Clothing intelligent matching system, image generation method thereof and storage medium

Abstract

The invention relates to an intelligent costume matching system, an image generation method thereof and a storage medium, and belongs to the technical field of deep learning and big data. The method comprises the following steps: on the basis of a to-be-recommended costume image input by a user, identifying and segmenting individual costumes in the to-be-recommended costume image to obtain a costume segmented image; inputting the clothing segmented image into a pre-established image library for retrieval recall to obtain a candidate clothing image set; obtaining a clothing suit edge detection image based on each clothing image, and obtaining a candidate clothing suit edge detection image set; and finally, inputting the clothing segmentation image and the clothing suit edge detection image into an image generation model, and outputting a generated clothing wearing and fitting suit image. According to the method, the image generation model is subjected to fine adjustment training in advance, so that the model has the capability of generating various high-quality matching suits for any given costumes, the diversity and novelty of the intelligent costume matching system are improved, and the overall effect is improved.

Description

Technical Field

[0001] The present application belongs to the field of deep learning and big data technology, and specifically relates to an intelligent clothing matching system and its image generation method and storage medium. Background Art

[0002] With the rapid development of artificial intelligence and big data technologies, how to reconstruct and upgrade traditional products and businesses based on artificial intelligence technology has become a hot research topic.

[0003] As part of the fashion industry, clothing has a large share in the consumer market. How to use artificial intelligence technology to improve the intelligence level and user experience of the clothing industry is a very challenging and meaningful topic. With the continuous improvement of economic level, fashionable wear has become a very common demand of consumers. Therefore, intelligent clothing matching has become one of the most popular research and application directions in the clothing industry.

[0004] Traditional clothing intelligent matching systems based on deep learning usually pre-build a dataset of positive and negative sample pairs of clothing matching, train a matching probability prediction model based on the matching dataset, and then complete the final clothing matching solution recommendation in combination with the business rule system built by business experts. Traditional methods usually rely more on manual data annotation or established rules, with a high degree of reliance on manual experts, and only focus on the matching feature modeling of single product granularity, lacking matching feature modeling of suit granularity, resulting in the problem of limited matching diversity and novelty.

[0005] Therefore, how to solve the problem of limited diversity and novelty of traditional clothing intelligent matching systems is a topic worthy of study in this field. Summary of the invention

[0006] In view of the above analysis, the embodiments of the present invention aim to provide an intelligent clothing matching system and an image generation method and storage medium thereof, so as to solve the problem of limited diversity and novelty of matching in traditional intelligent clothing matching systems based on deep learning.

[0007] In a first aspect of the present application, a method for generating an image of an intelligent clothing matching system is provided, comprising:

[0008] Obtaining the clothing image to be recommended input by the user;

[0009] Identifying and segmenting individual clothing in the clothing image to be recommended to obtain a clothing segmentation image;

[0010] Inputting the clothing segmentation image into a pre-established image library for retrieval and recall, obtaining clothing images matching the clothing segmentation image, and obtaining a candidate clothing image set;

[0011] Based on each clothing image, a corresponding clothing set edge detection image is obtained to obtain a candidate clothing set edge detection image set;

[0012] Inputting the clothing segmentation image and the clothing set edge detection image into an image generation model of a clothing intelligent matching system, and outputting a generated clothing set image;

[0013] Among them, a clothing intelligent matching image generation fine-tuning training data set is created in advance, including collecting large-scale labeled clothing images of different types and styles and corresponding clothing set edge detection images, and the image generation model is a model obtained by fine-tuning training on the clothing intelligent matching image generation fine-tuning training data set.

[0014] Optionally, the step of inputting the clothing segmentation image into a pre-established image library for retrieval and recall to obtain clothing images matching the clothing segmentation image, and obtaining a candidate clothing image set includes:

[0015] Vectorize the clothing segmentation image using a vectorization model to obtain a clothing segmentation image vector;

[0016] The clothing segmentation image vector is input into the image vector library for retrieval and recall. According to the similarity between the clothing segmentation image vector and the clothing image vector in the vector database, the top N results with high correlation with the clothing segmentation image vector are taken as candidate clothing image sets.

[0017] Optionally, fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training data set includes:

[0018] Collect clothing image materials as original clothing set images;

[0019] Performing image segmentation of clothing granularity on the original clothing set image, and dividing the original clothing set image into a plurality of sub-images according to the clothing segmentation result, each sub-image corresponding to a separate clothing, to obtain a clothing image;

[0020] Performing edge detection on the original clothing suit image to obtain a clothing suit edge detection image;

[0021] The clothing image and the clothing suit edge detection image are used as a clothing intelligent matching image generation fine-tuning training data set to perform fine-tuning training on the image generation model.

[0022] Optionally, the identifying and segmenting individual clothing in the clothing image to be recommended to obtain a clothing segmentation image comprises:

[0023] Preprocessing the clothing image to be recommended, wherein the preprocessing includes adjusting the clothing image to be recommended to a predetermined size, normalizing the clothing image to be recommended to a specific range, and performing data enhancement on the clothing image to be recommended;

[0024] Load a pre-trained SAM segmentation model, input the pre-processed image into the SAM segmentation model, and output the category label of each pixel in the image;

[0025] Based on the category label of each pixel, identify the connected regions in the segmented image, each region corresponds to a separate garment;

[0026] According to the boundary of each piece of clothing, the corresponding sub-image is cut out from the original image. Each sub-image corresponds to a separate piece of clothing, and a clothing segmentation image is obtained.

[0027] Optionally, acquiring the corresponding clothing suit edge detection image based on each clothing image to obtain the candidate clothing suit edge detection image set includes:

[0028] For each clothing image, use different edge detection algorithms or adjust the parameters of the edge detection algorithm to perform multiple edge detection processes on the clothing image to obtain M clothing set edge detection images as the candidate clothing set edge detection image set;

[0029] The edge detection algorithm includes: a Canny edge detection algorithm, a Sobel operator edge detection algorithm or a Laplacian operator edge detection algorithm; and the parameters of the edge detection algorithm include a threshold value or a filter size of the edge detection algorithm.

[0030] Optionally, after outputting the generated clothing set image, the method further includes:

[0031] Based on the user portrait information, the generated clothing outfit set images are sorted, and recommendation results for the clothing outfit set images are generated and fed back to the user.

[0032] Optionally, the image generation model adopts a GAN model; the step of inputting the clothing segmentation image and the clothing set edge detection image into the image generation model of the clothing intelligent matching system, and outputting the generated clothing set image comprises:

[0033] The generator receives the clothing image and the clothing set edge detection image as input, and aims to generate a clothing set image. The generator is composed of a multi-layer convolutional neural network, and constructs the final image by gradually adding details and textures.

[0034] The discriminator evaluates the similarity between the image output by the generator and the actual image; the discriminator is composed of a convolutional neural network and its task is to determine whether the input image is real or generated;

[0035] During the training process, the generator and the discriminator compete against each other until the training is completed.

[0036] Optionally, fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training dataset includes: fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training dataset using a LoRA fine-tuning method or a full parameter fine-tuning method.

[0037] According to a second aspect of the present application, there is provided an intelligent clothing matching system, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, any one of the above-mentioned intelligent clothing matching system image generation methods is implemented.

[0038] According to a third aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the image generation method of the intelligent clothing matching system described above is implemented.

[0039] The image generation method of the intelligent clothing matching system provided in the present application is based on the clothing image to be recommended input by the user, and obtains the clothing segmentation image by identifying and segmenting the individual clothing in the clothing image to be recommended; the clothing segmentation image is input into a pre-established image library for retrieval and recall to obtain a set of candidate clothing images; based on each clothing image, the corresponding clothing set edge detection image is obtained to obtain a set of candidate clothing set edge detection images; finally, the clothing segmentation image and the clothing set edge detection image are input into the image generation model of the intelligent clothing matching system, and the generated clothing set image is output. The present application pre-fine-tunes the image generation model with large-scale high-quality dressing sample data, so that the model has the ability to generate a variety of high-quality matching sets for any given clothing, thereby improving the diversity and novelty of the intelligent clothing matching system and improving the overall effect.

[0040] In addition, the present application also provides a clothing intelligent matching system and a computer-readable storage medium having the above-mentioned technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0042] Figure 1 A flowchart of a specific implementation method of the clothing intelligent matching system image generation method provided in this application;

[0043] Figure 2 Flowchart of the fine-tuning training process for the clothing intelligent matching image generation model;

[0044] Figure 3 A schematic diagram of a specific image example of the clothing intelligent matching system provided in this application;

[0045] Figure 4 Schematic diagram of the process of generating a model for fine-tuning training images;

[0046] Figure 5 This is a structural block diagram of the intelligent clothing matching system provided in this application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. It should be noted that, in the absence of conflict, the embodiments in the present disclosure and the features in the embodiments can be combined, separated, interchanged and / or rearranged with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0048] The terms used here are for the purpose of describing specific embodiments, and are not intended to be restrictive. As used here, unless the context clearly indicates otherwise, the singular forms "one (kind, person)" and "said (the)" are also intended to include plural forms. In addition, when the terms "comprise" and / or "include" and their variations are used in this specification, it is explained that there are stated features, integral bodies, steps, operations, parts, assemblies and / or their groups, but it is not excluded that there are or add one or more other features, integral bodies, steps, operations, parts, assemblies and / or their groups. It should also be noted that, as used here, the terms "substantially", "approximately" and other similar terms are used as approximate terms and not as degree terms, so that they are used to explain the inherent deviations of the measured values, calculated values ​​and / or the values ​​provided that will be recognized by those of ordinary skill in the art.

[0049] A flowchart of a specific implementation of the method for generating an image of an intelligent clothing matching system provided in this application is as follows: Figure 1 As shown, the method specifically includes:

[0050] S101: Obtaining a clothing image to be recommended input by a user.

[0051] The user uploads the clothing image that he / she wishes to match, which is used as the clothing image to be recommended, denoted as I. This can be achieved through an image upload interface, and the user can select an image from a user terminal or other human-computer interaction device and upload it to the system.

[0052] S102: Identify and segment individual clothing in the clothing image to be recommended to obtain a clothing segmentation image.

[0053] Image segmentation technology, such as deep learning model, is used to segment the input clothing image I to be recommended, identify and segment each independent clothing in the user uploaded image, and generate clothing segmentation image, denoted as IS. The image segmentation model can use SAM model, which is not limited here.

[0054] Before the image is identified and segmented, the method may further include: performing a preprocessing operation on the clothing image to be recommended. The preprocessing includes but is not limited to: adjusting the clothing image to be recommended to a predetermined size, normalizing the clothing image to be recommended to a specific range (e.g., 0 to 1), and performing a data enhancement operation on the clothing image to be recommended.

[0055] The segmentation process is specifically as follows: loading a pre-trained SAM segmentation model, inputting the preprocessed image into the SAM segmentation model, and outputting the category label of each pixel in the image; based on the category label of each pixel, identifying the connected areas in the segmented image, each area corresponds to a separate piece of clothing; according to the boundary of each piece of clothing, cutting out the corresponding sub-image from the original image, each sub-image corresponds to a separate piece of clothing, and obtaining a clothing segmentation image.

[0056] By accurately segmenting each piece of clothing in the image, the characteristics of each piece of clothing, such as design, color, and material, can be analyzed individually, which provides a basis for subsequent clothing matching and helps to improve the accuracy and effectiveness of recommendations.

[0057] S103: Inputting the clothing segmentation image into a pre-established image library for retrieval and recall, obtaining clothing images matching the clothing segmentation image, and obtaining a candidate clothing image set.

[0058] Create an image library that stores a large number of clothing images and related features. Use image recognition and matching algorithms to find clothing images that match the clothing segmentation images uploaded by users in style, color or pattern to form a candidate clothing image set.

[0059] As a specific implementation, a vector similarity calculation algorithm (such as cosine similarity, Euclidean distance, etc.) can be used for matching. The process specifically includes: using a vectorization model to vectorize the clothing segmentation image to obtain a clothing segmentation image vector, which is recorded as ISV. The vectorization model can use a mainstream model, such as CLIP (Contrastive Language-Image Pre-training), etc., which does not affect the implementation of this application.

[0060] The clothing segmentation image vector ISV is input into the image vector library for retrieval and recall. According to the similarity between the clothing segmentation image vector and the clothing image vector in the vector database, the top N results with high correlation with the clothing segmentation image vector are taken as the candidate clothing image set. The top N results are taken as the candidate clothing image set, denoted as KRS, and the value of N can be set to 3.

[0061] S104: acquiring corresponding clothing set edge detection images based on each clothing image, and obtaining a candidate clothing set edge detection image set.

[0062] Apply edge detection algorithms (such as Canny, Sobel) to each image in the candidate clothing image set to obtain edge detection images of clothing sets. These edge detection images emphasize the outline and details of clothing, which are crucial for subsequent image synthesis.

[0063] For each clothing image, a different edge detection algorithm is used or the parameters of the edge detection algorithm are adjusted to perform multiple edge detection processes on the clothing image, and M clothing set edge detection images are obtained as the candidate clothing set edge detection image set. The edge detection algorithm includes: Canny edge detection algorithm, Sobel operator edge detection algorithm or Laplacian operator edge detection algorithm; the parameters of the edge detection algorithm include the threshold value or filter size of the edge detection algorithm.

[0064] For each clothing image, M clothing set edge detection images are randomly selected to obtain a candidate clothing set edge detection image set. As an example, the value of M can be set to 5.

[0065] S105: Input the clothing segmentation image and the clothing set edge detection image into an image generation model of the clothing intelligent matching system, and output the generated clothing set image.

[0066] The clothing segmentation image IS and the clothing set edge detection image (N*M images) are taken as input and input into the trained image generation model, and the generated clothing set image (N*M images) is output, denoted as IG, which is presented to the user as the generated clothing intelligent matching effect.

[0067] The image generation model is pre-fine-tuned on a fine-tuning training dataset consisting of a large-scale labeled clothing image and its edge detection image to improve the accuracy and realism of the generated image. The image generation model can specifically adopt a generative adversarial network (GAN) or other models. The image generation model can be deployed on a cloud server.

[0068] Reference Figure 2 Flow chart of the fine-tuning training process of the clothing intelligent matching image generation model. The fine-tuning training process of the clothing intelligent matching image generation model includes:

[0069] S201: Create a clothing intelligent matching image generation fine-tuning training dataset.

[0070] Pre-create clothing intelligent matching image generation fine-tuning training dataset, including collecting large-scale labeled clothing images of different types and styles and corresponding clothing set edge detection images.

[0071] The clothing intelligent matching image generation fine-tuning training data set of the embodiment of the present invention is constructed using the model automatic annotation form. The input and output are both in the form of images, where the input includes clothing segmentation images and clothing set edge detection images, and the output is the generated clothing set image. The training data set format sample of the embodiment of the present invention is as follows:

[0072] {"input_segment":"input_segment.png","input_canny":"input_canny.png","output_target":"output_target.png"}

[0073] Among them, "input_segment" represents the input clothing segmentation image, "input_canny" represents the input clothing set edge detection image, and "output_target" represents the output clothing set image. A schematic diagram of a specific image example of the clothing intelligent matching system provided in this application is as follows: Figure 3 shown.

[0074] S202: Fine-tuning training of clothing intelligent matching image generation model.

[0075] The image generation model is a model obtained by fine-tuning training on the clothing intelligent matching image generation fine-tuning training data set. The image generation model can use existing models in the industry, such as Stable Diffusion, and use the clothing intelligent matching image generation model to complete fine-tuning on the clothing intelligent matching image generation fine-tuning training data set created above to enhance the image generation model's ability to generate clothing intelligent matching images.

[0076] The image generation model will output creative and practical clothing matching solutions based on the input image features, which can reflect the user's initial selection and style preferences. Users can give feedback on the generated matching results, which helps to further train and optimize the image generation model.

[0077] Fine-tuning the image generation model on the clothing intelligent matching image generation fine-tuning training dataset includes: fine-tuning the image generation model on the clothing intelligent matching image generation fine-tuning training dataset using a low-rank adaptation of large language models (LoRA) fine-tuning method or a full parameter fine-tuning method. The loss function can select the mean squared error loss function (MSE), which is used to measure the average difference between the predicted value and the true value of the model. In the task of image generation, this can be understood as calculating the difference between the generated image and the real image at each pixel. By minimizing this difference, the model can learn how to generate an output that is closer to the real image.

[0078] After the output of the generated clothing set image, it also includes: based on the user portrait information, sorting the generated clothing set images, generating recommendation results for the clothing set images, and feeding back to the user. The user portrait is a model built based on a variety of user data, which includes but is not limited to: basic information, historical behavior, preference settings, social media activities, etc. After the system generates multiple matching options through the image generation model), these options need to be processed and evaluated: key features are extracted from each generated clothing set image, including color, style, seasonality, clothing category, etc. The features of the generated clothing set image are compared with the preferences in the user portrait, and a collaborative filtering algorithm is used to evaluate the relevance and potential user satisfaction of each match. The clothing set images are sorted according to the matching scores, and the matching with a high score indicates that it is more in line with the user's preferences and needs.

[0079] Based on the sorting results, the user is presented with clothing matching options that best match their personal profile and preferences, usually in a visual form. The user can evaluate the recommended matching, such as choosing like, dislike, or purchase.

[0080] The image generation method of the intelligent clothing matching system provided in the present application is based on the clothing image to be recommended input by the user, and obtains the clothing segmentation image by identifying and segmenting the individual clothing in the clothing image to be recommended; the clothing segmentation image is input into a pre-established image library for retrieval and recall to obtain a set of candidate clothing images; based on each clothing image, the corresponding clothing set edge detection image is obtained to obtain a set of candidate clothing set edge detection images; finally, the clothing segmentation image and the clothing set edge detection image are input into the image generation model of the intelligent clothing matching system, and the generated clothing set image is output. The present application pre-fine-tunes the image generation model with large-scale high-quality dressing sample data, so that the model has the ability to generate a variety of high-quality matching sets for any given clothing, thereby improving the diversity and novelty of the intelligent clothing matching system and improving the overall effect.

[0081] It is understandable that the clothing images and clothing set edge detection images are pre-collected and constructed. Figure 4 A schematic diagram of the process of fine-tuning and training the image generation model. The process of fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training dataset specifically includes:

[0082] S401: Collect clothing image materials as original clothing set images.

[0083] This process is the data collection stage, the purpose of which is to build a database of diverse, high-quality clothing images. Data can be obtained in the following ways: paid purchase, purchase of high-quality, copyright-clear clothing images from professional image providers or clothing brands. The advantage of this method is that the image quality is usually high and there are fewer copyright issues. It can also be collected from the Internet, collecting public high-quality celebrity outfits or fashion promotional materials. This requires paying attention to copyright issues and ensuring that the images used comply with legal and ethical standards.

[0084] S402: performing image segmentation of the original clothing set image at the clothing granularity, and dividing the original clothing set image into a plurality of sub-images according to the clothing segmentation result, wherein each sub-image corresponds to a separate clothing, to obtain a clothing image.

[0085] Perform fine image segmentation on complex clothing sets to extract individual pieces of clothing. The specific process involves: using image segmentation techniques, such as deep learning models SAM (Space-Aware Multi-scale Model), to accurately segment individual pieces of clothing from complex backgrounds. This process can identify different objects in the image and separate them. Each segmented piece of clothing will be cut into a separate sub-image, which is more convenient for subsequent processing and analysis, such as feature extraction and similarity matching.

[0086] S403: Perform edge detection on the original clothing set image to obtain a clothing set edge detection image.

[0087] Edge detection is to extract edge information from clothing images, which is particularly important for understanding the structure and shape of clothing, and is beneficial for subsequent image generation. Specifically, the Canny edge detection algorithm can be used, which can effectively detect edges in images and has good edge connection and noise suppression capabilities, helping to highlight the outline of clothing. The edge detection result is usually an image that highlights all edges, which is particularly important for describing the geometric and structural features of clothing. Generating edge detection images of clothing sets provides key visual information about the outline of clothing, which can be used for further image analysis or as part of the input of a generative adversarial network.

[0088] S404: Using the clothing image and the clothing suit edge detection image as a clothing intelligent matching image generation fine-tuning training data set, and performing fine-tuning training on the image generation model.

[0089] The image generation model used in the embodiment of the present invention is Stable Diffusion. Of course, other image generation models, such as GAN, etc., can also be used to achieve the same effect as the present invention. The generative adversarial network (GAN) consists of two parts: a generator and a discriminator. The generator is responsible for generating images that are as realistic as possible, while the discriminator's task is to distinguish between generated images and real images. The two are in conflict with each other, and the quality of the generated images is continuously improved through this confrontation process. The following is a further introduction to the process of image generation using the GAN model. In this embodiment, the clothing segmentation image and the clothing suit edge detection image are input into the image generation model of the clothing intelligent matching system, and the output generated clothing matching suit image includes:

[0090] The generator receives the clothing image and the clothing set edge detection image as input, and aims to generate a clothing set image. The generator is composed of a multi-layer convolutional neural network, and constructs the final image by gradually adding details and textures.

[0091] The discriminator evaluates the similarity between the image output by the generator and the actual image; the discriminator is composed of a convolutional neural network and its task is to determine whether the input image is real or generated;

[0092] During the training process, the generator and the discriminator compete against each other until the training is completed.

[0093] In addition, the present application also provides a clothing intelligent matching system, such as Figure 5 As shown in the structural block diagram of the clothing intelligent matching system provided in the present application, the system specifically includes: a memory 51 and a processor 52, wherein the memory 51 stores a computer program, and when the computer program is executed by the processor 52, any one of the above-mentioned clothing intelligent matching system image generation methods is implemented. Figure 5 The smart clothing matching system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0094] In addition, the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the image generation method of the intelligent clothing matching system according to any of the above-mentioned methods is implemented.

[0095] It should be noted that the computer-readable medium mentioned above in the present disclosure may be a computer-readable storage medium, a computer-readable signal medium, or any combination of the above two.

[0096] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0097] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer readable program codes. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than a computer readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0098] The computer-readable medium may be included in the computing device; or it may exist independently without being assembled into the intelligent clothing matching system.

[0099] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the tester computer, partially on the tester computer, as a separate software package, partially on the tester computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the tester computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0100] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated system based on hardware that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0101] It can be understood that the specific implementation process of the intelligent clothing matching system and computer-readable storage medium provided by the present application can refer to the implementation described in the above-mentioned intelligent clothing matching system image generation method, and will not be repeated here.

[0102] This application pre-fine-tunes the image generation model with a large amount of high-quality outfit sample data, so that the model has the ability to generate a variety of high-quality matching sets for any given clothing, thereby improving the diversity and novelty of the clothing intelligent matching system and improving the overall effect.

[0103] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0104] The specific implementation methods described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only the specific implementation method of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for generating an image of an intelligent clothing matching system, characterized in that: include: Obtaining the clothing image to be recommended input by the user; Identifying and segmenting individual clothing in the clothing image to be recommended to obtain a clothing segmentation image; Inputting the clothing segmentation image into a pre-established image library for retrieval and recall, obtaining clothing images matching the clothing segmentation image, and obtaining a candidate clothing image set; Based on each clothing image, a corresponding clothing set edge detection image is obtained to obtain a candidate clothing set edge detection image set; Inputting the clothing segmentation image and the clothing set edge detection image into an image generation model of a clothing intelligent matching system, and outputting a generated clothing set image; Among them, a clothing intelligent matching image generation fine-tuning training data set is created in advance, including collecting large-scale labeled clothing images of different types and styles and corresponding clothing set edge detection images, and the image generation model is a model obtained by fine-tuning training on the clothing intelligent matching image generation fine-tuning training data set.

2. The image generation method according to claim 1, characterized in that: The clothing segmentation image is input into a pre-established image library for retrieval and recall, and clothing images matching the clothing segmentation image are obtained, and the candidate clothing image set is obtained, including: Vectorize the clothing segmentation image using a vectorization model to obtain a clothing segmentation image vector; The clothing segmentation image vector is input into the image vector library for retrieval and recall. According to the similarity between the clothing segmentation image vector and the clothing image vector in the vector database, the top N results with high correlation with the clothing segmentation image vector are taken as candidate clothing image sets.

3. The image generation method according to claim 1, characterized in that: Fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training data set includes: Collect clothing image materials as original clothing set images; Performing image segmentation of clothing granularity on the original clothing set image, and dividing the original clothing set image into a plurality of sub-images according to the clothing segmentation result, each sub-image corresponding to a separate clothing, to obtain a clothing image; Performing edge detection on the original clothing suit image to obtain a clothing suit edge detection image; The clothing image and the clothing suit edge detection image are used as a clothing intelligent matching image generation fine-tuning training data set to perform fine-tuning training on the image generation model.

4. The image generation method according to claim 3, characterized in that: The identifying and segmenting the individual clothing in the clothing image to be recommended to obtain the clothing segmentation image comprises: Preprocessing the clothing image to be recommended, wherein the preprocessing includes adjusting the clothing image to be recommended to a predetermined size, normalizing the clothing image to be recommended to a specific range, and performing data enhancement on the clothing image to be recommended; Load a pre-trained SAM segmentation model, input the pre-processed image into the SAM segmentation model, and output the category label of each pixel in the image; Based on the category label of each pixel, identify the connected regions in the segmented image, each region corresponds to a separate garment; According to the boundary of each piece of clothing, the corresponding sub-image is cut out from the original image. Each sub-image corresponds to a separate piece of clothing, and a clothing segmentation image is obtained.

5. The image generation method according to any one of claims 1 to 4, characterized in that: The step of acquiring the corresponding clothing suit edge detection image based on each clothing image to obtain the candidate clothing suit edge detection image set comprises: For each clothing image, use different edge detection algorithms or adjust the parameters of the edge detection algorithm to perform multiple edge detection processes on the clothing image to obtain M clothing set edge detection images as the candidate clothing set edge detection image set; The edge detection algorithm includes: a Canny edge detection algorithm, a Sobel operator edge detection algorithm or a Laplacian operator edge detection algorithm; and the parameters of the edge detection algorithm include a threshold value or a filter size of the edge detection algorithm.

6. The image generation method according to any one of claims 1 to 4, characterized in that: After the output of the generated clothing set image, the method further includes: Based on the user portrait information, the generated clothing outfit set images are sorted, and recommendation results for the clothing outfit set images are generated and fed back to the user.

7. The image generation method according to any one of claims 1 to 4, characterized in that: The image generation model adopts a GAN model; the clothing segmentation image and the clothing set edge detection image are input into the image generation model of the clothing intelligent matching system, and the clothing set image generated by outputting includes: The generator receives the clothing image and the clothing set edge detection image as input, and aims to generate a clothing set image. The generator is composed of a multi-layer convolutional neural network, and constructs the final image by gradually adding details and textures. The discriminator evaluates the similarity between the image output by the generator and the actual image; the discriminator is composed of a convolutional neural network and its task is to determine whether the input image is real or generated; During the training process, the generator and the discriminator compete against each other until the training is completed.

8. The image generation method according to claim 7, characterized in that: Fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training data set includes: fine-tuning and training the image generation model on the clothing intelligent matching image generation fine-tuning training data set using a LoRA fine-tuning method or a full parameter fine-tuning method.

9. An intelligent clothing matching system, characterized in that: It includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the image generation method of the clothing intelligent matching system according to any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the image generation method of the clothing intelligent matching system according to any one of claims 1-8 is implemented.

Images