A method and device for trying on clothing

Abstract

The present invention discloses a method and device for trying on clothing. The method for trying on clothing includes: constructing a standard human body model and standard clothing models that match the standard human body model and correspond to each piece of clothing; obtaining the body shape information of a user and the clothing selected by the user, and correcting the bone feature points and outer contour feature points of the standard human body model and the standard clothing model corresponding to the clothing selected by the user according to the body shape information to obtain a corrected human body model that matches the body shape information and a corrected clothing model that matches the corrected human body model; covering the model picture of the corrected clothing model on the model picture of the corrected human body model and presenting the try-on effect to the user. The method and device for trying on clothing according to the present invention can perform adaptive adjustment for users with different body shapes, so as to provide try-on effects that conform to the body shapes of different users and improve the user experience.

Description

Technical Field

[0001] The present invention relates to the technical field of virtual fitting, and more specifically, to a method and device for trying on and wearing clothing. Background Art

[0002] In the fashion industry, new products are introduced very quickly. Large clothing accessory suppliers usually have a huge product library, offering a large number of products to consumers.

[0003] In offline service scenarios, due to the limited display area of sales stores, usually only samples of some clothes can be displayed in the display area. The quantity of the same style of clothes displayed is limited, and there may even be only one piece of each style, reducing the probability that consumers can choose the clothes they like. In addition, it is time-consuming and laborious for consumers to try on clothes in the store, and the size they can wear may not be available for the styles they like.

[0004] To solve the above problems, some suppliers have launched online service scenarios. That is, through an application on a mobile device, users can see the fitting effects of the clothes they like on standard models. However, consumers can only judge the matching degree between the clothes and their own body types based on the fitting effects of the standard models, and cannot intuitively see the effects after putting on the clothes, resulting in serious after-sales situations after purchasing products, affecting the user experience and increasing the sales costs of suppliers. Summary of the Invention

[0005] The purpose of the present invention is to solve the above technical problems or defects, and provide a method and device for trying on and wearing clothing, which can perform adaptive adjustment for users with different body types to provide fitting and wearing effects that meet the body types of different users, and improve the user experience.

[0006] To achieve the above purpose, the first aspect of the present invention provides Technical Solution 1: A method for trying on and wearing clothing, which includes: Step 1: Construct a standard human body model and standard clothing models that match the standard human body model and correspond to each piece of clothing; both the standard human body model and the standard clothing models include model pictures marked with bone feature points and outer contour feature points; Step 2: Obtain the body type information of the user and the clothing selected by the user, and correct the bone feature points and outer contour feature points of the standard human body model and the standard clothing model corresponding to the clothing selected by the user according to the body type information to obtain a corrected human body model that matches the body type information and a corrected clothing model that matches the corrected human body model; Step 3: Cover the model picture of the corrected clothing model on the model picture of the corrected human body model and display the fitting and wearing effect to the user.

[0007] In Technical Solution 1, considering that different users not only have different bone characteristics, but may also have different outer contour characteristics even when their bone characteristics are the same, the present invention corrects both the bone feature points and the outer contour feature points of the standard human body model and the standard clothing model according to the user's body type information, so that the corrected human body model and clothing model highly match the user's body type in terms of both bone characteristics and outer contour characteristics, thereby being able to display a trial wearing effect adapted to the user's body type. After the user purchases clothing based on this trial wearing effect, the actual trial wearing effect will not deviate too far from the virtual trial wearing effect of the present invention, thus having a good user experience and being able to effectively reduce the sales cost of the supplier.

[0008] Based on Technical Solution 1, the present invention also provides Technical Solution 2: The standard human body model and the standard clothing model in Step 1 are constructed through the following steps: Step 11: Take photos of a standard model and photos of each piece of clothing worn on the standard model; Step 12: Use the standard model and each piece of clothing as the main objects, perform background removal on the obtained photos, and obtain model pictures with consistent sizes; Step 13: Mark the pixel points corresponding to the bones of the standard model on the model pictures as bone feature points, and mark the pixel points corresponding to the outer contours of the corresponding main objects on the model pictures as outer contour feature points.

[0009] Technical Solution 2 provides a specific implementation method for constructing the standard human body model and the standard clothing model. By performing background removal on the obtained photos through Technical Solution 2 and obtaining model pictures with consistent sizes, when performing overlay operations or other operations on the model pictures of the human body model and the clothing model subsequently, there is no need to perform benchmark alignment processing on the two, simplifying the processing amount during subsequent operations and improving accuracy.

[0010] Based on Technical Solution 2, the present invention also provides Technical Solution 3: Step 2 is specifically as follows: Step 21: Obtain and correct the bone feature points and outer contour feature points of the standard human body model according to the user's body type information to obtain the corrected human body model; Step 22: Calculate and respectively correct the bone feature points and outer contour feature points of the standard clothing model corresponding to the clothing after obtaining the clothing selected by the user according to the differences between the corrected human body model and the standard human body model in terms of bone feature points and outer contour feature points to obtain the corrected clothing model.

[0011] Solution 3 provides an implementation method for correcting a standard human body model and a standard clothing model. First, it corrects the bone feature points and outer contour feature points of the standard human body model, and then maps the differences between the corrected human body model and the standard human body model at the bone feature points and outer contour feature points to the standard clothing model correspondingly, so as to complete the correction of the standard clothing model at the bone feature points and outer contour feature points respectively. This not only overcomes the problem that it is difficult to correct the clothing model according to the user's body type information because there is no direct association between the clothing model and the user's body type information, but also simplifies the computational amount in the process of correcting the standard clothing model and improves the processing efficiency.

[0012] Based on Solution 3, the present invention also provides Solution 4: The body type information includes multiple body type indicators. Before step 21, a body type database is also established. The body type database stores the correspondence between body type indicators and multiple corresponding bone feature points. The correction of the bone feature points of the standard human body model in step 21 is carried out through the following steps: Step 211: Obtain and call, according to the body type information, the correspondence between the body type indicators included in the body type information and the corresponding bone feature points in the body type database, and calculate the positions of the target bone feature points corresponding to the body type information; Step 212: Adjust the bone feature points of the standard human body model according to the target bone feature points to obtain an intermediate human body model.

[0013] Solution 4 provides a specific implementation method for correcting the bone feature points of the standard human body model. By establishing a body type database in advance, the correspondence between body type indicators and bone feature points can be called according to the user's body type information and the target bone feature points can be calculated. According to the target bone feature points, the bone feature points of the standard human body model can be adjusted to obtain an intermediate human body model that highly coincides with the user's bones.

[0014] Based on Solution 4, the present invention also provides Solution 5: The body type database is constructed through the following steps: Step A: Take sample photos of multiple sample models with different body type information; Step B1: Use the AlphaPose system to train the bone feature points of each sample photo to obtain the bone feature points of each sample photo; Step C1: Establish a re - regression model function of each body type indicator and bone feature points according to the body type information of each sample model and the bone feature points of the corresponding sample photo.

[0015] Technical Solution 5 provides a specific implementation method for constructing a body shape database so that it has a corresponding relationship between body shape indicators and bone feature points. It obtains the bone feature points of the sample photos by using artificial intelligence training on a large number of sample photos corresponding to different body shape information, and uses the obtained bone feature points to establish a multi-regression function model with the body shape information, so as to facilitate calling the multi-regression function model according to the user's body shape information and calculating the bone feature points corresponding to the user's body shape information.

[0016] Based on technical solution four, the present invention also provides technical solution six: the multiple body shape indicators include at least one pixel association indicator, and the body shape database also stores the corresponding relationship between the pixel association indicator and the number of pixels of the model image of the human body model; the correction of the outer contour feature points of the standard human body model in the step 21 is performed by the following steps: step 213: calculating the number of pixels N1 of the intermediate human body model; step 214: calling the corresponding relationship between the pixel association indicator and the number of pixels of the model image in the body shape database according to the pixel association indicator of the body shape information, and calculating the target number of pixels N2 corresponding to the pixel association indicator; step 215: according to the difference between the number of pixels N1 and the target number of pixels N2 of the intermediate human body model, and taking other body shape indicators as fitting objects, using an image deformation algorithm to adjust the outer contour feature points of the intermediate human body model to obtain the corrected human body model; wherein the other body shape indicators include at least one of a chest circumference indicator, a waist circumference indicator and a hip circumference indicator.

[0017] Technical solution 6 provides a specific implementation method for correcting the outer contour feature points of the standard human body model, which establishes a body shape database including the correspondence between pixel association index and pixel number in advance, calls the correspondence between the pixel association index and the pixel number according to the user's body shape information to calculate the target pixel number, and then uses the image deformation algorithm to adjust the outer contour feature points based on the difference in the number of pixels, taking the chest circumference index, waist circumference index and hip circumference index as the fitting object, and uses the image deformation algorithm to adjust the outer contour feature points, which can well eliminate the difference in the number of pixels (i.e., the target number of pixels) that the intermediate human body model and the model image corresponding to the user's body shape should have, and obtain a corrected human body model that is highly consistent with the user's outer contour features. It can be seen that technical solution 6 fully considers the parts that have a greater impact on the user's outer contour features, and uses the different pixel numbers reflected in the model image corresponding to the user's body shape as the target value, and adjusts the outer contour feature points of the human body model based on the three-dimensional index as the fitting object, so that the outer contour feature points of the human body model can be accurately adjusted in accordance with the user's body shape, and the obtained corrected human body model has a high degree of consistency with the user's outer contour features, and the matching effect is good.

[0018] Based on Technical Solution Six, the present invention also provides Technical Solution Seven: The body shape database is constructed through the following steps: Step A: Take sample photos of multiple sample models with different body shape information; Step B2: Use a deep learning algorithm to train a semantic segmentation model for each sample photo, and obtain and calculate the number of pixels of the semantic segmentation image of each sample photo; Step C2: Establish a multiple regression model function between the pixel association index in the body shape information of each sample model and the number of pixels of the corresponding semantic segmentation image of the sample photo.

[0019] Technical Solution Seven provides a specific implementation method for constructing a body shape database to have a corresponding relationship between pixel association indexes and the number of pixels. It uses a deep learning algorithm to train a semantic segmentation model for a large number of sample photos corresponding to different body shape information to obtain semantic segmentation images, and uses the multiple regression function model of the number of pixels of the obtained semantic segmentation images and the pixel association indexes, which is convenient for calling the multiple regression function model according to the user's body shape information and calculating the target number of pixels corresponding to the user's body shape information, and then using it to adjust the outer contour feature points of the standard clothing model.

[0020] Based on Technical Solution Six, the present invention also provides Technical Solution Eight: The specific content of Step 22 is as follows: According to the difference between the bone feature points of the intermediate human body model and the standard human body model, adjust the bone feature points of the standard clothing model to obtain an intermediate clothing model; According to the difference between the outer contour feature points of the corrected human body model and the intermediate human body model, adjust the outer contour feature points of the intermediate clothing model to obtain the corrected clothing model.

[0021] Technical Solution Eight provides a specific implementation method for correcting the bone feature points and outer contour feature points of the standard clothing model. By mapping the correction results of the standard human body model on the bone feature points and outer contour feature points to the correction of the standard clothing model on the bone feature points and outer contour feature points respectively, the computational amount of the correction process of the standard clothing model is simplified, and the processing efficiency is improved.

[0022] Based on Technical Solution One, the present invention also provides Technical Solution Nine: The method further includes: Obtain the user's real photo; Use the area corresponding to the user's face in the user's real photo to replace the area corresponding to the face of the standard human body in the model picture of the corrected human body model to obtain a corrected human body model after face correction; Step 3 is to cover the model picture of the corrected clothing model on the model picture of the corrected human body model after face correction and display the try-on effect to the user.

[0023] Technical solution nine is a technical solution for functional expandability. By cutting out the face part of the user's real photo and using it to replace and correct the corresponding part of the human body model, the virtual try-on effect of the present invention is made more realistic and interesting.

[0024] To achieve the above object, the second aspect of the present invention provides technical solution ten: a clothing try-on device, including a processor; the processor is used to execute the clothing try-on method described in any one of the above technical solutions.

[0025] Technical solution ten provides a device corresponding to the above clothing try-on method, which inherits all the advantages of the above method. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0027] Figure 1 It is a schematic diagram when the try-on method of the embodiment of the present invention takes a model photo;

[0028] Figure 2 It is a schematic diagram of some body shape information obtained by the try-on method of the embodiment of the present invention;

[0029] Figure 3 It is a schematic diagram of the human body model of the try-on method of the embodiment of the present invention, on which bone feature points are marked;

[0030] Figure 4 It is a schematic diagram of the human body model of the try-on method of the embodiment of the present invention, on which outer contour feature points are marked;

[0031] Figure 5 It is a schematic diagram of the clothing model of the try-on method of the embodiment of the present invention, on which bone feature points are marked;

[0032] Figure 6 It is a schematic diagram of the clothing model of the try-on method of the embodiment of the present invention, on which outer contour feature points are marked. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are the preferred embodiments of the present invention and should not be regarded as excluding other embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0034] In the claims, description and above-mentioned drawings of the present invention, unless otherwise clearly defined, when using terms such as "first", "second" or "third", etc., they are used to distinguish different objects rather than to describe a specific order.

[0035] In the claims, description and above-mentioned drawings of the present invention, when using terms such as "comprising", "having" and their variants, the intention is "including but not limited to".

[0036] Referring to Figure 1-6 , an embodiment of the present invention provides a method for trying on clothing, which includes:

[0037] Step 1: Construct a standard human body model and standard clothing models that match the standard human body model and correspond to each piece of clothing. Both the standard human body model and the standard clothing models include model pictures marked with bone feature points and outer contour feature points.

[0038] Step 2: Obtain the body shape information of the user and the clothing selected by the user, and correct the bone feature points and outer contour feature points of the standard human body model and the standard clothing model corresponding to the clothing selected by the user according to the body shape information to obtain a corrected human body model that matches the body shape information and a corrected clothing model that matches the corrected human body model.

[0039] Step 3: Cover the model picture of the corrected clothing model on the model picture of the corrected human body model and show the try-on effect to the user.

[0040] It should be noted that the clothing referred to in the present invention is not limited to clothes, trousers and other garments, but may also include items suitable for the user to wear such as shoes, bags, accessories, etc. In addition, when the user needs to try on multiple pieces of clothing, the method of the embodiment of the present invention will use the model picture of the upper-layer clothing model as the basis for cutting, cut the model picture of the lower-layer clothing model, and then stack them on top of each other to form a single-sided try-on effect. The specific implementation method of this cutting process is already in the prior art, so it will not be elaborated here.

[0041] It can be seen that the embodiments of the present invention take into account that different users not only have different bone characteristics, but may also have different outer contour characteristics even when their bone characteristics are the same. Therefore, the embodiments of the present invention correct both the bone feature points and the outer contour feature points of the standard human body model and the standard clothing model according to the user body type information, so that the corrected human body model and clothing model are highly matched with the user body type in terms of both bone characteristics and outer contour characteristics, thereby being able to display a trial wearing effect adapted to the user body type. After the user purchases clothing based on this trial wearing effect, the actual trial wearing effect will not deviate too far from the virtual trial wearing effect of the present invention, thus having a good user experience and being able to effectively reduce the sales cost of the supplier.

[0042] The following details the specific implementation methods of the above steps.

[0043] Specifically, step 1 is as follows:

[0044] Step 11: Take photos of a standard model and photos of each piece of clothing worn on the standard model. In this embodiment, when taking photos, the model platform can be configured to be rotatable to take photos from multiple angles and provide users with different trial wearing preview angles. Further, the method of the embodiments of the present invention can also adopt a parallax scrolling method in step 3 to enhance the dynamic effect when users preview from multiple angles.

[0045] Step 12: Use the standard model and each piece of clothing as the main objects, perform background removal processing on the photos obtained by shooting, and obtain model pictures with the same size. The background removal processing is to remove other pixel points in the photo except the main part and form a transparent area; the same size means that the resolution of the pictures is the same.

[0046] Step 13: Mark the pixel points corresponding to the bones of the standard model on the model picture as bone feature points, and mark the pixel points corresponding to the outer contours of the corresponding main objects on the model picture as outer contour feature points. In this embodiment, the marking of the bone feature points and outer contour feature points on the model picture can be completed by recording the coordinates of the pixel points in a description file associated with the model picture.

[0047] It can be understood that the embodiments of the present invention perform background removal processing on the photos obtained by shooting and obtain model pictures with the same size, so that when performing overlay operations or other operations on the model pictures of the human body model and the clothing model in the subsequent process, there is no need to perform benchmark alignment processing on the two, which simplifies the processing amount during subsequent operations and improves the accuracy.

[0048] Specifically, step 2 is as follows:

[0049] Step 21: Obtain and correct the bone feature points and outer contour feature points of the standard human body model according to the body shape information of the user, so as to obtain the corrected human body model.

[0050] Step 22: Calculate and respectively correct the bone feature points and outer contour feature points of the standard clothing model corresponding to the clothing according to the differences between the corrected human body model and the standard human body model in terms of bone feature points and outer contour feature points after obtaining the clothing selected by the user, so as to obtain the corrected clothing model.

[0051] In the embodiment of the present invention, the bone feature points and outer contour feature points of the standard human body model are first corrected, and then the differences between the corrected human body model and the standard human body model in terms of bone feature points and outer contour feature points are correspondingly mapped to the standard clothing model, so as to respectively complete the correction of the standard clothing model in terms of bone feature points and outer contour feature points. This not only overcomes the problem that it is difficult to correct the clothing model according to the user's body shape information because there is no direct association between the clothing model and the user's body shape information, but also simplifies the computational amount of the correction process of the standard clothing model and improves the processing efficiency.

[0052] Specifically, the body shape information includes multiple body shape indexes, which include height index, weight index, shoulder width index, waist circumference index, chest circumference index, arm length index, leg length index, etc., as Figure 2 shown. Before the step 21, a body shape database is also established. The body shape database stores the corresponding relationships between multiple body shape indexes such as height index, shoulder width index, waist circumference index, arm length index, leg length index, etc. and multiple corresponding bone feature points. In addition, the weight index is defined as a pixel correlation index, which has a high correlation with the number of pixels of the model picture of the human body model. The body shape database also stores the corresponding relationship between the weight index and the number of pixels of the model picture of the human body model.

[0053] The correction of the bone feature points of the standard human body model in the step 21 is carried out through the following steps:

[0054] Step 211: Obtain and call the corresponding relationship between the body shape indexes included in the body shape information and the corresponding bone feature points in the body shape database according to the body shape information, and calculate the positions of the target bone feature points corresponding to the body shape information.

[0055] Step 212: Adjust the bone feature points of the standard human body model according to the target bone feature points to obtain an intermediate human body model.

[0056] In an embodiment of the present invention, a body type database is established in advance, so that the corresponding relationship between body type indexes and skeletal feature points can be called according to the body type information of the user, and the target skeletal feature points can be calculated. According to the target skeletal feature points, the skeletal feature points of the standard human model can be adjusted to obtain an intermediate human model that highly coincides with the user's bones.

[0057] Correspondingly, the correction of the outer contour feature points of the standard human model in step 21 is performed through the following steps:

[0058] Step 213: Calculate the number of pixels N1 of the intermediate human model.

[0059] Step 214: Call the corresponding relationship between the weight index and the number of pixels of the model picture in the body type database according to the weight index, and calculate the target number of pixels N2 corresponding to the weight index.

[0060] Step 215: According to the difference between the number of pixels N1 of the intermediate human model and the target number of pixels N2, and taking the waist circumference index and the chest circumference index as fitting objects, use an image deformation algorithm to adjust the outer contour feature points of the intermediate human model to obtain the corrected human model. Among them, the image deformation algorithm can be the MLSR deformation algorithm.

[0061] In an embodiment of the present invention, a body type database including the corresponding relationship between the weight index and the number of pixels is established in advance. According to the body type information of the user, the corresponding relationship between the weight index and the number of pixels is called to calculate the target number of pixels. Subsequently, according to the difference in the number of pixels, taking body type indexes such as the chest circumference index, the waist circumference index, and the hip circumference index as fitting objects, an image deformation algorithm is used to adjust the outer contour feature points, which can well eliminate the difference in the number of pixels that the intermediate human model and the model picture corresponding to the user's body type should have (i.e., the target number of pixels), and obtain a corrected human model that highly coincides with the user's outer contour features. It can be understood that the embodiment of the present invention fully considers the parts that have a greater impact on the user's outer contour features, and uses the different numbers of pixels reflected in the model picture corresponding to the user's body type as the target value, and takes the three-dimensional measurement indexes as fitting objects to adjust the outer contour feature points of the human model, so as to accurately and conform to the user's body type to adjust the outer contour feature points of the human model. The obtained corrected human model has a high degree of coincidence with the user's outer contour features and a good matching effect.

[0062] Further, step 22 is specifically:

[0063] According to the difference between the skeletal feature points of the intermediate human model and the standard human model, adjust the skeletal feature points of the standard clothing model to obtain an intermediate clothing model.

[0064] Adjust the outer contour feature points of the intermediate clothing model according to the difference between the outer contour feature points of the modified human body model and the intermediate human body model, so as to obtain the modified clothing model.

[0065] In the embodiment of the present invention, the correction results of the standard human body model at the bone feature points and the outer contour feature points are used to respectively map the correction of the standard clothing model at the bone feature points and the outer contour feature points, which simplifies the computational complexity of the correction process of the standard clothing model and improves the processing efficiency.

[0066] The implementation method of establishing the body type database is introduced in detail below, which includes:

[0067] Step A: Take sample photos of multiple sample models with different body type information. Among them, the shooting method in Step A can refer to the shooting method shown in Figure 1 the shown shooting method.

[0068] Step B1: Use the AlphaPose system to train the bone feature points of each sample photo to obtain the bone feature points of each sample photo, as shown in Figure 3 the shown.

[0069] For example, the bone feature points of the human body model are numbered in sequence as: 1. right shoulder, 2. right elbow, 3. right wrist, 4. left shoulder, 5. left elbow, 6. left wrist, 7. right hip, 8. right knee, 9. right ankle, 10. left hip, 11. left knee, 12. left ankle, 13. top of the head, 14. neck.

[0070] After training, the bone feature point data is returned as follows:

[0071]

[0072] Step C1: Establish a re - regression model function of each body type index and bone feature points according to the body type information of each sample model and the bone feature points of the corresponding sample photo.

[0073] For example, (1) Use the coordinate data of 1. right shoulder (X1, Y1) and 4. left shoulder (X4, Y4) in the bone feature point data and the shoulder width data J of the corresponding sample model to establish a shoulder width re - regression model function: J = f(X4 - X1);

[0074] (2) Use the coordinate data of 13. top of the head (X13, Y13), 9. right ankle (X9, Y9), and 12. left ankle (X12, Y12) in the bone feature point data and the height data H of the corresponding sample model to establish a height re - regression model function: H = f[Y13 - (Y9 + Y12) / 2];

[0075] (3) Use the coordinate data of 7. right hip (X7, Y7) and 10. left hip (X10, Y10) in the bone feature point data to establish a waist circumference multiple regression model function with the waist circumference data Y of the corresponding sample model: Y = f(X7 - X10);

[0076] (4) Use the coordinate data of 1. right shoulder (X1, Y1), 2. right elbow (X2, Y2), 3. right wrist (X3, Y3), 4. left shoulder (X4, Y4), 5. left elbow (X5, Y5), and 6. left wrist (X6, Y6) in the bone feature point data to establish an arm length multiple regression model function with the arm length data B of the corresponding sample model: B = [f(X1, Y1, X2, Y2) + f(X2, Y2, X3, Y3) + f(X4, Y4, X5, Y5) + f(X5, Y5, X6, Y6)];

[0077] (5) Use the coordinate data of 7. right hip (X7, Y7), 8. right knee (X8, Y8), 9. right ankle (X9, Y9), 10. left hip (X10, Y10), 11. left knee (X11, Y11), and 12. left ankle (X12, Y12) in the bone feature point data to establish a leg length multiple regression model function with the leg length data T of the corresponding sample model: T = [f(X7, Y7, X8, Y8) + f(X8, Y8, X9, Y9) + f(X10, Y10, X11, Y11) + f(X11, Y11, X12, Y12)].

[0078] Step B2: Use a deep learning algorithm to train a semantic segmentation model for each sample photo, and obtain and calculate the number of pixels N of the semantic segmentation image of each sample photo. Among them, the deep learning algorithm can be a convolutional neural network method.

[0079] Step C2: According to the weight index in the body type information of each sample model and the number of pixels N of the semantic segmentation image of the corresponding sample photo, establish a multiple regression model function between the weight index and the number of pixels: G = f(N).

[0080] In an embodiment of the present invention, bone feature points of sample photos are obtained by using artificial intelligence training for a large number of sample photos corresponding to different body type information, and a re - regression function model between the obtained bone feature points and the body type information is established, so as to facilitate calling the re - regression function model according to the user's body type information and calculating the bone feature points corresponding to the user's body type information. In addition, in an embodiment of the present invention, a semantic segmentation model is trained by using a deep - learning algorithm for a large number of sample photos corresponding to different body type information to obtain a semantic segmentation image, and a re - regression function model between the number of pixels of the obtained semantic segmentation image and the weight index is used, so as to facilitate calling the re - regression function model according to the user's body type information and calculating the target number of pixels corresponding to the user's body type information, and further used to adjust the outer contour feature points of the standard clothing model.

[0081] In an extensible manner, in the embodiment of the present invention, the method further includes cutting out the face part of the user's real photo and using it to replace and correct the corresponding part of the human body model, so that the virtual try - on effect of the present invention is more realistic and interesting. Specifically, it includes the following steps:

[0082] Obtain the user's real photo.

[0083] Use the area corresponding to the user's face in the user's real photo to replace the area corresponding to the face of the standard human body in the model picture of the corrected human body model to obtain a corrected human body model after face correction.

[0084] In step 3, cover the model picture of the corrected clothing model on the model picture of the corrected human body model after face correction to show the try - on effect to the user.

[0085] Furthermore, an embodiment of the present invention also provides a clothing try - on device, including a processor, and the processor is used to execute the clothing try - on method of the embodiment of the present invention, which inherits all the advantages of the above - mentioned method.

[0086] The above description of the specification and embodiments is used to explain the protection scope of the present invention, but does not constitute a limitation to the protection scope of the present invention. Through the inspiration of the present invention or the above - mentioned embodiments, those of ordinary skill in the art, combined with common general knowledge, general technical knowledge in the field and / or existing technologies, through logical analysis, reasoning or limited experiments, can obtain modifications, equivalent replacements or other improvements to the embodiments of the present invention or some of its technical features, which should all be included within the protection scope of the present invention.

Claims

1. A method for trying on clothing, characterized in that, Including: Step 1: Construct a standard human body model and standard clothing models that match the standard human body model and correspond to each piece of clothing; Both the standard human body model and the standard clothing models include model pictures marked with bone feature points and outer contour feature points; construct a body type database that stores the correspondence between body type indicators and bone feature points, and the correspondence between pixel association indicators and the number of pixels of the model pictures of the human body models; Step 2: Obtain the body type information of the user and the clothing selected by the user, and correct the bone feature points and outer contour feature points of the standard human body model and the standard clothing model corresponding to the clothing selected by the user to obtain a corrected human body model that matches the body type information and a corrected clothing model that matches the corrected human body model; Among them, the correction of the outer contour feature points of the standard human body model includes: calling the mapping relationship between the pixel association indicator and the number of pixels in the body type database according to the pixel association indicator in the user's body type information, calculating the target number of pixels, and combining at least one of the chest circumference indicator, waist circumference indicator, and hip circumference indicator through the difference between the target number of pixels and the number of pixels of the intermediate human body model, and dynamically adjusting the outer contour feature points of the intermediate human body model by using an image deformation algorithm; among them, the intermediate human body model is a human body model obtained by adjusting the bone feature points of the standard human body model according to the target bone feature points; the target bone feature points are calculated according to the correspondence between the body type indicator and the bone feature points; Step 3: Cover the model picture of the corrected clothing model on the model picture of the corrected human body model and display the try-on effect to the user.

2. The clothing try-on method according to claim 1, characterized in that The standard human body model and the standard clothing models in Step 1 are constructed through the following steps: Step 11: Take pictures of a standard model and pictures of each piece of clothing worn on the standard model respectively; Step 12: Take the standard model and each piece of clothing as the main objects, perform background removal on the pictures obtained by shooting, and obtain model pictures with the same size; Step 13: Mark the pixel points corresponding to the bones of the standard model on the model picture as bone feature points, and mark the pixel points corresponding to the outer contour of the corresponding main object on the model picture as outer contour feature points.

3. The clothing try-on method according to claim 2, characterized in that, The specific content of Step 2 is: Step 21: Obtain and correct the bone feature points and outer contour feature points of the standard human body model according to the body type information of the user to obtain the corrected human body model; Step 22: Calculate and respectively correct the bone feature points and outer contour feature points of the standard clothing model corresponding to the clothing according to the differences between the corrected human body model and the standard human body model in terms of bone feature points and outer contour feature points after obtaining the clothing selected by the user to obtain the corrected clothing model.

4. The clothing try-on method according to claim 3, characterized in that: The body type information includes multiple body type indicators; The correction of the bone feature points of the standard human body model in Step 21 is carried out through the following steps: Step 211: Obtain and, based on the body shape information, call in the body shape database the correspondence between the body shape indicators included in the body shape information and the corresponding skeletal feature points, and calculate the positions of the target skeletal feature points corresponding to the body shape information; Step 212: Adjust the skeletal feature points of the standard human model according to the target skeletal feature points to obtain an intermediate human model.

5. The clothing try-on method according to claim 4, wherein, The body shape database is constructed through the following steps: Step A: Take sample photos of multiple sample models with different body shape information; Step B1: Use the AlphaPose system to train the skeletal feature points for each sample photo to obtain the skeletal feature points of each sample photo; Step C1: Based on the body shape information of each sample model and the skeletal feature points of the corresponding sample photo, establish a re - regression model function between each body shape indicator and the skeletal feature points.

6. The clothing try - on method according to claim 4, wherein: The multiple body shape indicators include at least one pixel - related indicator; The correction of the outer contour feature points of the standard human model in step 21 is carried out through the following steps: Step 213: Calculate the number of pixels N1 of the intermediate human model; Step 214: Based on the pixel - related indicator of the body shape information, call in the body shape database the correspondence between the pixel - related indicator and the number of pixels of the model picture, and calculate the target number of pixels N2 corresponding to the pixel - related indicator; Step 215: Based on the difference between the number of pixels N1 of the intermediate human model and the target number of pixels N2, and taking other body shape indicators as the fitting object, use an image deformation algorithm to adjust the outer contour feature points of the intermediate human model to obtain the corrected human model; wherein, the other body shape indicators include at least one of the chest circumference indicator, waist circumference indicator, and hip circumference indicator.

7. The clothing try-on method according to claim 6, wherein The body shape database is constructed through the following steps: Step A: Take sample photos of multiple sample models with different body shape information; Step B2: Use a deep - learning algorithm to train the semantic segmentation model for each sample photo, and obtain and calculate the number of pixels of the semantic segmentation image of each sample photo; Step C2: Based on the pixel - related indicator in the body shape information of each sample model and the number of pixels of the semantic segmentation image of the corresponding sample photo, establish a re - regression model function between the pixel - related indicator and the number of pixels.

8. The clothing try-on method according to claim 6, characterized in that, Step 22 is specifically: Based on the difference between the skeletal feature points of the intermediate human model and the standard human model, adjust the skeletal feature points of the standard clothing model to obtain an intermediate clothing model; Based on the difference between the outer contour feature points of the corrected human model and the intermediate human model, adjust the outer contour feature points of the intermediate clothing model to obtain the corrected clothing model.

9. The clothing try-on method according to claim 1, characterized in that, The method further includes: Obtain the user's real photo; Use the area corresponding to the user's face in the user's real photo to replace the area corresponding to the face of the standard human in the model picture of the corrected human model to obtain a corrected human model with face correction; Step 3 is to overlay the model picture of the corrected clothing model on the model picture of the corrected human body model after face correction, and display the try-on effect to the user.

10. A clothing try-on device, characterized in that, Comprising: A processor; The processor is used to execute the clothing try-on method according to any one of claims 1-9.

Images