Commodity picture display method and device, equipment and medium

Abstract

The invention relates to a commodity picture display method and device, equipment and a medium in the technical field of e-commerce, and the method comprises the steps: judging whether a user belongs to a target user or not according to a behavior accumulation data set of each user in an online shop, and for each target user, determining whether the user belongs to the target user or not; according to the behavior log of the target user, determining an interested commodity associated with the target behavior in the behavior log, constructing a picture preference vector of the target user according to a picture feature vector corresponding to a commodity display picture of each interested commodity, and on this basis, obtaining a picture preference vector of the target user; determining a preference score between the picture feature vector of each commodity picture of the commodities in the online shop and the picture preference vector of each target user in the target user pool, and fusing the preference scores corresponding to all the target users for each commodity picture to obtain a comprehensive score, and taking the commodity picture with the optimal comprehensive score as a commodity display picture of the corresponding commodity. According to the invention, commodity pictures attracting users can be displayed.

Description

Technical Field

[0001] The present application relates to the field of e-commerce technology, and in particular to a product image display method and corresponding devices, electronic devices, and computer-readable storage media. Background Art

[0002] On e-commerce websites, a product is usually accompanied by multiple pictures. Selecting a picture to be shown to users first during the product display process will help attract user attention and promote the transaction process, such as clicking on the product, adding the product to the shopping cart, etc., until it is finally purchased, thereby achieving the goal of improving the conversion rate of the product and increasing store revenue.

[0003] In traditional technology, the primary display pictures of products are selected by the operators of online stores. However, such pictures selected based solely on the operators' own experience and / or personal ideas may not be able to attract users, resulting in the product display failing to achieve the expected results.

[0004] In view of the shortcomings of traditional practices, the applicant has been engaged in research in related fields for a long time and has taken a different approach to solve the problems in the industry. Summary of the invention

[0005] The primary purpose of the present application is to solve at least one of the above problems and provide a product picture display method and corresponding device, electronic device, and computer program product.

[0006] In order to meet the various objectives of this application, this application adopts the following technical solutions:

[0007] A method for displaying product images provided to meet one of the purposes of this application includes the following steps:

[0008] Based on the accumulated data set of each user's behavior in the online store, determine whether the user is a target user and construct a target user pool corresponding to each target user;

[0009] For each target user in the target user pool, determine the interest products associated with the target behavior according to the behavior log of the target user, and construct an interest product list of the target user corresponding to the target behavior with each interest product;

[0010] constructing a picture preference vector of the target user according to the picture feature vector corresponding to the product display picture of each product of interest in the list of products of interest;

[0011] Determine the preference score between the image feature vector of each product image of the products in the online store and the image preference vector of each target user in the target user pool, integrate the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and use the product image with the best comprehensive score as the product display image of the corresponding product.

[0012] On the other hand, a product image display device provided to meet one of the purposes of the present application includes a user pool construction module, an interest mining module, a preference representation module and a display optimization module, wherein the user pool construction module is used to determine whether a user belongs to a target user based on the accumulated data set of the behavior of each user in the online store, and construct a target user pool corresponding to each target user; the interest mining module is used to determine, for each target user in the target user pool, the interest products associated with the target behavior according to the behavior log of the target user, and construct an interest product list of the target user corresponding to the target behavior with each interest product; the preference representation module is used to construct an image preference vector of the target user based on the image feature vector corresponding to the product display image of each interest product in the interest product list; the display optimization module is used to determine the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, and obtain a comprehensive score by integrating the preference scores corresponding to all target users for each product image, and use the product image with the best comprehensive score as the product display image of the corresponding product.

[0013] On the other hand, an electronic device provided to meet one of the purposes of the present application includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the product image display method described in the present application.

[0014] On the other hand, a computer program product provided to meet another purpose of the present application includes a computer program / instruction, which, when executed by a processor, implements the steps of the method described in any embodiment of the present application.

[0015] The technical solution of this application has many advantages, including but not limited to the following aspects:

[0016] This application analyzes the cumulative performance of various behaviors of each user in the online store, identifies each target user who meets the standard to construct a target user pool, and then for each target user, uses the behavior log to determine the various product display pictures that the target user is interested in, and constructs the target user's picture preference vector based on the picture feature vectors corresponding to these pictures. Further, the preference score between the picture feature vector of each product picture in the online store and the picture preference vector of each target user is determined, and the preference scores corresponding to all target users are integrated for each product picture to obtain a comprehensive score, and the product picture with the best comprehensive score is selected as the product display picture of the corresponding product. It can be seen that the target user group of the online store can be specifically refined, and the preference of each target user for the product picture can be accurately and quantitatively expressed, ensuring that the preference score determined according to the quantitative expression corresponding to the picture preference and the product picture is accurate and reliable. On this basis, for each product picture, the comprehensive score is obtained by integrating the preferences of all target users for the product picture, and the one with the highest comprehensive score is selected as the product display picture, ensuring that when it displays the product, it can attract the attention of the target user, enhance the user's shopping experience, and help promote the completion of product transactions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0018] Figure 1 A schematic diagram of a typical network deployment architecture for implementing the technical solution of this application;

[0019] Figure 2 This is a flowchart of a typical embodiment of the product image display method of the present application;

[0020] Figure 3 This is a functional block diagram of the product image display device of this application;

[0021] Figure 4 This is a schematic diagram of the structure of an electronic device used in this application.

[0022] Specific implementation method

[0023] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be interpreted as limiting the present application.

[0024] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be an intermediate element. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0025] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as those generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with those in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as here.

[0026] like Figure 1 In the network architecture shown, the e-commerce platform 82 is deployed on the Internet to provide corresponding services to its users. Similarly, the devices 80 of the merchant users of the e-commerce platform 82 and the devices 81 of the consumer users are also connected to the Internet to use the services provided by the e-commerce platform.

[0027] The exemplary e-commerce platform 82 provides supply and demand matching of products and / or services to the general public with the help of Internet infrastructure. In the e-commerce platform 82, products and / or services are provided as commodity information. To simplify the description, the concepts of commodity, product, etc. are used in this application to refer to the products and / or services in the e-commerce platform 82, which may specifically be physical products, digital products, tickets, service subscriptions, other offline services, etc.

[0028] In reality, all entities can access the e-commerce platform 82 as users, use various online services provided by the e-commerce platform 82, and achieve the purpose of participating in the business activities achieved by the e-commerce platform 82. These entities can be natural persons, legal persons or social organizations, etc. Corresponding to the two types of entities, merchants and consumers in business activities, the e-commerce platform 82 has two types of users, merchant users and consumer users. All entities in the product distribution chain in business activities, including manufacturers, sellers, retailers, logistics providers, etc., can use online services in the e-commerce platform 82 as merchant users, while consumers in business activities, including real or potential consumers, can use online services in the e-commerce platform 82 as their corresponding consumer users. In actual business activities, the same entity can act as both a merchant user and a consumer user, and this should be understood flexibly.

[0029] The infrastructure used to deploy the e-commerce platform 82 mainly includes the backend architecture and frontend equipment. The backend architecture runs various online services through the service cluster, including middleware or frontend services for the platform, services for consumers, services for merchants, etc., to enrich and improve its service functions; the frontend equipment mainly covers the terminal devices used by users to access the e-commerce platform 82 as clients, including but not limited to various mobile terminals, personal computers, point-of-sale devices, etc. For example, merchant users can enter product information for their online stores through their terminal devices 80, or generate their product information using the interfaces opened by the e-commerce platform; consumer users can access the webpage of the online store implemented by the e-commerce platform 82 through their terminal devices 81, trigger the shopping process through the shopping buttons provided on the webpage, and call various online services provided by the e-commerce platform 82 during the shopping process, so as to achieve the purpose of shopping orders.

[0030] In some embodiments, the e-commerce platform 82 may be implemented by a processing facility including a processor and a memory, the processing facility storing a set of instructions that, when executed, cause the e-commerce platform 82 to perform the e-commerce and support functions involved in the present application. The processing facility may be part of a server, client, network infrastructure, mobile computing platform, cloud computing platform, fixed computing platform, or other computing platform, and provides electronic components, merchant equipment, payment gateways, application developers, marketing channels, transportation providers, customer equipment, point-of-sale equipment, etc. of the e-commerce platform 82.

[0031] The e-commerce platform 82 can be implemented as an online service such as cloud computing service, software as a service (SaaS), infrastructure as a service (IaaS), platform as a service (PaaS), desktop as a service (DaaS), hosted software as a service, mobile backend as a service (MBaaS), information technology management as a service (ITMaaS), etc. In some embodiments, the various functional components of the e-commerce platform 82 can be implemented to be suitable for operation on various platforms and operating systems. For example, for an online store, its administrator users enjoy the same or similar functions regardless of various embodiments such as iOS, Android, HomonyOS, or web pages.

[0032] The e-commerce platform 82 can realize the corresponding independent station for each merchant to run its corresponding online store, and provide the merchant with the corresponding business management engine instance for the merchant to establish, maintain and run one or more online stores in one or more independent stations. The business management engine instance can be used for content management, task automation and data management of one or more online stores, and various specific business processes of the online store can be configured through interfaces or built-in components to support the implementation of business activities. The independent station is the infrastructure of the e-commerce platform 82 with cross-border service functions. Merchants can maintain their online stores more centrally and autonomously based on the independent station. The independent station usually has a domain name and storage space dedicated to the merchant, and different independent stations are relatively independent. The e-commerce platform 82 can provide standardized or personalized technical support for a large number of independent stations, so that merchant users can customize their own business management engine instance and use this business management engine instance to maintain one or more online stores they own.

[0033] The online store can implement backend configuration and maintenance by having the merchant user log in to its business management engine instance as an administrator. With the support of various online services provided by the infrastructure of the e-commerce platform 82, the merchant user can configure various functions in its online store as an administrator, view various data, etc. For example, the merchant user can manage various aspects of its online store, such as viewing the online store's recent activities, updating the online store's product catalog, managing orders, recent visit activities, total order activities, etc.; the merchant user can also view more detailed information about the business and visitors to the merchant's online store by obtaining reports or metrics, such as showing a sales summary of the merchant's overall business, specific sales and participation data of active sales marketing channels, etc.

[0034] The e-commerce platform 82 may provide communication facilities and associated merchant interfaces for providing electronic communications and marketing, such as utilizing electronic message aggregation facilities to collect and analyze communication interactions between merchants, consumers, merchant devices, customer devices, point-of-sale devices, etc., aggregating and analyzing communications, such as for increasing the potential for providing product sales, etc. For example, a consumer may have a question related to a product, which may generate a dialogue between the consumer and the merchant (or an automated processor-based agent on behalf of the merchant), where the communication facility is responsible for interacting and providing the merchant with analysis on how to increase the probability of a sale.

[0035] See also Figure 2 A method for displaying product images of the present application, in a typical embodiment, comprises the following steps:

[0036] Step S1100: Determine whether a user is a target user based on the accumulated behavior data set of each user in the online store, and construct a target user pool corresponding to each target user;

[0037] The user is usually a commodity buyer of an online store. The behavior cumulative data set includes at least one behavior performed by the corresponding user on the commodity in the recent period and its cumulative execution times.

[0038] The behavior accumulation data set includes at least one behavior and its accumulated execution times.

[0039] The recent period may be any one of the past three days, the past week, the past half month, the past month, the past three months, the past six months, etc., and those skilled in the art may set it as needed, for example, within six months.

[0040] User behaviors can be captured by setting event listeners at the interaction points on the front-end interactive page. For example, a listening function can be bound to the click events of the product page and the purchase button. When the user performs these behaviors, the trigger function confirms and records these behaviors. Furthermore, a counter with an initial value of zero is set for each behavior, so that for each behavior, every time the user performs the behavior once, the cumulative execution count of the counter of the behavior is increased by one.

[0041] The various actions that users can perform on products include any one or more of single-item purchase, multiple-item purchase, adding to shopping cart, repeat purchase, return, positive review, negative review, search, share, favorite, direct click, search click, related click, customer service consultation, etc. The single-item purchase refers to a user purchasing a single item. The multiple-item purchase refers to a user purchasing multiple identical items. The direct click refers to a user directly clicking on the item without searching, adding to shopping cart, or purchasing the item first. The search click refers to a user clicking on one of the multiple items searched out. The related click refers to a user clicking on one of at least one item recommended and matched with the item after adding to shopping cart or purchasing the item first.

[0042] In one embodiment, the behavior cumulative data set of each user in the online store includes single-item purchases and their cumulative execution times, as well as multiple-item purchases and their cumulative execution times. For each user, it is determined whether any cumulative execution times in the behavior cumulative data set of the user is greater than zero. When it is greater than zero, it means that the user has recently purchased goods from the online store. It is not difficult to understand that such a user is a valuable user who can generate revenue for the online store, and thus, it is confirmed that the user belongs to the target user. Further, all target users are collected in a target user pool to obtain the target user pool.

[0043] Step S1200: for each target user in the target user pool, determine the interest products associated with the target behavior according to the behavior log of the target user, and construct an interest product list corresponding to the target behavior of the target user with each interest product;

[0044] The behavior log is usually generated by the background, and the generation of each log is triggered by the user performing any behavior on the product. The log includes the corresponding behavior and at least one product associated with it.

[0045] The target behavior can show that the user is interested in the corresponding product, and can be flexibly set by those skilled in the art according to the disclosure herein, such as any one or more of direct click, search click, related click, adding to shopping cart, single purchase, multiple purchase, etc.

[0046] Taking a target user as an example, the behavior logs of the target user are traversed to determine the logs whose behaviors belong to the target behaviors, the products in these logs are obtained as interest products, and the interest products are assembled to construct a product list as the interest product list of the target user corresponding to the target behavior.

[0047] Step S1300: constructing a picture preference vector of the target user according to the picture feature vector corresponding to the product display picture of each product of interest in the list of products of interest;

[0048] The product display image is a product image of the corresponding product that the target user sees. It is not difficult to understand that due to the limited page length, when displaying a product, generally one product image of at least one product image of the product is selected for display, and the displayed product image is the product display image.

[0049] The image representation model is pre-trained to a convergence state and acquires the ability to represent the image semantic features of the input image. The specific selection of the image representation model can be any readily available open source model such as Resnet series, DenseNet, EfficientNet, VGG, Mobilenet series, etc.

[0050] In one embodiment, the image representation model is implemented using a Resnet50 model. For each product display image of the product of interest, the product display image is input into the Resnet50 model after conventional preprocessing, and the image features corresponding to the image are gradually extracted by the stem block and four residual blocks of the Resnet50 model. The shallow stage extracts the basic features such as details and edges corresponding to the product in the image, and then extracts deep semantic features and high-level logical features in the deep stage, and finally obtains the image features of the vectorized representation of the image output by the last stage, i.e., the Res5 stage, i.e., the image feature vector. The preprocessing is required to adapt to the input of the Resnet50 model, and the product images input into the model are subjected to corresponding processing operations, such as image resizing, pixel normalization, data enhancement, channel rearrangement, etc. Those skilled in the art can flexibly implement it according to the disclosure herein.

[0051] The sum of the feature vectors of each image is calculated and then divided by the total number of these vectors to obtain the corresponding mean vector. The mean vector is divided by its Euclidean norm to obtain a normalized vector. Since this vector comprehensively considers each item of interest, that is, it can reflect the item preference of the target user at the vector representation level, this vector is used as the image preference vector. The Euclidean norm is the square root of the sum of the squares of the components in the corresponding vector, which can measure and represent the length of the vector in the Euclidean space.

[0052] Step S1400, determine the preference score between the image feature vector of each product image of the products in the online store and the image preference vector of each target user in the target user pool, integrate the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and use the product image with the best comprehensive score as the product display image of the corresponding product.

[0053] For each product in the online store, the picture representation model is used to determine the picture feature vector of each product picture of the product.

[0054] The cosine similarity algorithm is used to determine the cosine similarity between the image feature vector of each product image in the online store and the image preference vector of each target user in the target user pool. Because it can quantify the degree of preference of the corresponding target user for the product image, it is used as the preference score.

[0055] Furthermore, for each product image, the sum of the preference scores corresponding to the product image and each target user is calculated, and then divided by the total number of these target users to obtain the corresponding mean. Because it can quantify the degree of preference of all target users for the product image, it is used as a comprehensive score.

[0056] For each product in the online store, the product picture with the largest comprehensive score is selected from the comprehensive scores of the product pictures of the product as the product display picture of the product.

[0057] It can be known from the typical embodiments of the present application that the technical solution of the present application has many advantages, including but not limited to the following aspects:

[0058] This application analyzes the cumulative performance of various behaviors of each user in the online store, identifies each target user who meets the standard to construct a target user pool, and then for each target user, uses the behavior log to determine the various product display pictures that the target user is interested in, and constructs the target user's picture preference vector based on the picture feature vectors corresponding to these pictures. Further, the preference score between the picture feature vector of each product picture in the online store and the picture preference vector of each target user is determined, and the preference scores corresponding to all target users are integrated for each product picture to obtain a comprehensive score, and the product picture with the best comprehensive score is selected as the product display picture of the corresponding product. It can be seen that the target user group of the online store can be specifically refined, and the preference of each target user for the product picture can be accurately and quantitatively expressed, ensuring that the preference score determined according to the quantitative expression corresponding to the picture preference and the product picture is accurate and reliable. On this basis, for each product picture, the comprehensive score is obtained by integrating the preferences of all target users for the product picture, and the one with the highest comprehensive score is selected as the product display picture, ensuring that when it displays the product, it can attract the attention of the target user, enhance the user's shopping experience, and help promote the completion of product transactions.

[0059] In a further embodiment, step S1100, judging whether a user belongs to a target user based on the accumulated data set of behaviors of each user in the online store, and constructing a target user pool corresponding to each target user, comprises the following steps:

[0060] Step S1110: for each user in the online store, determine the value scores of various behaviors corresponding to the user's behavior accumulation data set;

[0061] The value weights of the various behaviors can be pre-set, and the value weights quantitatively represent the importance of the corresponding behaviors for generating revenue for the online store. Those skilled in the art can set them according to the disclosure herein and in combination with business needs. The behavior weights of the various behaviors described in the exemplary examples are: single purchase: 0.5, multiple purchase: 1, add to shopping cart: 0.3, repeat purchase: 0.7, return: -2, positive review: 0.8, negative review: -1, search: 0.1, share: 0.4, favorite: 0.2, direct click: 0.3, search click: 0.2, related click: 0.3, customer service consultation: 0.3. The business needs corresponding to the weights of each behavior are: single purchase: a direct source of income, multiple purchases: usually means higher sales than single purchases, add to cart: a prelude to purchase behavior, but not necessarily leading to sales, repeat purchase: indicates user loyalty to the brand, which has a positive impact on long-term income, returns: have a high negative impact on the store, positive reviews: can improve the store's credibility, attract new customers and build word-of-mouth, negative reviews: damage the store's reputation and reduce sales, search: indicates that the user has a vague interest in the product, share: increase the exposure of the product, help attract potential customers and / or new users, collect: indicates that the user is slightly interested in the product, direct click: indicates that the user has a clear interest in the product, search click: indicates that the user has a clear interest in the product, related click: indicates that the user is interested in related products, customer service consultation: indicates that the user may have the intention to buy and needs further communication.

[0062] In one embodiment, the behavior cumulative data set of each user in the online store includes single purchase, multiple purchase, adding to shopping cart, repeat purchase, return, positive review, negative review, search, share, favorite, direct click, search click, related click, customer service consultation, and the cumulative execution times of each behavior. For each user, the cumulative execution times of each behavior in the user's behavior cumulative data set are multiplied by the value weight corresponding to each behavior to obtain the value score corresponding to each behavior.

[0063] Step S1120: Determine the value score comprehensive value score of the user according to the various value scores, and when the value score comprehensive value score meets a preset condition, confirm the user as a target user.

[0064] The sum of the value scores of various behaviors in the user's behavior cumulative data set is obtained as a comprehensive value score, which quantitatively represents the value contribution of the user to the online store's revenue generation. When the comprehensive value score is higher than the preset threshold, it means that the comprehensive value score at this time can generate sufficient economic benefits for the online store. It is not difficult to understand that such a user is a valuable user who can generate revenue for the online store, and thus, the user is confirmed to be a target user. The preset threshold can be flexibly set by a person skilled in the art according to the disclosure herein.

[0065] In this embodiment, a comprehensive value score is determined for each user in the online store, and target users are selected based on the score. In this process, the value scores of various behaviors are comprehensively considered to ensure the accuracy and reliability of the comprehensive value score obtained subsequently.

[0066] In a further embodiment, after step S1400, determining the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, the following steps are included:

[0067] Step S2400: for each target user, determine the maximum value of each preference score corresponding to the target user and each product image of the product;

[0068] Taking a single target user as an example, the maximum preference score, ie, the maximum value, is screened out from the preference scores corresponding to the target user and each product image of the product.

[0069] Step S2410: Use the product image corresponding to the maximum value as the product display image used to display the product to the target user.

[0070] It is not difficult to understand that the target user is more interested in the product picture corresponding to the maximum value than other product pictures of the product. Therefore, the product picture corresponding to the maximum value is used as the product display picture of the product, so that when the target user browses the product, the product display picture is displayed to the target user.

[0071] In this embodiment, for the target user, the product display pictures of the products that the target user is most interested in are selected, so as to attract the attention of the target user to the greatest extent, promote the transaction of the corresponding products, and hopefully improve the conversion rate of the products.

[0072] In a further embodiment, step S1400, determining the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, fusing the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and using the product image with the best comprehensive score as the product display image of the corresponding product, comprises the following steps:

[0073] Step S1500: for each product in the online store, construct a training sample using each product image of the product;

[0074] Obtain each product image of each product in the online store, and for each product, use each product image of the product to form a single training sample.

[0075] Step S1510: for each training sample, a supervisory label of the training sample is marked accordingly according to whether each product image in the training sample belongs to a product display image;

[0076] Taking a single training sample as an example, in one embodiment, the product image in the training sample that belongs to the product display picture is marked as 1, indicating that the possibility that the product image belongs to the product display picture is 1, and the remaining product images are marked as 0, respectively indicating that the possibility that the corresponding product image belongs to the product display picture is 0. The various labeled values ​​are concatenated to form the supervision label of the training sample. For ease of understanding, an exemplary example is given, in which the product contains 4 product images, the second product image belongs to the product display picture, and the remaining product images do not belong to the product display picture. The supervision label of the corresponding training sample is [0,1,0,0].

[0077] Step S1520: construct a training set using all training samples and their supervision labels, and use the training set to train the image optimization model until it converges, so that it can learn the ability to select product display images from multiple product images of input products.

[0078] A single training sample in the training set and its supervision label are obtained. The image representation layer in the image optimization model extracts the image feature vector of each product image in the training sample. The classification layer classifies and maps the image feature vectors to determine the predicted probability that each product image belongs to the product display image. The sum of these predicted probabilities is 1, and the prediction results are concatenated to form the predicted results.

[0079] The cross entropy loss function is used to calculate the loss value corresponding to the cross entropy loss of the prediction result based on the supervisory label of the training sample. When the loss value reaches the preset threshold, it indicates that the picture optimization model has been trained to a convergence state, so that the training of the picture optimization model can be terminated; when the loss value does not reach the preset threshold, it indicates that the picture optimization model has not converged, so the picture optimization model is gradient updated according to the loss value, and the weight parameters of each corresponding link in the picture optimization model are usually corrected by back propagation to make the picture optimization model further close to convergence, and then continue to call other training samples in the training set and their supervisory labels to perform iterative training on the picture optimization model until the picture optimization model is trained to a convergence state. The preset threshold can be set as needed by those skilled in the art according to the disclosure herein.

[0080] The image representation layer in the image optimization model may be selected from the Resnet series, DenseNet, EfficientNet, VGG, Mobilenet series, etc., and the classification layer may be selected from the MLP and / or fully connected layer, which can be flexibly implemented by those skilled in the art.

[0081] In this embodiment, training is implemented to obtain a picture selection model that can select a picture corresponding to a product display picture from multiple product pictures of an input product, ensuring that its reasoning results are reliable, accurate and highly interpretable.

[0082] In a further embodiment, after step S1520, constructing a training set with all training samples and their supervision labels, and training the image optimization model so that it learns the ability to select a product display image from multiple input product images, the following steps are included:

[0083] Step S1530: According to the business category set of the online store and the business category sets of multiple secondary online stores, determine secondary online stores similar to the online store as target online stores, wherein the user volume of each secondary online store is lower than the user volume of the online store;

[0084] It is not difficult to understand that each product in an online store has its own product category. The product category refers to a way of classifying the products in an online store according to their attributes, uses, functions or consumer needs.

[0085] First, the accumulated behavioral data set of all users of the online store and the accumulated behavioral data set of all users of other online stores are obtained. For each online store, all target users of the online store are determined, and the total number of these target users is used as the user volume of the online store. Then, the difference between the user volume of the online store and the user volume corresponding to each other online store is calculated, and each other online store corresponding to the difference exceeding the first preset threshold is used as a secondary online store. The first preset threshold can be set by a person skilled in the art as needed.

[0086] Secondly, obtain the business category set corresponding to the online store and each secondary online store, and the business category set includes at least one commodity category operated by the corresponding online store. The business category set can usually be provided by the operator of the online store. If not provided, the category of each commodity in the online store can be determined by adopting a pre-defined commodity classification model. Specifically, taking a single commodity as an example, the commodity description text and commodity image of the commodity are input into the commodity classification model, and the deep semantic information of the commodity image is extracted from the image representation layer in the commodity classification model, and the deep semantic information is vectorized to obtain the corresponding image feature vector. The deep semantic information of the commodity description text is extracted from the text representation layer in the commodity classification model, and the deep semantic information is vectorized to obtain the corresponding text feature vector. The image feature vector and the text feature vector are fused by the image-text fusion layer in the commodity classification model to obtain the image-text fusion vector. The commodity classification model classifies and maps the image-text fusion vector to confirm the commodity category with the highest classification probability. Further, the commodity categories of all commodities are grouped into the commodity category set, and after deduplication, the operating category set of the online store is obtained.

[0087] The commodity classification model is pre-trained to convergence, and acquires the ability to determine the commodity category of the commodity based on the commodity description text and commodity image of the input commodity. The technicians in this field can flexibly implement the training of the commodity classification model according to the above reasoning process. The commodity classification model includes an image representation layer, a text representation layer, an image-text fusion layer, and a classification layer. The image-text fusion layer is connected to the image representation layer and the text representation layer, and the classification layer is connected to the image-text fusion layer. The recommended model for the image representation layer is the Resnet model, and any other model such as EfficientNet, DenseNet, MobileNet, LeNet, ViT, VGG, etc. can also be used. The recommended model for the text representation layer is the mBERT (Multilingual BERT) model, and any other model such as XLM (Cross-lingual Language Model), mT5 (multilingual T5), XLM-R (XLM-RoBERTa), etc. can also be used, which is suitable for multiple languages. The image-text fusion layer can be implemented by weighted averaging, gating, etc., and the technicians in this field can choose one to implement as needed. The classification layer can be selected as MLP and / or fully connected layer, which can be flexibly implemented by those skilled in the art. The output results of the image representation layer and the text representation layer are mapped to the same semantic space. The product description text includes all texts describing the corresponding product, such as product title, product details text, etc.

[0088] Further, the intersection and union between the business category set of the online store and the business category sets of each secondary online store are confirmed, and when the number of commodity categories that differs between the union and the intersection is less than a second preset threshold, it is confirmed that the corresponding online store and the secondary online store are similar, and each similar secondary online store is used as a target online store. The second preset threshold can be set by a person skilled in the art as needed.

[0089] Step S1531: for each product in the target online store, use the image optimization model trained to a convergent state to determine a product display image from multiple product images of the product.

[0090] The image optimization model learns the ability to select a product display image from multiple product images of the input product. Thus, for each product in the target online store, multiple product images of the product are input into the image optimization model, and the product display image of the product can be determined.

[0091] In this embodiment, by determining the target online stores that are similar to the online stores and have a smaller user base, and then determining the product display pictures of multiple product pictures of each product in each target online store, it can be seen that the experience of user group preference for product pictures obtained from the larger user base of the online stores can be utilized and applied to the target stores, so that the product display effect of the target online stores is better, especially for the newly opened target online stores with no user base or very low user base, so as to achieve forward-looking and high-quality product display effects.

[0092] In a further embodiment, step S1400, determining the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, fusing the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and using the product image with the best comprehensive score as the product display image of the corresponding product, comprises the following steps:

[0093] Step S1600: for each product in the online store, the attribute text set of the product display image of the product is used as the optimal attribute text set, and the attribute text sets of the remaining product images of the product are used as suboptimal attribute text sets, and the optimal attribute text set and the suboptimal attribute text sets form a comparison instance;

[0094] The attribute text set of the product image includes descriptive text of at least one attribute of the corresponding product, which may be any one or more of resolution, brightness, contrast, product perspective description text (for example: front, back, left, right), objects that set off the product (for example: pants, tops, female models, male models, computers, table mats), display purposes (for example: showing the ultra-thin design of a mobile phone, showing the comfortable fit of sports shoes, showing the one-key operation interface of a coffee machine), etc.

[0095] Since the product images of commodities are generally obtained by photographing and editing the commodities by the operators of the corresponding online stores, the operators may further provide descriptive texts of various attributes of the product images.

[0096] Step S1610: embed each control instance into a preset prompt template, obtain corresponding prompt text and input it into the large language model to guide the model to summarize the characteristic attribute text describing the product image used to display the product.

[0097] The large language model is suitable for text processing in the field of NLP. It is pre-trained with an extremely large corpus until convergence, acquires the ability to generate human language, and has a certain degree of accurate text semantic understanding and logical reasoning ability. The large language model includes OpenAI, Falcon, Chinchilla, PaLM, LLaMA 2, text-embedding-ada-002, etc. Those skilled in the art can choose one to implement as needed.

[0098] The preset prompt template includes a task description and various comparison examples to be embedded. Those skilled in the art may refer to the following disclosure to flexibly preset prompt templates. For example, the task description in the preset prompt template is: "Analyze according to the following given comparison examples, for each comparison example, find the difference between the attribute text set of the product display image of the same product and the attribute text set of the other product images in the comparison example, and summarize the commonalities between the differences found by each comparison example, summarize these commonalities, derive features suitable for use as the main image of the product for product display, and output text describing these features". The various comparison examples to be embedded in the preset prompt template: {

Comparison Example 1

[0099] - Attribute text set of product display pictures: {$optimal attribute text set$};

[0100] - The attribute text sets of the remaining product images: {$suboptimal attribute text set$}, {$suboptimal attribute text set$}…, {$suboptimal attribute text set$}; ...

[0102]

Comparative Example N

[0103] - Attribute text set of product display pictures: {$optimal attribute text set$};

[0104] -The attribute text sets of the remaining product images: {$suboptimal attribute text set$}, {$suboptimal attribute text set$}…, {$suboptimal attribute text set$};}.

[0105] A description text output by a large language model according to the prompt text is obtained as the feature attribute text.

[0106] In this embodiment, by constructing a reference instance corresponding to each product based on the attribute text set of the product display image of each product in the online store and the attribute text set of the remaining product images of the product, the large language model is guided to determine the text describing the characteristics of the product display image based on each reference instance. It can be seen that the characteristics of the product images preferred by the user group can be mined, and these characteristics can be visualized to construct corresponding description texts, so that they can be provided to the operators of the online store to perform corresponding shooting and editing of the products, and obtain product images that attract more user attention.

[0107] In a further embodiment, step S1400, determining the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, fusing the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and using the product image with the best comprehensive score as the product display image of the corresponding product, comprises the following steps:

[0108] Step S1700: In response to a timed trigger event, obtain the exposure effectiveness score of each product in the online store, screen out multiple exposure effectiveness scores that meet preset conditions, and calculate the corresponding average;

[0109] An event is triggered every predetermined time, and the server responds to the event. Specifically, the server responds to the event by obtaining the sum of the number of clicks after exposure, the number of exposures, and the number of single and multiple purchases after clicks of the corresponding product within the predetermined time, and dividing the number of clicks after exposure by the number of exposures to obtain the click-through rate, and dividing the sum of the number of single and multiple purchases after clicks by the number of clicks after exposure to obtain the conversion rate, and then multiplying the click-through rate and conversion rate by their respective weights and adding them together to obtain the exposure effectiveness score. The weights corresponding to the click-through rate and conversion rate can be set by technicians in this field as needed, such as 0.3 and 0.7. The exposure refers to the display of the product to any user of the online store. The preset duration can be set by technicians in this field as needed.

[0110] Furthermore, the exposure effectiveness scores are sorted from large to small, and the top N 1 and the last N 2 The corresponding exposure effectiveness scores are summed up and then divided by the total number of these exposure effectiveness scores to obtain the corresponding mean. 1 and N 2 All can be set by those skilled in the art as required.

[0111] Step S1710: When the mean value satisfies a preset condition, a product display image of each product in the online store is re-determined.

[0112] In one embodiment, the mean value is subtracted from the mean value obtained in the last response to the timing trigger event to obtain the corresponding difference. When the difference value is lower than the first preset threshold, it is considered to meet the preset condition. In addition, it indicates that the user group's preference for product images has changed significantly at this time. Therefore, based on the typical embodiment of the present application, the product display image of each product in the online store is re-determined. The first preset threshold value can be set as needed by those skilled in the art based on the disclosure herein.

[0113] In another embodiment, when the difference is lower than a second preset threshold, the preset condition is considered to be met, and in addition, it indicates that the user group's preference for product images has changed significantly at this time, and thus, based on the typical embodiment of the present application, the product display image of each product in the online store is re-determined. The second preset threshold can be set as needed by those skilled in the art based on the disclosure herein.

[0114] In this embodiment, the selection of product display pictures in the online store is updated on demand at a regular interval to ensure the attractiveness and conversion rate of the products, so as to be able to respond to and adapt to changes in the user group's preference for product pictures in a timely manner.

[0115] See also Figure 3 , a product image display device provided to meet one of the purposes of the present application is a functional embodiment of the product image display method of the present application. On the other hand, the device is a product image display device provided to meet one of the purposes of the present application, including a user pool construction module 1100, an interest mining module 1200, a preference expression module 1300 and a display preference module 1400, wherein the user pool construction module 1100 is used to determine whether a user belongs to a target user based on the accumulated data set of the behavior of each user in the online store, and to construct a target user pool corresponding to each target user; the interest mining module 1200 is used to determine whether a user belongs to a target user based on the accumulated data set of the behavior of each user in the target user pool, and to construct a target user pool corresponding to each target user; The log determines the interest items associated with the target behavior, and constructs the interest item list of the target user corresponding to the target behavior with each interest item; the preference representation module 1300 is used to construct the image preference vector of the target user according to the image feature vector corresponding to the product display image of each interest item in the interest item list; the display optimization module 1400 is used to determine the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, and obtains a comprehensive score by integrating the corresponding preference scores of all target users for each product image, and uses the product image with the best comprehensive score as the product display image of the corresponding product.

[0116] In a further embodiment, the user pool construction module 1100 includes: a score determination submodule, which is used to determine, for each user in the online store, the value scores of various behaviors corresponding to the user's behavior cumulative data set; a target determination submodule, which is used to determine the user's value score comprehensive value score based on the various value scores, and when the value score comprehensive value score meets the preset conditions, confirm that the user is a target user.

[0117] In a further embodiment, after the display preference module 1400, it includes: a score determination submodule, which is used to determine, for each target user, the maximum value of the preference scores corresponding to the target user and each product image of the product; and a picture screening submodule, which is used to use the product image corresponding to the maximum value as the product display image used to display the product to the target user.

[0118] In a further embodiment, after the display optimization module 1400, it includes: a sample construction submodule, which is used to construct a training sample with each product image of each product in the online store; a sample labeling submodule, which is used to label the supervision label of each training sample according to whether each product image in the training sample belongs to the product display image; a model training submodule, which is used to construct a training set with all training samples and their supervision labels, and use the training set to train the picture optimization model to a convergence state, so that it learns the ability to select a product display image from multiple product images of the input product.

[0119] In a further embodiment, after the model training submodule, it includes: a store determination submodule, which is used to determine a secondary online store similar to the online store as a target online store based on the business category set of the online store and the business category sets of multiple secondary online stores, and the user volume of each secondary online store is lower than the user volume of the online store; a display selection submodule, which is used to determine, for each product in the target online store, a product display image from multiple product images of the product using the picture selection model trained to a convergence state.

[0120] In a further embodiment, after the display optimization module 1400, it includes: an actual construction submodule, which is used to, for each product in the online store, use the attribute text set of the product display image of the product as the optimal attribute text set, and use the attribute text sets of the remaining product images of the product as suboptimal attribute text sets, and form a comparison instance with the optimal attribute text set and each suboptimal attribute text set; a feature induction submodule, which is used to embed each comparison instance into a preset prompt template, obtain the corresponding prompt text and input it into the large language model to guide the model to summarize the feature attribute text that describes the product image used to display the product.

[0121] In a further embodiment, after the display optimization module 1400, it includes: an event response submodule, which is used to respond to a timed trigger event, obtain the exposure effectiveness score of each product in the online store, screen out multiple exposure effectiveness scores that meet preset conditions, and calculate the corresponding average; a display re-selection submodule, which is used to re-determine the product display picture of each product in the online store when the average meets the preset conditions.

[0122] In order to solve the above technical problems, the present application also provides an electronic device. Figure 4 As shown, a schematic diagram of the internal structure of an electronic device. The electronic device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the electronic device stores an operating system, a database, and computer-readable instructions. The database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor can implement the product image display method. The processor of the electronic device is used to provide computing and control capabilities to support the operation of the entire electronic device. The memory of the electronic device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute the product image display method described in this application. The network interface of the electronic device is used to connect and communicate with a terminal. Those skilled in the art will understand that Figure 4 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0123] In this execution mode, the processor is used to execute Figure 3 The memory stores the program code and various data required to execute the above modules or modules. The network interface is used to transmit data between user terminals or servers. The memory in this execution mode stores the program code and data required to execute all modules / modules in the product image display device of this application, and the server can call the program code and data of the server to execute the functions of all modules.

[0124] The present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the method of any embodiment of the present application.

[0125] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a computer-readable storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0126] In summary, the present application can select the product display pictures that attract users, and can timely update the selection mechanism to adapt to the changes in the preferences of user groups for product pictures.

[0127] It will be understood by those skilled in the art that the various operations, methods, steps, measures, and schemes in the processes discussed in this application may be alternated, changed, combined, or deleted. Furthermore, other steps, measures, and schemes in the various operations, methods, and processes discussed in this application may also be alternated, changed, rearranged, decomposed, combined, or deleted. Furthermore, the steps, measures, and schemes in the various operations, methods, and processes in the prior art that are open source and in this application may also be alternated, changed, rearranged, decomposed, combined, or deleted.

[0128] The above is only a partial implementation of the present application. It should be pointed out that ordinary technicians in this technical field can make several improvements and modifications without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for displaying a product image, characterized in that: The steps include: Based on the accumulated data set of each user's behavior in the online store, determine whether the user is a target user and construct a target user pool corresponding to each target user; For each target user in the target user pool, determine the interest products associated with the target behavior according to the behavior log of the target user, and construct an interest product list of the target user corresponding to the target behavior with each interest product; constructing a picture preference vector of the target user according to the picture feature vector corresponding to the product display picture of each product of interest in the list of products of interest; Determine the preference score between the image feature vector of each product image of the products in the online store and the image preference vector of each target user in the target user pool, integrate the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and use the product image with the best comprehensive score as the product display image of the corresponding product.

2. The method for displaying product images according to claim 1, characterized in that: Based on the accumulated data set of each user's behavior in the online store, determine whether the user is a target user, and construct a target user pool corresponding to each target user, including the following steps: For each user in the online store, determine the value scores of various behaviors corresponding to the user's behavior cumulative data set; A value score comprehensive value score of the user is determined according to the various value scores, and when the value score comprehensive value score meets a preset condition, the user is confirmed as a target user.

3. The method for displaying product images according to claim 1, characterized in that: After determining the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, the following steps are included: For each target user, determining the maximum value of each preference score corresponding to the target user and each product image of the product; The product image corresponding to the maximum value is used as the product display image used to display the product to the target user.

4. The method for displaying product images according to claim 1, characterized in that: Determine the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, merge the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and use the product image with the best comprehensive score as the product display image of the corresponding product, including the following steps: For each product in the online store, construct training samples using each product image of the product; For each training sample, the supervision label of the training sample is marked according to whether each product image in the training sample belongs to a product display image; A training set is constructed with all training samples and their supervision labels, and the image optimization model is trained with the training set until convergence, so that it can learn the ability to select product display images from multiple product images of the input product.

5. The method for displaying product images according to claim 4, characterized in that: After constructing a training set with all training samples and their supervision labels, and training the image selection model to enable it to learn the ability to select product display images from multiple input product images, the following steps are included: According to the business category set of the online store and the business category sets of multiple secondary online stores, a secondary online store similar to the online store is determined as a target online store, wherein the user volume of each secondary online store is lower than the user volume of the online store; For each product in the target online store, the image optimization model trained to a convergent state is used to determine a product display image from multiple product images of the product.

6. The method for displaying product images according to claim 1, characterized in that: Determine the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, merge the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and use the product image with the best comprehensive score as the product display image of the corresponding product, including the following steps: For each product in the online store, the attribute text set of the product display image of the product is used as the optimal attribute text set, and the attribute text sets of the remaining product images of the product are used as suboptimal attribute text sets, and the optimal attribute text set and the suboptimal attribute text sets form a comparison instance; Each control instance is embedded into a preset prompt template to obtain the corresponding prompt text and input it into the large language model to guide the model to summarize the characteristic attribute text describing the product image used to display the product.

7. The method for displaying product images according to claim 1, characterized in that: Determine the preference score between the image feature vector of each product image of the product in the online store and the image preference vector of each target user in the target user pool, merge the corresponding preference scores of all target users for each product image to obtain a comprehensive score, and use the product image with the best comprehensive score as the product display image of the corresponding product, including the following steps: In response to a timed trigger event, an exposure effectiveness score of each product in the online store is obtained, multiple exposure effectiveness scores that meet preset conditions are screened out, and a corresponding average is calculated; When the mean value satisfies a preset condition, a product display image of each product in the online store is re-determined.

8. A commodity picture display device, characterized in that: It includes: The user pool construction module is used to determine whether a user belongs to a target user based on the accumulated data set of each user's behavior in the online store, and to construct a target user pool corresponding to each target user; An interest mining module is used to determine, for each target user in the target user pool, interest commodities associated with the target behavior according to the behavior log of the target user, and construct an interest commodity list of the target user corresponding to the target behavior with each interest commodity; A preference representation module, used to construct a picture preference vector of the target user according to a picture feature vector corresponding to a product display picture of each product of interest in the list of products of interest; The display optimization module is used to determine the preference score between the image feature vector of each product image of the products in the online store and the image preference vector of each target user in the target user pool, and to obtain a comprehensive score by integrating the corresponding preference scores of all target users for each product image, and to use the product image with the best comprehensive score as the product display image of the corresponding product.

9. An electronic device, comprising a central processing unit and a memory, characterized in that: The central processing unit is used to call and run the computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 7.

10. A non-volatile storage medium, characterized in that: It stores a computer program implemented according to the method described in any one of claims 1 to 7 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images