Neural network based method of fitting a model to a garment

Abstract

The embodiment of the specification provides a model and clothes wearing method based on a neural network, which comprises receiving an initial model image of a target model and a clothes image of a target clothes; inputting the initial model image and the clothes image into a pre-trained neural network model to obtain a target model image of the target model wearing the target clothes; and returning the target model image. The method is applied to the field of virtual reality or augmented reality, and can adaptively wear the target clothes on the target model through the pre-neural neural network model, without the need for manual intervention to adjust the size of the target clothes in the process of wearing the target clothes on the target model, thereby solving the problem that low-cost mass production of model clothes wearing cannot be realized in a short time due to manual intervention in model clothes wearing.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of computer technology, and in particular to two methods for wearing mannequins and clothing based on neural networks. Background Technology

[0002] With the popularization and development of the internet, e-commerce has permeated personal life, and the concept of online shopping has gradually taken root in people's hearts. Currently, online shopping typically involves merchants providing users with references by showing them photos of their products. For example, when the product is clothing, after a garment is designed, it needs to be photographed on a mannequin before the photos are shown to users for purchase reference. Specifically, this involves first selecting a specific mannequin, then manually putting the garment on the mannequin for a photo shoot. If a different mannequin is used, manual intervention is required to adjust the size of the garment before it is put on the mannequin for another photo shoot. Because each garment requires manual adjustment, merchants cannot achieve low-cost, large-scale, and short-term photography of garments on mannequins for display. Summary of the Invention

[0003] In view of this, embodiments of this specification provide two methods for wearing mannequins and clothing based on neural networks. One or more embodiments of this specification also relate to a mannequin and clothing wearing device based on neural networks, a virtual reality device or augmented reality device, a computing device, a computer-readable storage medium, and a computer program, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this specification, a method for wearing clothing on a model based on a neural network is provided, comprising:

[0005] Receive the initial model image of the target model and the clothing image of the target garment;

[0006] The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing;

[0007] Return the image of the target model.

[0008] The neural network model is obtained by training using a semi-supervised method based on labeled and unlabeled training samples.

[0009] According to a second aspect of the embodiments of this specification, a wearable device for a mannequin and clothing based on a neural network is provided, comprising:

[0010] The image receiving module is configured to receive an initial model image of the target model and a clothing image of the target garment;

[0011] The image acquisition module is configured to input the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing;

[0012] The image display module is configured to return the image of the target model.

[0013] The neural network model is obtained by training using a semi-supervised method based on labeled and unlabeled training samples.

[0014] According to a third aspect of the embodiments of this specification, a method for wearing clothing on a model based on a neural network is provided, comprising:

[0015] Display an image input interface to the user based on the user's request;

[0016] Receive the initial model image of the target model and the clothing image of the target garment input by the user through the image input interface;

[0017] The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing;

[0018] The target model image is displayed to the user through the image input interface.

[0019] The neural network model is obtained by training using a semi-supervised method based on labeled and unlabeled training samples.

[0020] According to a fourth aspect of the embodiments of this specification, a virtual reality device or an augmented reality device is provided, comprising:

[0021] Memory and processor;

[0022] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, perform the following steps:

[0023] Receive the initial model image of the target model and the clothing image of the target garment;

[0024] The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing;

[0025] Return the target model image and render it onto the display interface of the virtual reality device or the display of the augmented reality device.

[0026] According to a fifth aspect of the embodiments of this specification, a computing device is provided, comprising:

[0027] Memory and processor;

[0028] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the above-described steps of the model and clothing wearing method based on neural networks are implemented.

[0029] According to a sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores computer-executable instructions, which, when executed by a processor, implement the steps of the above-described neural network-based method for wearing clothing on a model.

[0030] According to a seventh aspect of the embodiments of this specification, a computer program is provided, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described neural network-based method for wearing clothing on a model.

[0031] One embodiment of this specification implements a method for modeling clothing based on a neural network. The method includes receiving an initial model image of a target model and a clothing image of a target garment; inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target garment; and returning the target model image. The neural network model is trained using a semi-supervised method based on labeled and unlabeled training samples.

[0032] Specifically, this neural network-based method for dressing mannequins and clothing can adaptively dress target clothing on target mannequins using a pre-built neural network model. This eliminates the need for manual intervention to adjust the size of the clothing during the dressing process, thus solving the problem of not being able to achieve low-cost, large-scale, and short-term clothing dressing of mannequins when manually dressed. Furthermore, it can directly output images of the target mannequins wearing the clothing, eliminating the need to photograph the mannequins and further improving the user experience. Attached Figure Description

[0033] Figure 1 This is a schematic diagram illustrating a specific application scenario of a neural network-based method for wearing clothing, as provided in one embodiment of this specification.

[0034] Figure 2 This is a flowchart illustrating a method for wearing clothing on a model based on a neural network, as provided in one embodiment of this specification.

[0035] Figure 3 This is a schematic diagram illustrating the processing steps of a neural network-based method for wearing clothing, as provided in one embodiment of this specification.

[0036] Figure 4 This is a schematic diagram of the structure of a neural network-based mannequin and clothing wearing device provided in one embodiment of this specification;

[0037] Figure 5 This is a flowchart illustrating another method for wearing clothing on a model based on a neural network, as provided in one embodiment of this specification.

[0038] Figure 6 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0039] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0040] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0041] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0042] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0043] 3D: 3D is short for 3 Dimensions. In Chinese, it refers to three dimensions, three coordinates, that is, it has length, width and height, and is visually three-dimensional. 3D is the concept of space, that is, space composed of three axes: X, Y and Z, which is relative to a plane (2D) that only has length and width.

[0044] CNN: Convolutional Neural Network, a structure that can learn from data.

[0045] XR: Extended Reality. XR is a new concept referring to an interactive environment that combines real and virtual elements, created through computer technology and wearable devices. Extended Reality includes various forms such as Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR). In other words, to avoid confusion, XR is a general term encompassing AR, VR, and MR. XR is divided into multiple levels, ranging from virtual worlds with limited sensor input to fully immersive virtual worlds.

[0046] Semi-supervised learning is a learning method that combines supervised and unsupervised learning. It uses a large amount of unlabeled data, along with labeled data, to perform pattern recognition tasks. When using semi-supervised learning, it requires as few people as possible to perform the work, while still achieving relatively high accuracy.

[0047] This specification provides two methods for wearing mannequins and clothing based on neural networks. One or more embodiments of this specification also relate to a mannequin and clothing wearing device based on neural networks, a virtual reality device or augmented reality device, a computing device, a computer-readable storage medium, and a computer program, which are described in detail in the following embodiments.

[0048] See Figure 1 , Figure 1 The illustration shows a specific application scenario of a neural network-based method for wearing clothing, according to one embodiment of this specification.

[0049] Figure 1 The system includes a terminal 102 and a server 104. The terminal 102 includes, but is not limited to, mobile phones, tablets, desktop computers, etc. The server 104 can be understood as a physical server or a cloud server. For ease of understanding, this specification describes the embodiments using a cloud server as an example for detailed explanation.

[0050] The following is a detailed explanation of the application of the neural network-based model and clothing wearing method provided in the embodiments of this specification to a merchant's clothing display scenario.

[0051] In practice, the terminal 102 uploads the image of the mannequin to be worn and the image of the clothing to be worn on the mannequin for display to the server 104 according to the mannequin clothing wearing instruction triggered by the merchant. The mannequin image and the clothing image can be taken by the merchant through the camera embedded in the terminal 102, or they can be taken by the merchant through an external camera and uploaded to the terminal 102. Of course, they can also be uploaded by the merchant through other storage devices (such as hard drives) to the terminal 102.

[0052] Server 104 inputs the mannequin image and the clothing image into a pre-trained convolutional neural network. Through processing by the convolutional neural network, a final mannequin image suitable for clothing display is obtained, and this image is returned to terminal 102 for display on the terminal's interface. Alternatively, this mannequin image can be used in the merchant's virtual store creation project.

[0053] The method for wearing clothing on a mannequin based on a neural network provided in this specification can adaptively wear clothing on mannequins of completely different heights, weights, and builds based on a pre-trained convolutional neural network. The convolutional neural network achieves the matching of clothing size with the mannequin size. This fully automatic method completes the wearing of clothing on mannequins, saving costs and improving the efficiency of wearing clothing on mannequins.

[0054] See Figure 2 , Figure 2 A flowchart is shown of a method for wearing clothing on a model based on a neural network according to an embodiment of this specification, which specifically includes the following steps.

[0055] Step 202: Receive the initial model image of the target model and the clothing image of the target garment.

[0056] Specifically, the neural network-based model and clothing wearing method provided in this specification can be applied to XR extended reality e-commerce projects. For example, after a merchant builds a virtual store in an XR extended reality e-commerce project, they can display clothing for sale worn by a mannequin in the virtual store for buyers to reference. Of course, this neural network-based model and clothing wearing method can also be applied to ordinary e-commerce scenarios. For example, when a merchant joins a shopping platform, they can use this method to generate multiple clothing display images of clothing for sale worn by a mannequin, and use these images to list and display clothing in their store on the shopping platform for buyers browsing the platform to reference. Therefore, the specific application scenarios of the neural network-based model and clothing wearing method provided in this specification can be set according to actual needs, and this specification does not impose any limitations on them.

[0057] The initial model image of the target model and the clothing image of the target garment can be uploaded by the user or received from other devices that need to acquire images of the target model wearing the target garment. For ease of understanding, the initial model image of the target model and the clothing image of the target garment will be used as an example to illustrate the process of receiving the initial model image of the target model and the clothing image of the target garment sent by the user.

[0058] The user in the embodiments of this specification can be understood as the merchant in the above embodiments, or other users who want to obtain images of target models wearing target clothing.

[0059] The target model can be understood as the aforementioned mannequin, i.e., a human model. It can be a real person or a mannequin, etc., and the target model does not have a specific height or weight. The initial model image can be understood as an image containing the target model. Taking a real person as an example, if the target model is model 'a', the initial model image is an image containing model 'a', that is, an image taken of model 'a'. The target clothing can be understood as clothing of any style, size, and color to be matched with the target model (i.e., clothing to be worn by the target model). The clothing image can be understood as an image containing the target clothing, that is, an image taken of the target clothing.

[0060] In practical applications, the initial model image and the clothing image of the target model can be obtained by the user through a professional camera, meaning these images are noise-free (clean background, no clutter, etc.); or they can be obtained by randomly shooting the target model and clothing, meaning these images contain noise (complex background, clutter, etc.). If the initial model image and clothing image are noise-free, they can be input into a pre-trained neural network model for processing. If they contain noise, they can be processed first before being input into the pre-trained neural network model to ensure the accuracy of the model's output.

[0061] Taking the user as the merchant as an example, specifically, receiving the initial model image of the target model and the clothing image of the target garment can be understood as receiving the initial model image of the target model and the clothing image of the target garment sent by the merchant. For example, receiving the initial model image of model 'a' and the clothing image of a dress with model number 'cc' sent by the merchant.

[0062] Step 204: Input the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of the target model wearing the target clothing.

[0063] The neural network model is obtained by training using a semi-supervised method based on labeled and unlabeled training samples.

[0064] Specifically, the neural network model provided in the embodiments of this specification can be understood as a convolutional neural network model. During training, the neural network model provided in the embodiments of this specification can be trained using a portion of labeled training samples and a portion of unlabeled training samples through a semi-supervised learning method. Specifically, the neural network model is trained through the following steps:

[0065] Obtain images of at least two sets of sample clothing and images of at least two sample models;

[0066] Based on the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models, the labeled training samples and the unlabeled training samples are determined, wherein the number of unlabeled training samples is greater than the number of labeled training samples.

[0067] The neural network model is trained using the labeled training samples and the unlabeled training samples.

[0068] Among them, sample clothing can be understood as clothing of any style, color and size; sample models can be understood as human models or mannequins of any height, weight and build.

[0069] Specifically, obtain sample clothing images of at least two sets of sample garments and sample model images of at least two sample models; this can be understood as obtaining multiple sets of garments of various styles as sample garments and determining the sample garment image corresponding to each set of sample garments, and obtaining multiple human models of various body types as sample models and determining the sample model image corresponding to each sample model.

[0070] Then, based on the sample clothing images of multiple sets of sample clothing and the sample model images of multiple sample models, labeled training samples and unlabeled training samples are determined; finally, the neural network model is trained based on the labeled training samples and the unlabeled training samples.

[0071] In practical applications, since labeled training samples require manual annotation, which incurs labor costs, a small number of labeled training samples and a large number of unlabeled training samples are obtained when determining the training samples. By using a small number of labeled training samples and a large number of unlabeled training samples to train the neural network model, the labor costs can be reduced while ensuring the accuracy of model training.

[0072] The neural network model training steps provided in this specification, after obtaining sample clothing images of multiple sets of sample clothing and sample model images of multiple sample models, determine a small number of labeled training samples and a large number of unlabeled training samples based on the sample clothing images of multiple sets of sample clothing and sample model images of multiple sample models. Through this semi-supervised learning method, relatively high accuracy can be achieved while minimizing manual costs.

[0073] In practice, after determining the images of multiple sets of sample clothing and multiple sample models, labeled training samples can be identified using a preset sample processing method. Once the labeled training samples are determined, unlabeled training samples can be quickly identified based on them. This means that both labeled and unlabeled training samples can be acquired in a single data collection, improving training sample acquisition efficiency and reducing the training time of the neural network model. The specific implementation method is as follows:

[0074] The step of determining the labeled training samples and the unlabeled training samples based on the images of the at least two sets of sample clothing and the images of the at least two sample models includes:

[0075] According to a preset sample processing method, the labeled training samples are determined from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models.

[0076] Based on the labeled training samples, the unlabeled training samples are determined from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models.

[0077] The preset sample processing method can be understood as a physical simulation method.

[0078] Specifically, after determining the sample clothing images of multiple sets of sample clothing and the sample model images of multiple sample models, a physical simulation method can be used to match certain sample clothing with certain sample models from the sample clothing images of multiple sets of sample clothing and the sample model images of multiple sample models. The sample clothing images and sample model images that match the sample clothing with the sample models are used as labeled training samples.

[0079] After identifying the labeled training samples, all remaining images of the sample clothing and sample models from multiple sets of sample clothing and multiple sample models are used as unlabeled training samples. In this way, two types of training samples can be obtained after one data collection, without the need to collect labeled and unlabeled training samples separately, thus improving the efficiency of training sample collection.

[0080] In practical applications, labeled training samples include training samples and their corresponding labels. The specific implementation method for determining training samples and their corresponding labels from multiple sets of sample clothing images and multiple sample model images is as follows:

[0081] The step of determining the labeled training samples from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models according to a preset sample processing method includes:

[0082] Select the sample clothing image and the sample model image to be matched from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models;

[0083] According to the preset sample processing method, the image of the clothing to be matched is matched with the image of the model to be matched to obtain the target sample model image of the model wearing the clothing to be matched.

[0084] The clothing image and the model image to be matched are determined as training samples, and the model image of the target sample is determined as the sample label of the training samples.

[0085] The labeled training samples are determined based on the training samples and the sample labels.

[0086] In practice, the process begins by selecting a sample clothing image and a sample model image to be matched from multiple sets of sample clothing images and multiple sample model images. The sample clothing in the sample clothing image to be matched must match one, two, or more sample models in the sample model image to be matched. In other words, the sample clothing in the sample clothing image to be matched can be appropriately worn on one, two, or more sample models in the sample model image to be matched.

[0087] Then, the image of the sample clothing to be matched is matched with the image of the sample model to be matched using a physical simulation method to obtain a target sample model image of the sample model wearing the sample clothing to be matched. For example, by using a physical simulation method, the sample clothing in the image of the sample clothing to be matched is worn on a suitable sample model in the image of the sample model to be matched, thereby obtaining a target sample model image of the sample model wearing the suitable sample clothing to be matched. That is, the target sample model image contains the sample model wearing the sample clothing to be matched.

[0088] Finally, the matching sample clothing image and the matching sample model image are determined as training samples. Through physical simulation, the matching sample clothing image and the matching sample model image are matched, and the resulting target sample model image is used as the sample label. The training sample and the sample label corresponding to the training sample constitute the labeled training sample.

[0089] In the embodiments of this specification, after obtaining sample clothing images of multiple sets of sample clothing and sample model images of multiple sample models, certain sample clothing and certain sample models can be matched by physical simulation. That is, by making reasonable wearing of the clothing in terms of size and dimensions, labeled training samples can be quickly and accurately constructed to ensure the accuracy of subsequent neural network model training.

[0090] After obtaining labeled and unlabeled training samples, to further ensure the training accuracy of the neural network model, an appropriate number of labeled and unlabeled training samples can be selected according to a preset ratio. This ensures the training accuracy of the neural network model while avoiding the high manual cost caused by excessive acquisition of labeled training samples through physical simulation methods. The specific implementation method is as follows:

[0091] The step of training the neural network model based on the labeled training samples and the unlabeled training samples includes:

[0092] The labeled training samples and the unlabeled training samples are determined according to a preset ratio.

[0093] The labeled training samples are input into the neural network model to obtain a first output label, and the unlabeled training samples are input into the neural network model to obtain a second output label;

[0094] The neural network model is trained based on the first output label and the second output label.

[0095] The preset ratio can be set according to the actual application. This specification does not limit this in any way. For example, the preset ratio can be 10% for labeled training samples and 90% for unlabeled training samples.

[0096] A detailed explanation will be given using a preset ratio of 10% labeled training samples and 90% unlabeled training samples as an example.

[0097] Specifically, 10% of the training samples are labeled and 90% are unlabeled. When the input training samples are labeled, the labeled training samples are input into the neural network model to obtain the first output label. When the input training samples are unlabeled, the unlabeled training samples are input into the neural network model to obtain the second output label. The neural network model is trained based on the first output label and the second output label.

[0098] In specific implementation, training the neural network model based on the first output label and the second output label includes:

[0099] The loss function is calculated based on the sample labels of the labeled training samples and the first output label, and the neural network model is adjusted based on the loss function.

[0100] Based on the confidence level of the second output label, a target training sample is determined from the unlabeled training samples. The target training sample and the second output label corresponding to the target training sample are added to the labeled training samples, so as to train the neural network model based on the updated labeled training samples.

[0101] The neural network model is obtained when the model training termination condition is met.

[0102] Among them, the conditions for ending model training include, but are not limited to, the loss function of the neural network model having met the preset loss threshold, or the number of iterations of the neural network model having met the preset number threshold, etc.

[0103] In practical applications, when training a neural network model using both unlabeled and labeled training samples, if the input training samples are labeled, the model outputs a first output label, such as an image of a model wearing arbitrary clothing. This first output label is then compared with the original labels of the labeled training samples to calculate a loss function and return gradient information, thus adjusting the neural network model's training. If the input training samples are unlabeled, the model outputs a second output label, also such as an image of a model wearing arbitrary clothing. This second output label can then be used as a pseudo-label for the unlabeled training samples. Finally, based on the confidence level of this pseudo-label, target training samples are determined from the unlabeled training samples. For example, the top 5 unlabeled training samples with the highest pseudo-label confidence levels are selected as target training samples. These target training samples and their corresponding pseudo-labels form labeled training samples, which are then added to the original labeled training samples. These labeled training samples are then used as labeled training samples in the next neural network model training iteration.

[0104] When the loss function of the neural network model meets the preset loss threshold or the number of iterations of the neural network model has met the preset number threshold (such as 20,000 times), it is determined that the neural network model meets the model training termination condition. At this time, the neural network model can be understood as a pre-trained neural network model, which can be used for subsequent applications.

[0105] In the embodiments of this specification, the labeled training samples can be expanded according to the confidence level of the neural network model for unlabeled training samples, thereby training the neural network model with more labeled training samples and further improving the accuracy of the neural network model.

[0106] After the neural network model is pre-trained, the initial model image and clothing image sent by the user can be input into the neural network model to obtain the target model image of the target model wearing the target clothing. That is, the target model image includes the target model wearing the target clothing.

[0107] In the embodiments of this specification, the pre-trained neural network model can adaptively fit clothing onto models of matching or mismatched sizes. By adjusting the size, dimensions, and matching relationships within the neural network model, an image of a target model wearing the target clothing can be obtained without the user noticing or needing to do anything extra.

[0108] Furthermore, the neural network model provided in this specification can not only adaptively wear the target clothing on the target model when the user provides a target model and a set of target clothing; it can also adaptively wear the set of target clothing on each target model when the user provides multiple target models and a set of target clothing; it can also adaptively wear each set of target clothing on the target model when the user provides a target model and multiple sets of target clothing; and it can also adaptively wear each set of target clothing on each target model when the user provides multiple target models and multiple sets of target clothing.

[0109] Specifically, when the user provides multiple target models and a set of target clothing, the specific implementation method for obtaining the target model image is as follows:

[0110] The receiving of the initial model image of the target model and the clothing image of the target garment includes:

[0111] Receive initial model images of at least two target models and an image of a set of target clothing sent by a user, wherein the at least two target models have different sizes;

[0112] Accordingly, the step of inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing includes:

[0113] The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of at least two target models wearing the target clothing.

[0114] The following is a detailed introduction using the example of three target models: Target Model 1, Target Model 2, and Target Model 3, with at least two target models.

[0115] In practice, the system receives initial model images of target model 1, target model 2, target model 3, and a set of target clothing sent by the user. The system simultaneously inputs these initial model images into a pre-trained neural network model to obtain target model images of target model 1, target model 2, target model 3, and a set of target clothing.

[0116] In the embodiments of this specification, when the user provides multiple target models and a set of target clothing, the pre-trained neural network model can adaptively dress the set of target clothing on each target model, thereby obtaining images of target model 1, target model 2, and target model 3 wearing the target clothing, thus meeting the user's personalized needs.

[0117] Therefore, when a user provides a target model and multiple sets of target clothing, the specific implementation method for obtaining the target model's image is as follows:

[0118] The receiving of the initial model image of the target model and the clothing image of the target garment includes:

[0119] Receive an initial model image of a target model and clothing images of at least two sets of target clothing sent by a user, wherein the clothing sizes of the at least two sets of target clothing are different;

[0120] Accordingly, the step of inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing includes:

[0121] The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of a target model wearing at least two sets of target clothing.

[0122] The following is a detailed introduction using the example of three target outfits (at least two target outfits): target outfit a, target outfit b, and target outfit c.

[0123] In practice, the system receives an initial model image of a target model, as well as clothing images of target garment a, target garment b, and target garment c, sent by the user. Simultaneously, the initial model image of a target model, along with clothing images of target garment a, target garment b, and target garment c, are input into a pre-trained neural network model to obtain target model images of the target model wearing target garment a, target model wearing target garment b, and target model wearing target garment c.

[0124] In the embodiments of this specification, when a user provides a target model and multiple sets of target clothing, a pre-trained neural network model can adaptively dress the multiple sets of target clothing on a target model, thereby obtaining images of the target model wearing target clothing a, target clothing b, and target clothing c, respectively. This allows for obtaining images of a single model wearing three sets of clothing, thus meeting the user's personalized needs.

[0125] When the user provides multiple target models and multiple sets of target clothing, the specific implementation method for obtaining the target model images is as follows:

[0126] The receiving of the initial model image of the target model and the clothing image of the target garment includes:

[0127] Receive initial model images of at least two target models and clothing images of at least two sets of target clothing sent by a user, wherein the at least two target models have different model sizes and the at least two sets of target clothing have different clothing sizes;

[0128] Accordingly, the step of inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing includes:

[0129] The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of at least two target models wearing at least two sets of target clothing.

[0130] Combining the two embodiments above, a detailed introduction will be given using the example of at least two target models as three target models: target model 1, target model 2, and target model 3; and at least two sets of target clothing as three sets of target clothing: target clothing a, target clothing b, and target clothing c.

[0131] In specific implementation, the system receives initial model images of target model 1, target model 2, target model 3, and target clothing a, target clothing b, and target clothing c sent by the user. Simultaneously, the initial model images of target model 1, target model 2, target model 3, target clothing a, target clothing b, and target clothing c are input into a pre-trained neural network model. This allows the system to obtain target model images of target model 1 wearing target clothing a, target model 1 wearing target clothing b, target model 1 wearing target clothing c, target model 2 wearing target clothing a, target model 2 wearing target clothing b, target model 2 wearing target clothing c, target model 3 wearing target clothing a, target model 3 wearing target clothing b, and target model 3 wearing target clothing c.

[0132] In the embodiments of this specification, when the user provides multiple target models and multiple sets of target clothing, a pre-trained neural network model can adaptively dress each set of target clothing on each of the multiple target models, thereby obtaining an image of each model wearing three sets of clothing, thus meeting the user's personalized needs.

[0133] Step 206: Return the image of the target model.

[0134] Specifically, after obtaining the target model image through the neural network model, the target model image is returned, such as by returning the target model image to the user.

[0135] In practical applications, to make user operations more convenient and improve the user experience, a visual user interface can be provided. Users can upload initial model images and clothing images of the target model through this interface. These images are then processed by a neural network model to obtain the final target model image, which can then be displayed to the user through the user interface, further enhancing the user experience. The specific implementation method is as follows:

[0136] The receiving of the initial model image of the target model and the clothing image of the target garment includes:

[0137] Receive the initial model image of the target model and the clothing image of the target garment sent by the user through the user interaction interface;

[0138] Accordingly, returning the target model image includes:

[0139] The target model image is returned and displayed to the user through the user interface.

[0140] The neural network-based mannequin and clothing wearing method provided in this specification can adaptively wear the target clothing onto the target mannequin using a pre-built neural network model. This eliminates the need for manual intervention to adjust the size of the target clothing during the wearing process, thus solving the problem of not being able to achieve low-cost, large-scale, and short-term mannequin clothing wearing when manual intervention is required. Furthermore, it can directly output the target mannequin's image wearing the target clothing, eliminating the need to take photos of the target mannequin wearing the target clothing, further improving the user experience.

[0141] The following is in conjunction with the appendix Figure 3 Taking the application of the neural network-based model and clothing wearing method provided in this specification in a human model clothing wearing scenario as an example, the neural network-based model and clothing wearing method will be further explained. Figure 3 The illustration shows a schematic diagram of a neural network-based method for wearing clothing, which includes the following steps, according to one embodiment of this specification.

[0142] Step 1: Obtain 90% of the unlabeled data and 10% of the labeled data.

[0143] Specifically, first, multiple human models of various body types and multiple sets of clothing of various styles and sizes are obtained; through physical simulation, some clothing that can be properly sized and worn is matched with some human models to form labeled data; the remaining human models and clothing are used as unlabeled data; then, from the labeled data and unlabeled data, according to the above proportions, 90% of the unlabeled data and 10% of the labeled data are obtained as model training data.

[0144] Step 2: Input 90% of the unlabeled data and 10% of the labeled data into the convolutional neural network model.

[0145] Step 3: When the input data to the convolutional neural network model is labeled data, the convolutional neural network model calculates the loss function between the output human model dressed in clothing and the human model dressed in clothing generated through physical simulation from the labeled data, and then returns the gradient information; when the input data to the convolutional neural network model is unlabeled data, the convolutional neural network model outputs the human model dressed in clothing and the confidence score of the human model, and then filters according to the confidence score, selecting the top 5 unlabeled data as labeled data for the next training step.

[0146] Through iterative processes described above, training ends when the convolutional neural network model meets the training termination condition. In the subsequent inference process, the human model and clothing can be directly input into the convolutional neural network model, which can directly predict and output the effect of the human model wearing appropriately sized clothing.

[0147] The neural network-based method for model and clothing wearing provided in this specification employs a semi-supervised approach for network learning. This not only avoids the need for large amounts of labeled data required by fully supervised methods, but also allows the final convolutional neural network to outperform unsupervised learning. Furthermore, it utilizes end-to-end convolutional neural network learning (i.e., both model training and inference can be performed on the device), making the entire process fully automated and eliminating the need for manual intervention. This solves the problems of slow and costly manual generation of mannequin clothing wearing effects. Moreover, by adopting a training-then-inference approach, the final inference process is achieved at the millisecond level, greatly improving the efficiency of mannequin clothing wearing.

[0148] Corresponding to the above method embodiments, this specification also provides embodiments of a wearable device for models and clothing based on neural networks. Figure 4 This specification illustrates a schematic diagram of a neural network-based mannequin and clothing wearing device according to one embodiment. Figure 4 As shown, the device includes:

[0149] The image receiving module 402 is configured to receive an initial model image of the target model and a clothing image of the target garment;

[0150] The image acquisition module 404 is configured to input the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing;

[0151] Image display module 406 is configured to return the image of the target model.

[0152] The neural network model is obtained by training using a semi-supervised method based on labeled and unlabeled training samples.

[0153] Optionally, the image receiving module 402 is further configured to:

[0154] Receive initial model images of at least two target models and an image of a set of target clothing sent by a user, wherein the at least two target models have different sizes;

[0155] Accordingly, the image acquisition module 404 is further configured to:

[0156] The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of at least two target models wearing the target clothing.

[0157] Optionally, the image receiving module 402 is further configured to:

[0158] Receive an initial model image of a target model and clothing images of at least two sets of target clothing sent by a user, wherein the clothing sizes of the at least two sets of target clothing are different;

[0159] Accordingly, the image acquisition module 404 is further configured to:

[0160] The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of a target model wearing at least two sets of target clothing.

[0161] Optionally, the image receiving module 402 is further configured to:

[0162] Receive initial model images of at least two target models and clothing images of at least two sets of target clothing sent by a user, wherein the at least two target models have different model sizes and the at least two sets of target clothing have different clothing sizes;

[0163] Accordingly, the image acquisition module 404 is further configured to:

[0164] The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of at least two target models wearing at least two sets of target clothing.

[0165] Optionally, the image acquisition module 404 is further configured to:

[0166] The neural network model is obtained through the following steps:

[0167] Obtain images of at least two sets of sample clothing and images of at least two sample models;

[0168] Based on the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models, the labeled training samples and the unlabeled training samples are determined, wherein the number of unlabeled training samples is greater than the number of labeled training samples.

[0169] The neural network model is trained using the labeled training samples and the unlabeled training samples.

[0170] Optionally, the image acquisition module 404 is further configured to:

[0171] According to a preset sample processing method, the labeled training samples are determined from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models.

[0172] Based on the labeled training samples, the unlabeled training samples are determined from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models.

[0173] Optionally, the image acquisition module 404 is further configured to:

[0174] Select the sample clothing image and the sample model image to be matched from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models;

[0175] According to the preset sample processing method, the image of the clothing to be matched is matched with the image of the model to be matched to obtain the target sample model image of the model wearing the clothing to be matched.

[0176] The clothing image and the model image to be matched are determined as training samples, and the model image of the target sample is determined as the sample label of the training samples.

[0177] The labeled training samples are determined based on the training samples and the sample labels.

[0178] Optionally, the image acquisition module 404 is further configured to:

[0179] The labeled training samples and the unlabeled training samples are determined according to a preset ratio.

[0180] The labeled training samples are input into the neural network model to obtain a first output label, and the unlabeled training samples are input into the neural network model to obtain a second output label;

[0181] The neural network model is trained based on the first output label and the second output label.

[0182] Optionally, the image acquisition module 404 is further configured to:

[0183] The loss function is calculated based on the sample labels of the labeled training samples and the first output label, and the neural network model is adjusted based on the loss function.

[0184] Based on the confidence level of the second output label, a target training sample is determined from the unlabeled training samples. The target training sample and the second output label corresponding to the target training sample are added to the labeled training samples, so as to train the neural network model based on the updated labeled training samples.

[0185] The neural network model is obtained when the model training termination condition is met.

[0186] Optionally, the image receiving module 402 is further configured to:

[0187] Receive the initial model image of the target model and the clothing image of the target garment sent by the user through the user interaction interface;

[0188] Accordingly, the image display module 406 is further configured as follows:

[0189] The target model image is returned and displayed to the user through the user interface.

[0190] The neural network-based mannequin and clothing wearing device provided in this specification can adaptively wear target clothing onto a target mannequin using a pre-built neural network model. This eliminates the need for manual intervention to adjust the size of the target clothing during the wearing process, thus solving the problem of not being able to achieve low-cost, large-scale, and short-term mannequin clothing wearing when manually intervening. Furthermore, it can directly output the target mannequin's image wearing the target clothing, eliminating the need to take photos of the target mannequin wearing the target clothing, further improving the user experience.

[0191] The above is an illustrative scheme of a mannequin and clothing wearing device based on a neural network according to this embodiment. It should be noted that the technical solution of this mannequin and clothing wearing device based on a neural network and the technical solution of the aforementioned mannequin and clothing wearing method based on a neural network belong to the same concept. For details not described in detail in the technical solution of the mannequin and clothing wearing device based on a neural network, please refer to the description of the technical solution of the aforementioned mannequin and clothing wearing method based on a neural network.

[0192] In addition, one embodiment of this specification also provides a method for training an image processing model, which specifically includes the following steps.

[0193] Obtain images of at least two sets of sample clothing and images of at least two sample models.

[0194] The labeled training samples and the unlabeled training samples are determined based on the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models.

[0195] Wherein, the number of unlabeled training samples is greater than the number of labeled training samples;

[0196] The image processing model is trained using the labeled training samples and the unlabeled training samples, wherein the image processing model is a neural network model.

[0197] The image processing model can be understood as the neural network model or convolutional neural network model in the above embodiments.

[0198] In practical applications, the training method of the image processing model in this embodiment is the same as the specific implementation of the training method of the neural network model in the above embodiment. For details of the technical solution of the image processing model training method that are not described in detail, please refer to the description of the technical solution of the neural network model training method in the above-mentioned neural network-based model and clothing wearing method.

[0199] The image processing model training method provided in the embodiments of this specification adopts a semi-supervised method for network learning. This not only avoids the need for a large amount of labeled data as in fully supervised methods, but also makes the final convolutional neural network perform better than unsupervised learning. Furthermore, it adopts end-to-end (i.e., model training can be implemented on the device) convolutional neural network learning, making the entire process fully automatic without manual intervention, thus solving the problems of slow and costly manual generation of training samples.

[0200] Accordingly, one embodiment of this specification provides a training apparatus for an image processing model, comprising:

[0201] The sample image acquisition module is configured to acquire sample clothing images of at least two sets of sample clothing and sample model images of at least two sample models;

[0202] The training sample determination module is configured to determine the labeled training samples and the unlabeled training samples based on the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models, wherein the number of unlabeled training samples is greater than the number of labeled training samples.

[0203] The model training module is configured to train the image processing model based on the labeled training samples and the unlabeled training samples, wherein the image processing model is a neural network model.

[0204] The above is a schematic scheme of an image processing model training device according to this embodiment. It should be noted that the technical solution of this image processing model training device and the technical solution of the image processing model training method described above belong to the same concept. For details not described in detail in the technical solution of the image processing model training device, please refer to the description of the technical solution of the image processing model training method described above.

[0205] See Figure 5 , Figure 5 A flowchart illustrating another method for wearing clothing based on a neural network, according to an embodiment of this specification, is shown, specifically including the following steps.

[0206] Step 502: Display the image input interface to the user based on the user's request;

[0207] Step 504: Receive the initial model image of the target model and the clothing image of the target garment input by the user through the image input interface;

[0208] Step 506: Input the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of the target model wearing the target clothing;

[0209] Step 508: Display the target model image to the user through the image input interface.

[0210] The neural network model is obtained by training using a semi-supervised method based on labeled and unlabeled training samples.

[0211] The neural network-based mannequin and clothing wearing method provided in this specification can adaptively wear the target clothing onto the target mannequin using a pre-built neural network model. This eliminates the need for manual intervention to adjust the size of the target clothing during the wearing process, thus solving the problem of not being able to achieve low-cost, large-scale, and short-term mannequin clothing wearing when manual intervention is required. Furthermore, it can directly output the target mannequin's image wearing the target clothing, eliminating the need to take photos of the target mannequin wearing the target clothing, further improving the user experience.

[0212] The above is an illustrative scheme of another neural network-based method for wearing clothing on a mannequin. It should be noted that this other neural network-based method for wearing clothing on a mannequin belongs to the same concept as the aforementioned method. Details not described in detail in this other method can be found in the description of the aforementioned method.

[0213] See Figure 6 , Figure 6 A structural block diagram of a computing device 600 according to one embodiment of this specification is shown. The components of the computing device 600 include, but are not limited to, a memory 610 and a processor 620. The processor 620 is connected to the memory 610 via a bus 630, and a database 650 is used to store data.

[0214] The computing device 600 also includes an access device 640, which enables the computing device 600 to communicate via one or more networks 660. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 640 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, a Near Field Communication (NFC) interface, and so on.

[0215] In one embodiment of this specification, the above-described components of the computing device 600 and Figure 6 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 6The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0216] The computing device 600 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 600 can also be a mobile or stationary server.

[0217] The processor 620 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described neural network-based method for wearing clothing on a model.

[0218] The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the above-described neural network-based method for wearing clothing on a model belong to the same concept. For details not described in detail in the technical solution of the computing device, please refer to the description of the technical solution of the above-described neural network-based method for wearing clothing on a model.

[0219] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described neural network-based method for wearing clothing on a model.

[0220] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium belongs to the same concept as the technical solution of the above-described method for wearing clothing on a model based on a neural network. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-described method for wearing clothing on a model based on a neural network.

[0221] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described neural network-based method for wearing models and clothing.

[0222] The above is an illustrative example of a computer program according to this embodiment. It should be noted that the technical solution of this computer program belongs to the same concept as the aforementioned method for wearing clothing on a model based on a neural network. Details not described in detail in the computer program's technical solution can be found in the description of the aforementioned method for wearing clothing on a model based on a neural network.

[0223] This specification also provides an embodiment of a virtual reality device or augmented reality device, including:

[0224] Memory and processor;

[0225] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, perform the following steps:

[0226] Receive the initial model image of the target model and the clothing image of the target garment;

[0227] The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing;

[0228] Return the target model image and render it onto the display interface of the virtual reality device or the display of the augmented reality device.

[0229] The above is an illustrative scheme of a virtual reality device or augmented reality device according to this embodiment. It should be noted that the technical solution of this virtual reality device or augmented reality device belongs to the same concept as the above-described method for wearing clothing on a model based on neural networks. For details not described in detail in the technical solution of the virtual reality device or augmented reality device, please refer to the description of the above-described method for wearing clothing on a model based on neural networks.

[0230] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0231] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0232] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0233] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0234] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for wearing clothing on a model based on a neural network, comprising: Receive the initial model image of the target model and the clothing image of the target garment; The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing; Return the image of the target model. The neural network model is trained using a semi-supervised method based on labeled and unlabeled training samples. This model is used to obtain images of a target model with one or more different model sizes combined with one or more sets of target clothing in different sizes. The training method includes: calculating a loss function and adjusting the neural network model based on the sample labels and first output labels of the labeled training samples; determining target training samples from the unlabeled training samples based on the confidence level of the second output label; adding the target training samples and their corresponding second output labels to the labeled training samples; and training the neural network model based on the updated labeled training samples. The number of unlabeled training samples is greater than the number of labeled training samples.

2. The method for wearing clothing on a model based on a neural network according to claim 1, wherein the initial model image of the target model and the clothing image of the target clothing include: Receive initial model images of at least two target models and an image of a set of target clothing sent by a user, wherein the at least two target models have different sizes; Accordingly, the step of inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing includes: The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images.

3. The method for wearing clothing on a model based on a neural network according to claim 1, wherein receiving the initial model image of the target model and the clothing image of the target clothing includes: Receive an initial model image of a target model and clothing images of at least two sets of target clothing sent by a user, wherein the clothing sizes of the at least two sets of target clothing are different; Accordingly, the step of inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing includes: The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images.

4. The method for wearing clothing on a model based on a neural network according to claim 1, wherein receiving the initial model image of the target model and the clothing image of the target clothing includes: Receive initial model images of at least two target models and clothing images of at least two sets of target clothing sent by a user, wherein the at least two target models have different model sizes and the at least two sets of target clothing have different clothing sizes; Accordingly, the step of inputting the initial model image and the clothing image into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing includes: The initial model image and the clothing image are input into a pre-trained neural network model to obtain at least two target model images of at least two target models wearing at least two sets of target clothing.

5. The method for wearing clothing on a model based on a neural network according to any one of claims 1-4, wherein the neural network model is obtained through the following steps: Obtain images of at least two sets of sample clothing and images of at least two sample models; The labeled training samples and the unlabeled training samples are determined based on the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models. The neural network model is trained using the labeled training samples and the unlabeled training samples.

6. The method for model and clothing wearing based on a neural network according to claim 5, wherein determining the labeled training samples and the unlabeled training samples based on the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models includes: According to a preset sample processing method, the labeled training samples are determined from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models. Based on the labeled training samples, the unlabeled training samples are determined from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models.

7. The method for model and clothing wearing based on neural networks according to claim 6, wherein determining the labeled training samples from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models according to a preset sample processing method includes: Select the sample clothing image and the sample model image to be matched from the sample clothing images of the at least two sets of sample clothing and the sample model images of the at least two sample models; According to the preset sample processing method, the image of the clothing to be matched is matched with the image of the model to be matched to obtain the target sample model image of the model wearing the clothing to be matched. The clothing image and the model image to be matched are determined as training samples, and the model image of the target sample is determined as the sample label of the training samples. The labeled training samples are determined based on the training samples and the sample labels.

8. The method for wearing clothing on a model based on a neural network according to claim 5, wherein training the neural network model based on the labeled training samples and the unlabeled training samples comprises: The labeled training samples and the unlabeled training samples are determined according to a preset ratio. The labeled training samples are input into the neural network model to obtain a first output label, and the unlabeled training samples are input into the neural network model to obtain a second output label; The neural network model is obtained by training based on the first output label and the second output label, provided that the model training termination condition is met.

9. The method for wearing clothing on a model based on a neural network according to claim 1, wherein receiving the initial model image of the target model and the clothing image of the target clothing includes: Receive the initial model image of the target model and the clothing image of the target garment sent by the user through the user interaction interface; Accordingly, returning the target model image includes: The target model image is returned and displayed to the user through the user interface.

10. A method for wearing clothing on a model based on a neural network, comprising: Display an image input interface to the user based on the user's request; Receive the initial model image of the target model and the clothing image of the target garment input by the user through the image input interface; The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing; The target model image is displayed to the user through the image input interface. The neural network model is trained using a semi-supervised method based on labeled and unlabeled training samples. This model is used to obtain images of a target model with one or more different model sizes combined with one or more sets of target clothing in different sizes. The training method includes: calculating a loss function and adjusting the neural network model based on the sample labels and first output labels of the labeled training samples; determining target training samples from the unlabeled training samples based on the confidence level of the second output label; adding the target training samples and their corresponding second output labels to the labeled training samples; and training the neural network model based on the updated labeled training samples. The number of unlabeled training samples is greater than the number of labeled training samples.

11. A virtual reality device or augmented reality device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, perform the following steps: Receive the initial model image of the target model and the clothing image of the target garment; The initial model image and the clothing image are input into a pre-trained neural network model to obtain a target model image of a target model wearing the target clothing. The neural network model is trained using a semi-supervised method based on labeled and unlabeled training samples. The neural network model is used to obtain a target model image of one model size or multiple target models of different sizes combined with one set of target clothing or multiple sets of target clothing of different sizes. The training method of the neural network model includes: calculating a loss function and adjusting the neural network model based on the sample labels and first output labels of the labeled training samples; determining target training samples from the unlabeled training samples based on the confidence of the second output label; adding the target training samples and their corresponding second output labels to the labeled training samples; and training the neural network model based on the updated labeled training samples. The number of unlabeled training samples is greater than the number of labeled training samples. Return the target model image and render it onto the display interface of the virtual reality device or the display of the augmented reality device.

12. A computing device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the neural network-based model and clothing wearing method according to any one of claims 1 to 10.

13. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the neural network-based method for wearing clothing as described in any one of claims 1 to 10.

Images