit grade prediction model to obtain the overall clothing fit grade.

Images