Commodity information providing model construction method and commodity information providing method

Abstract

The embodiment of the invention discloses a commodity information providing model construction method and device, a model training data obtaining method and device, a commodity information providing method and device and AR glasses. And the commodity information providing model is deployed in AR glasses provided by a physical store, and shopping guide service is provided for the customer user in the store through the AR glasses. The method comprises the steps that multiple sets of training data sequentially collected according to the browsing situation of a sample user in a shop in the whole process that the sample user wears AR glasses are obtained, and each set of training data comprises position information of the sample user, identification information of a checked commodity and information of the interest degree of the sample user in the checked commodity; constructing an initial model for providing a shopping guide service; and carrying out model training on the initial model by taking a plurality of groups of training data sequentially collected in the training sample as model input to obtain a commodity information providing model, so that the commodity information providing model has the capability of providing a browsing route recommendation service for the commodities displayed in the physical store.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a method and device for constructing a product information providing model, a method and device for obtaining model training data, a method and device for providing product information, and AR glasses. Background Art

[0002] Under the impact of the digital wave, the offline retail industry is facing unprecedented challenges. The convenience and price advantages of e-commerce platforms have made consumers more and more inclined to shop online, resulting in a continuous decline in the customer flow of physical stores. In such a market environment, offline retailers need to continue to innovate and optimize, and only by providing high-quality products and services, improving customer experience, and increasing the conversion rate of goods transactions can they gain a firm foothold in the fierce market competition. Therefore, how to reduce store costs and increase the conversion rate of goods transactions has become the key to the survival and development of offline retailers. Summary of the invention

[0003] The present application provides a method and device for constructing a product information provision model, a method and device for obtaining model training data, a method and device for providing product information, and AR glasses, which improve user shopping efficiency and increase product transaction conversion rate by providing more intelligent shopping guide services to consumer users who come to the store.

[0004] This application provides the following solutions:

[0005] A method for constructing a product information provision model, wherein the product information provision model is deployed in augmented reality (AR) glasses provided by a physical store, and shopping guide services are provided to in-store consumers through the AR glasses. The method comprises:

[0006] Obtain multiple training samples, each training sample includes multiple sets of training data collected sequentially during the whole process of the sample user wearing the AR glasses according to the browsing situation of the sample user in the physical store, and each set of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product;

[0007] Build an initial model for providing shopping guide services;

[0008] A plurality of sets of training data collected in the order included in the training samples are used as inputs of the initial model, and the initial model is trained to obtain the product information providing model, so that the product information providing model has the ability to provide browsing route recommendation services for the products displayed in the physical store, and the browsing route is generated based on the browsing movements of the sample users in the physical store determined by the location information, and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information.

[0009] The interest level information includes at least one of the following: the length of time the sample user stays at the viewed product, the number of times the sample user performs voice interaction with the viewed product, the number of times the sample user performs physical interaction with the viewed product, and whether an operation behavior indicating purchase intention is performed.

[0010] In the process of training the initial model using the training samples, the popular products of the physical store determined according to the interest level information are obtained as the products that meet the user's general shopping preferences, and then the browsing route of the physical store is generated based on the browsing movements of the sample users in the physical store determined according to the location information collected in the sequence.

[0011] The operation behavior indicating the purchase intention is that the sample user performs a purchase operation on the viewed product.

[0012] In the process of training the initial model using the training samples, the product associations between multiple purchased products are also obtained, and the products with product associations are used as the products that meet the user's general shopping preferences, so as to generate a browsing route of the physical store based on the products with product associations, the popular products and the browsing routes.

[0013] Based on the commodity association, the commodity information providing model is also enabled to have the ability to identify matching commodities associated with the commodity currently purchased by the consumer user and provide commodity information of the matching commodities.

[0014] Among them, the product information providing model also has the ability to provide the identification information of the popular products and / or the identification information of the products with product association to the manager user associated with the physical store, so that the manager user can adjust the product display method of the physical store.

[0015] The sample user performs a purchase operation on the viewed product, and the purchase operation is realized after a shopping guide information is pushed to the sample user, the shopping guide information is pre-configured by the physical store according to the user's focus associated with the viewed product, and the type of the shopping guide information includes at least product knowledge answer information and / or product store promotion information;

[0016] The training data also includes: the push rounds for pushing shopping guide information to the sample users, and the types and contents of the shopping guide information pushed in different push rounds;

[0017] The method also includes: using the push rounds and the types and contents of the shopping guide information pushed in different push rounds as inputs of the initial model, performing model training on the initial model, and enabling the product information push model to obtain the degree of influence of different types of shopping guide information on shopping decisions, and determine the shopping guide information push method corresponding to the target product currently viewed by the consumer user according to the degree of influence, wherein the shopping guide information push method includes the type of shopping guide information associated with the target product and the push priority of different types of shopping guide information.

[0018] The training data also includes: the consulting questions input by the sample users during voice interaction with the viewed products,

[0019] The user attention points associated with the viewed product are obtained in the following manner:

[0020] The consulting question is used as the input of the initial model to perform model training on the initial model, and the product information providing model is also enabled to have the ability to determine the user's focus points associated with the viewed product based on the consulting question.

[0021] A method for obtaining model training data, applied to augmented reality (AR) glasses provided by physical stores, the method comprising:

[0022] During the whole process of the sample user wearing the AR glasses and browsing the displayed commodities in the physical store, multiple sets of training data related to the viewed commodities are sequentially collected, and the multiple sets of training data are used as a training sample for model training to obtain a commodity information provision model for providing shopping guide services to in-store consumer users;

[0023] The training data includes: location information of the sample user, identification information of the viewed commodity, and information on the interest level of the sample user in the viewed commodity;

[0024] The location information is obtained by positioning the AR glasses through multiple sensors deployed in the physical store, and the location information in the multiple sets of training data collected sequentially is used to represent the browsing route of the sample user in the physical store;

[0025] The identification information of the viewed product and the interest level information are determined based on information collected by the AR glasses, where the information collected by the AR glasses includes image information within the field of view of the AR glasses and / or voice interaction information between the AR glasses and the sample user.

[0026] A method for providing product information is applied to augmented reality (AR) glasses provided by a physical store, and provides shopping guide services to in-store consumers based on a product information provision model deployed in the AR glasses. The method comprises:

[0027] When it is determined that the consumer user has a store navigation requirement, obtaining the current target location of the consumer user;

[0028] The product information providing model is called, and the model is trained based on multiple training samples to obtain the ability to provide browsing route recommendation services for the products displayed in the physical store; wherein each training sample includes multiple sets of training data collected sequentially according to the browsing situation of the sample user in the physical store during the whole process of the sample user wearing the AR glasses, and each set of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product;

[0029] The target location is used as a model input to obtain a target navigation route output by the model based on the browsing route, wherein the target navigation route takes the target location as a starting point and is used to guide the consumer user to browse the merchandise displayed on the target navigation route through the AR glasses worn by the consumer user; wherein the browsing route is generated based on the browsing movements of the sample user in the physical store determined by the location information, and the merchandise in the physical store that can satisfy the user's general shopping preferences determined by the interest level information.

[0030] After the consumer user enters the store and wears the AR glasses, it is determined that the consumer user has a store navigation need.

[0031] The obtaining of the target location where the consumer user is currently located includes: obtaining the starting point of the browsing route generated by the model training as the target location;

[0032] Then, after the starting point of the browsing route is input into the model, the target navigation route output by the model is obtained, and the target navigation route is the browsing route.

[0033] When receiving a navigation request submitted by the consumer user through the AR glasses worn during product browsing, it is determined that the consumer user has a store navigation demand, then

[0034] The obtaining of the target location where the consumer user is currently located comprises: obtaining the target location by positioning the AR glasses;

[0035] Then, after the target location is input into the model, a target navigation route output by the model is obtained, and the target navigation route is part or all of the browsing route.

[0036] A device for constructing a product information provision model, wherein the product information provision model is deployed in augmented reality (AR) glasses provided by a physical store, and shopping guide services are provided to consumers who visit the store through the AR glasses. The device comprises:

[0037] A sample obtaining unit, configured to obtain a plurality of training samples, each of which includes a plurality of sets of training data sequentially collected during the whole process of a sample user wearing the AR glasses according to the browsing situation of the sample user in the physical store, each set of training data including the location information of the sample user, the identification information of the viewed commodity, and the interest level information of the sample user in the viewed commodity;

[0038] A model building unit, used for building an initial model for providing shopping guide services;

[0039] A model training unit is used to use multiple groups of training data collected in the order included in the training samples as inputs of the initial model, perform model training on the initial model, and obtain the product information providing model, so that the product information providing model has the ability to provide browsing route recommendation services for the products displayed in the physical store, and the browsing route is generated based on the browsing movement of the sample user in the physical store determined by the location information, and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information.

[0040] A device for obtaining model training data, applied to augmented reality (AR) glasses provided by a physical store, comprising:

[0041] A data collection unit is used to sequentially collect multiple sets of training data related to the viewed products during the whole process of the sample user wearing the AR glasses to browse the displayed products in the physical store, and the multiple sets of training data are used as a training sample for model training to obtain a product information provision model for providing shopping guide services to in-store consumer users;

[0042] The training data includes: location information of the sample user, identification information of the viewed commodity, and information on the interest level of the sample user in the viewed commodity;

[0043] The location information is obtained by positioning the AR glasses through multiple sensors deployed in the physical store, and the location information in the multiple sets of training data collected sequentially is used to represent the browsing route of the sample user in the physical store;

[0044] The identification information of the viewed product and the interest level information are determined based on information collected by the AR glasses, where the information collected by the AR glasses includes image information within the field of view of the AR glasses and / or voice interaction information between the AR glasses and the sample user.

[0045] A commodity information providing device is applied to augmented reality (AR) glasses provided by a physical store, and provides shopping guide services to consumers who come to the store based on a commodity information providing model deployed in the AR glasses. The device includes:

[0046] A location obtaining unit, configured to obtain a current target location of the consumer user when it is determined that the consumer user has a store navigation requirement;

[0047] A model calling unit, used to call the product information providing model, wherein the model is trained based on a plurality of training samples to obtain the ability to provide browsing route recommendation services for the products displayed in the physical store; wherein each training sample includes a plurality of sets of training data sequentially collected during the whole process of the sample user wearing the AR glasses according to the browsing situation of the sample user in the physical store, and each set of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product;

[0048] A navigation route obtaining unit is used to take the target location as a model input, and obtain a target navigation route output by the model based on the browsing route, wherein the target navigation route takes the target location as a starting point, and is used to guide the consumer user to browse the commodities displayed on the target navigation route through the AR glasses worn by the consumer user; wherein the browsing route is generated based on the browsing movement of the sample user in the physical store determined by the location information, and the commodities in the physical store that can meet the user's general shopping preferences determined by the interest level information.

[0049] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any of the methods described above.

[0050] A pair of AR glasses, comprising:

[0051] one or more processors; and

[0052] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of any of the aforementioned methods.

[0053] A computer program product comprises a computer program / computer executable instructions, wherein the computer program / computer executable instructions, when executed by a processor in an AR glasses, implement the steps of any of the aforementioned methods.

[0054] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0055] In the embodiment of the present application, multiple sets of training data can be collected sequentially during the whole process of the sample user browsing the physical store through the AR glasses worn by the sample user, and the actual shopping situation of the sample user can be described by the multiple sets of training data. In this way, when the model is trained using the training samples composed of multiple sets of training data, some objective phenomena existing in the physical store can be discovered.

[0056] This objective phenomenon can be reflected as a store browsing route that conforms to the shopping habits and shopping preferences of most users. Specifically, during the model training process, the browsing route of the physical store can be generated by combining the browsing routes of sample users in the physical store determined by the location information and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information. This browsing route conforms to the shopping habits of most consumer users in the physical store and is universal and universal. In this way, when the model provides intelligent shopping guide services to consumer users, it can guide consumer users to shop along the browsing route, allowing users to efficiently and conveniently view the products provided by the physical store, which helps to improve user shopping efficiency and product transaction conversion rate.

[0057] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0059] Figure 1 It is a flowchart of a method for constructing a commodity information providing model provided in an embodiment of the present application;

[0060] Figure 2 This is a schematic diagram of merchandise display on some shelves provided in an embodiment of the present application;

[0061] Figure 3 is a flow chart of a method for providing commodity information provided by an embodiment of the present application;

[0062] Figure 4 It is a schematic diagram of a device for constructing a commodity information providing model provided in an embodiment of the present application;

[0063] Figure 5 It is a schematic diagram of a commodity information providing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0065] In order to reduce store costs and increase the conversion rate of commodity transactions, a new shopping model for offline physical stores is proposed. In this model, physical stores can provide multiple low-cost augmented reality (AR) glasses for consumers to choose after entering the store; at the same time, a small number of commodities can be displayed on the shelves in the store as samples for consumers to view. In this way, when consumers browse the commodities displayed in the store through the AR glasses they wear, the AR glasses can collect and identify the commodities within the user's field of view, and display the relevant information of the commodities in an augmented reality way.

[0066] That is to say, consumers can see the real scene of the physical store and the AR content rendered based on the relevant information of the goods through the AR glasses they wear. Preferably, the real scene of the store can be the background, and the AR content is transparently superimposed on the real scene of the store to ensure the visual experience of consumers wearing AR glasses.

[0067] Usually, the relevant information displayed to consumers through AR glasses is related to the users' shopping decisions. It can assist users in solving problems in the shopping process without a shopping guide, thereby improving the transaction conversion rate of goods.

[0068] For example, for offline clothing stores, the relevant information displayed by AR glasses may include at least one of the following items: product size information, color information, price information, material information, design inspiration, and outfit recommendations. For offline supermarkets, the relevant information displayed by AR glasses may include at least one of the following items: product ingredient list information, shelf life information, origin information, processing technology information, and usage scenario information (taking vegetable products as an example, the usage scenario information can be reflected in the recipes available for the vegetable and other ingredients that can be paired with it).

[0069] Under this new shopping model, how to provide more intelligent shopping guide services to consumers who come to the store is a key technical problem that needs to be solved in order to improve user shopping efficiency and increase product transaction conversion rate.

[0070] Correspondingly, the embodiment of the present application provides a solution for providing intelligent shopping guide services based on an AI model. The model can be deployed on AR glasses provided by physical stores to provide shopping guide services to consumers who come to the store through AR glasses.

[0071] The implementation process of the model building method provided in the embodiment of the present application is explained below with reference to specific examples. Figure 1 The flowchart shown may include:

[0072] S101: Obtain multiple training samples, each training sample includes multiple groups of training data collected sequentially during the whole process of the sample user wearing the AR glasses according to the browsing behavior of the sample user in the physical store, and each group of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product.

[0073] Usually, most users have basically similar shopping habits and preferences. Therefore, the embodiments of the present application can use the browsing routes of sample users in physical stores as a training basis. According to the actual browsing situations of multiple sample users in the physical store, a browsing route that conforms to the habits of most users is determined, thereby guiding consumer users to browse the products provided by the physical store that can meet the users' general shopping preferences.

[0074] Taking sample user 1 as an example, from the time he enters the store and puts on AR glasses to the time he takes off the glasses and leaves the store, during the entire process of browsing products through the AR glasses, multiple sets of training data can be collected sequentially according to his browsing route, and usually each set of training data is for different products viewed by the sample user, which may specifically include: the sample user's location information, the identification information of the viewed product, and the sample user's interest level in the viewed product.

[0075] Among them, the location information in the multiple sets of training data collected sequentially can be used to describe the browsing routes of sample users in the store; the information on the interest level of sample users in the viewed products can be used to determine the products that meet the user's general shopping preferences. The browsing route of the physical store generated based on these two aspects of information is the best browsing route that objectively exists in the physical store discovered during the model training process, which conforms to the shopping habits and shopping preferences of most users.

[0076] As an example, the interest level information may include at least one of: the length of time the sample user stays at the viewed product, the number of times the sample user performs voice interaction with the viewed product, the number of times the sample user performs physical interaction with the viewed product, and whether an operation behavior indicating purchase intention is performed. Generally, the longer the sample user stays at the product, the more times the sample user performs voice consultation on the product, the more times the sample user performs physical interaction with the product such as touching or checking the tag, and the more operations indicating purchase intention are performed, the higher the sample user's interest in the product; otherwise, the lower the interest level.

[0077] Among them, the operation behavior indicating the purchase intention may at least include: performing at least one of the following operations: adding the product to the viewed product (i.e., adding the product to the set of products to be settled, commonly known as the "shopping cart"), placing an order, settling the product, and making a payment.

[0078] Of course, in a specific shopping scenario, the interest level information can also be reflected as an operation behavior unique to the specific shopping scenario. For example, in a clothing shopping scenario, a user submits a try-on request for the viewed product, thereby indicating a high level of interest in the product. The present application embodiment does not specifically limit the specific form of the interest level information, the levels of interest level, the judgment thresholds corresponding to each level, etc.

[0079] Taking the training samples collected for sample user A as an example, it may include the following N groups of training data: (location 1, product 1, stay time 1), (location 2, product 2, number of voice interactions 2), ..., (location N, product N, number of body interactions N).

[0080] In practical applications, after the training samples are collected, the samples can be screened out first, for example, samples including a small amount of training data can be screened out, and samples with more diverse training data can be retained. For example, training samples including more interaction methods and / or training samples including more operation behaviors can be regarded as samples with diverse training data. Among them, the interaction methods may include resident viewing, voice interaction, body interaction, etc.; the operation behaviors may include behaviors with purchase intentions, such as the above-mentioned add-to-cart operations, order operations, settlement operations, payment operations, etc., and may also include behaviors without purchase intentions, such as deleting items from the shopping cart.

[0081] S102: Construct an initial model for providing shopping guide services.

[0082] In the embodiment of the present application, model training can be performed based on traditional deep learning methods. For example, the network topology can be determined based on convolutional neural networks (CNN), an initial model can be constructed, and then model training can be performed based on a large number of training samples and a preset loss function to obtain a commodity information provision model that meets the usage requirements.

[0083] Alternatively, an existing general model, such as a pre-trained large model, may be determined as the initial model, and then based on fine-tuning technology, a product information provision model that meets usage requirements may be obtained through a small amount of training samples.

[0084] S103: Using multiple sets of training data collected in the order included in the training samples as inputs of the initial model, the initial model is trained to obtain the product information providing model, so that the product information providing model has the ability to provide browsing route recommendation services for the products displayed in the physical store, and the browsing route is generated based on the browsing movements of the sample users in the physical store determined by the location information, and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information.

[0085] The embodiment of the present application uses the browsing route of sample users in physical stores as the training basis, so the multiple sets of training data included in different training samples can be used as the input of the initial model to participate in the model training, so that the model can find a browsing route recommendation plan that is more suitable for the physical store. On the one hand, the browsing route is in line with the shopping habits of general users in the physical store, which helps to improve the shopping experience of in-store consumer users; on the other hand, based on the browsing route, the goods in the physical store that meet the general shopping preferences of users can be provided to consumer users for browsing as comprehensively as possible, which also helps to improve the shopping efficiency of in-store consumer users and the conversion rate of goods transactions.

[0086] As an example, products that meet the user's general shopping preferences can be reflected as popular products in the store. For example, products with a large number of visitors, a long browsing time, and high sales can all be regarded as popular products in the embodiments of the present application.

[0087] Correspondingly, the following solution for obtaining a browsing route of a physical store through model training can be provided: in the process of training the initial model using the training samples, the popular products of the physical store determined according to the interest level information are obtained as the products that meet the user's general shopping preferences, and then the browsing route of the physical store is generated based on the browsing movements of the sample users in the physical store determined according to the location information collected in the sequence.

[0088] See also Figure 2 As shown in FIG. 1 , regarding the merchandise display schematic diagram of some shelves 1 and 2 of physical store A, if merchandise A and merchandise E are determined to be popular merchandises according to the interest level information in the training samples, the route between merchandise A and merchandise E may be included when generating the browsing route of physical store A. For example, the partial route may be embodied as: after browsing merchandise A at shelf 1, navigate directly to shelf 2, and navigate to merchandise E after passing merchandise D to browse merchandise E.

[0089] As another example, the products that meet the user's general shopping preferences can be further embodied as related products that can be sold together, that is, products with product association in the embodiments of the present application. For example, tops and skirts that can be worn together, and a variety of ingredients that can be cooked together, etc., can all be regarded as related products in the embodiments of the present application.

[0090] Correspondingly, if the operation behavior indicating the purchase intention is that the sample user performs a purchase operation on the viewed product, the following scheme for obtaining the browsing route of the physical store through model training can be provided: in the process of training the initial model using the training sample, the product association between multiple purchased products can also be obtained, and the products with product associations are used as the products that meet the user's general shopping preferences. In this way, the browsing route of the physical store can be generated based on the products with product associations, the popular products, and the browsing route.

[0091] Among them, product correlation refers to the possibility of different products being sold together. The higher the possibility of selling products together, the higher the product correlation.

[0092] Still Figure 2 For example, if it is found that there is a correlation between product C and product E according to the product purchase information in the training sample, the route from product A to product E to product C can be included when generating the browsing route of physical store A. For example, this part of the route can be reflected as: after browsing product A at shelf 1, navigating to product C after passing product B, and after browsing product C, navigating to product E on shelf 2 to browse product E.

[0093] The popular products and products with product associations in the embodiments of the present application are mainly determined based on the actual browsing situation of sample users in the physical store, which is also an objective phenomenon existing in the physical store discovered during the model training process. This phenomenon is consistent with the interest level of most users in store products and their preference for purchasing related products together. Generating a store browsing route based on this helps consumers quickly discover products that meet their shopping needs while shopping along the browsing route, thereby improving user shopping efficiency and product transaction conversion rate.

[0094] Optionally, based on the popular products and products with product associations determined by model training, the product information providing model of the embodiment of the present application may also preferably provide the following services: providing the identification information of the popular products and / or the identification information of the products with product associations to the manager user associated with the physical store, so that the manager user can adjust the product display method of the physical store accordingly.

[0095] The embodiments of the present application can provide a variety of implementation solutions for adjusting the display method of store merchandise, which are illustrated below with examples.

[0096] In one solution, if the display locations of popular products and / or products with product associations are relatively scattered in the store, the display locations of these products can be adjusted to display them in a centralized manner. For example, a fixed area can be determined in the physical store as a product hot zone, and all popular products can be aggregated and displayed in the product hot zone for consumers to browse in a centralized manner. In this way, the browsing path of the store can also be optimized, so that consumers can complete product browsing in the shortest possible path, further improving the efficiency of shopping, and thus providing consumers with a better offline shopping experience.

[0097] Under another solution, you can adjust the display position of non-popular products and place them together with popular products to expose them and increase their transaction conversion rate.

[0098] In another solution, the display positions of related commodities can be adjusted, and at least two related commodities can be placed together. This will help increase the possibility of selling commodities in combination, bring exposure to more commodities, and also help increase the transaction conversion rate of commodities.

[0099] In addition, in offline shopping scenarios, offline shopping guides usually provide targeted answers to questions that consumers have during the shopping process. In response to the special needs of shopping guides in offline shopping scenarios, the physical store in the embodiment of the present application can pre-configure shopping guide information for products based on the user's concerns associated with different products. That is, for questions that may exist during the user's shopping process and need to be answered by the shopping guide, pre-configure relevant answers and use them as shopping guide information. When the sample user browses the products, the shopping guide information is pushed for the products viewed by the user to promote the transaction of the product.

[0100] In terms of information type, shopping guide information may at least include: product knowledge answer information and / or product store promotion information.

[0101] Of course, in actual applications, the above-mentioned types of shopping guide information can be further subdivided according to usage needs. For example, for knowledge answer information, it can be subdivided into answer information for materials, answer information for outfit recommendations, answer information for size selection, answer information for ingredients, answer information for processing technology, answer information for product usage scenarios, and so on; for store promotion information, it can be subdivided into store member discount information, single product discount information, store full reduction information, and so on. Generally, the finer the classification of shopping guide information types, the more accurately the factors that affect users' purchasing decisions can be located. In this way, when providing shopping guide services to consumers who come to the store in the future, targeted shopping guide information can be pushed to consumers more accurately, which also helps to improve the conversion rate of product transactions.

[0102] That is to say, if the sample user performs a purchase operation on the viewed product, and the purchase operation is realized after shopping guide information is pushed to the sample user, the training data may also include: the push rounds for pushing shopping guide information to the sample user, and the types and contents of the shopping guide information pushed in different push rounds.

[0103] Correspondingly, when performing model training, the push rounds and the types and contents of the shopping guide information pushed in different push rounds can also be used as inputs of the initial model to perform model training on the initial model, so that the product information push model has the ability to obtain the degree of influence of different types of shopping guide information on shopping decisions, and determine the shopping guide information push method corresponding to the target product currently viewed by the consumer user based on the degree of influence. The shopping guide information push method includes the type of shopping guide information associated with the target product and the push priority of different types of shopping guide information.

[0104] That is to say, after adding the training data related to the push of shopping guide information to the model training process, the model can learn and discover a new objective phenomenon determined by the actual purchase behavior of sample users, which can indicate the degree of influence of different types of shopping guide information on user shopping decisions. Based on this, when in-store consumers view target products of interest, a shopping guide information push method can be determined in a targeted manner to clarify the rounds of pushing shopping guide information to consumer users and the corresponding type of shopping guide information pushed in each round.

[0105] In an embodiment of the present application, the shopping guide information can be configured based on the human experience of the administrator user and the offline shopping guide staff, and / or, the user's focus on different products can be determined through model training, and the shopping guide information can be configured based on the focus.

[0106] In one example, AR glasses are equipped with a voice interaction module and can provide intelligent question-and-answer services to users wearing AR glasses. Correspondingly, the consulting questions input by sample users during voice interaction for the viewed products can be obtained and used as training data for model training. These consulting questions can better reflect the real needs of sample users for the viewed products, so the user focus determined in this way is more objective and accurate.

[0107] Specifically, the user focus points associated with the viewed product can be obtained in the following manner: taking the consulting question as the input of the initial model, training the initial model, and enabling the product information providing model to have the ability to determine the user focus points associated with the viewed product based on the consulting question.

[0108] In the embodiment of the present application, the commodity can be used as the basic granularity to determine the user attention points associated with different commodities; or a coarse-grained aggregation can be performed, for example, the commodity category can be used as the basic granularity to determine the user attention points associated with commodities of different categories. The specific granularity can be determined according to the use needs.

[0109] It is understandable that in order to enable the product information providing model to provide more accurate shopping guide services for store products, more information related to the store type can be collected according to the store type of the physical store, such as clothing stores, supermarkets, etc., and used as model input to participate in model training, so as to optimize the product information providing model into a vertical model that provides accurate shopping guide services for stores of specified types.

[0110] Correspondingly, the embodiment of the present application can also provide a method for obtaining model training data. The method can be applied to augmented reality AR glasses provided by physical stores, and the training data required for model training is obtained according to the following method:

[0111] During the whole process of the sample user wearing the AR glasses and browsing the displayed goods in the physical store, multiple sets of training data related to the viewed goods are sequentially collected, and the multiple sets of training data are used as a training sample for model training to obtain a product information provision model for providing shopping guide services to in-store consumer users. The training data may include: the location information of the sample user, the identification information of the viewed goods, and the interest level information of the sample user in the viewed goods.

[0112] In the embodiment of the present application, training data from the first-person perspective of a sample user can be collected through AR glasses, which helps to improve the accuracy and convenience of obtaining training data, and further improve the accuracy of the model trained based on the training data. The following is an example of the implementation process of collecting training data.

[0113] For location information, it refers to the location where the sample user stayed, that is, during the entire product browsing process, the location where the sample user viewed the products of interest was identified. For the products passed by during the browsing process, it is usually not necessary to obtain the corresponding location information. The location information in multiple sets of training data collected sequentially can describe the browsing route of the sample user in the physical store.

[0114] Under one solution, multiple sensors can be deployed at different locations in a physical store, and the location information of sample users wearing the AR glasses can be obtained by positioning the AR glasses through multiple sensors.

[0115] In another solution, the product images captured by AR glasses can be used to identify the products viewed by the sample users, and then the location information of the sample users can be determined based on the display location of the products in the store.

[0116] In another solution, the background image captured by the AR glasses can be used to identify the location of the store corresponding to the background image, thereby determining the location information of the sample user.

[0117] The identification information and interest level information of the viewed product can be determined based on the information collected by the AR glasses. The information collected by the AR glasses may include: image information within the field of view of the AR glasses and / or voice interaction information between the AR glasses and the sample user.

[0118] For example, the identification information of the viewed product can be identified from the product image captured by the AR glasses through image feature extraction, QR code recognition, etc.; and / or, the product name, product number, etc. entered by the sample user for the viewed product can be identified from the voice captured by the AR glasses through voice recognition, so as to clarify the identification information of the viewed product.

[0119] For another example, image recognition and / or voice recognition can be used to determine the length of time the sample user stays at the viewed product, the number of times the sample user has voice interaction with the viewed product, the number of times the sample user has physical interaction with the viewed product, whether operations indicating purchase intention are performed, etc., thereby clarifying the sample user's interest level in the viewed product.

[0120] Correspondingly, the embodiment of the present application can also provide a method for providing product information based on an AI model. The method can be applied to augmented reality AR glasses provided by physical stores, and provide shopping guide services to in-store consumers based on the product information provision model deployed in the AR glasses. Figure 3 As shown in the flow chart, the method may include:

[0121] S301: When it is determined that the consumer user has a store navigation requirement, the current target location of the consumer user is obtained.

[0122] S302: Calling the product information providing model, wherein the model is trained based on multiple training samples to obtain the ability to provide browsing route recommendation services for the products displayed in the physical store; wherein each training sample includes multiple groups of training data collected sequentially during the whole process of the sample user wearing the AR glasses according to the browsing behavior of the sample user in the physical store, and each group of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product.

[0123] S303: Using the target location as a model input, obtaining a target navigation route output by the model based on the browsing route, wherein the target navigation route takes the target location as a starting point, and is used to guide the consumer user to browse the merchandise displayed on the target navigation route through the AR glasses worn by the consumer user; wherein the browsing route is generated based on the browsing movement of the sample user in the physical store determined by the location information, and the merchandise in the physical store that can satisfy the user's general shopping preferences determined by the interest level information.

[0124] The embodiment of the present application provides users with intelligent shopping guide services through an AI model, which can be reflected in at least the following two aspects:

[0125] First, provide store shopping guidance services to consumers.

[0126] That is, AR glasses are used to guide consumers to browse the products displayed in physical stores according to the target navigation route. In this aspect, the product information provided by the AI ​​model can be reflected as path information related to the products displayed in the physical store.

[0127] Combined with the needs of consumer users, the target navigation route output by the model can be reflected in the following two situations:

[0128] Under one usage requirement, after consumers enter the store and wear AR glasses, store guidance services can be provided.

[0129] Specifically, the following solutions can be used to identify the store navigation needs of consumer users:

[0130] For example, AR glasses can detect the usage status, and when the AR glasses are in use, determine that the consumer user currently wearing the AR glasses has a need for store navigation. The usage status detection can be reflected in at least one of the following items: detecting the wearing action, detecting that the AR glasses enter the working state, detecting that the consumer user scans the AR glasses through a terminal device such as a mobile phone to bind, detecting that the consumer user performs a preset gesture to use the AR glasses function, etc., all of which can determine that the AR glasses are currently in use.

[0131] For another example, after a consumer user enters a store and wears AR glasses, he or she can express his or her store navigation needs through voice interaction, such as voice input of "how to shop". Correspondingly, the AR glasses can identify whether the consumer user has store navigation needs based on the voice input.

[0132] Under this usage requirement, the AR glasses can obtain the starting point of the browsing route generated by the model training as the target location. Correspondingly, after the starting point of the browsing route is input into the model, the target navigation route output by the model can be obtained, and the target navigation route is the browsing route.

[0133] The starting point of the browsing route may be a preset location determined during the model training process. For example, the preset location may be an AR glasses display area or an entrance to a merchandise display area of ​​a physical store.

[0134] That is to say, after a consumer user enters a store and wears AR glasses, the model deployed in the AR glasses can actively provide the consumer user with a store-shopping guidance service, and the target navigation route provided under this service is the entire browsing route generated by model training. In this way, the consumer user is guided to shop along the browsing route, and can efficiently and conveniently view the products displayed in the physical store.

[0135] Under another usage requirement, store guidance services can be provided to consumers while they are browsing products while wearing AR glasses.

[0136] That is to say, if the store shopping guidance service is not provided after the consumer user enters the store and puts on the AR glasses, or if a browsing route is recommended to the consumer user but the user does not follow the route to shop around the store, then while the consumer user is shopping around the store wearing the AR glasses, he or she can submit store navigation requests through the AR glasses at any time, triggering the product information recommendation model to provide him or her with store shopping guidance services.

[0137] Specifically, AR glasses can identify navigation requests submitted by consumer users through the AR glasses they are wearing during the product browsing process through gesture detection, voice recognition, etc., so as to clarify when consumer users have store navigation needs.

[0138] Under this usage requirement, the target position is obtained by positioning the AR glasses; then, after the target position is input into the model, the target navigation route output by the model can be obtained, and the target navigation route is part or all of the browsing route.

[0139] That is to say, when consumers are browsing products through the AR glasses they are wearing, the AR glasses can passively provide store shopping guidance services to consumers when triggered by user requests, and the target navigation route provided under this service is part or all of the browsing route generated by model training.

[0140] For example, the path node closest to the target location can be determined from the browsing route, and the target location can be used as the starting point of the target navigation route. After navigating to the path node, the consumer user can be guided to shop along the route of the browsing route including the path node. Alternatively, after navigating from the target location to the path node, the consumer user can be guided to shop starting from the path node and following the route of the entire browsing route until the consumer user returns to the path node and ends the shop.

[0141] Secondly, provide product explanation services to consumers.

[0142] That is, after the target product currently viewed by the consumer user is identified through the AR glasses worn, shopping guide information related to the target product can be pushed in a targeted manner. In this aspect, the product information provided by the AI ​​model can be reflected as shopping guide information related to the target product.

[0143] In one example, after the target product currently viewed by the consumer user is identified through AR glasses, the product information providing model can actively determine the shopping guide information push method corresponding to the target product and provide the consumer user with the corresponding shopping guide information. The specific implementation process can be found in the above introduction, and no example will be given here.

[0144] In another example, after the AR glasses recognize the consumer user's voice input consultation questions about the target product currently being viewed, the product information provision model is triggered to generate answers corresponding to the consultation questions and provide them to the consumer user. In other words, the product information provision model can passively provide the consumer user with the product explanation service required by the user under the trigger of the user's voice.

[0145] In summary, the embodiment of the present application can sequentially collect multiple sets of training data during the whole process of the sample user browsing the physical store through the AR glasses worn, and describe the actual shopping situation of the sample user through the multiple sets of training data. In this way, when using the training samples composed of multiple sets of training data to perform model training, some objective phenomena existing in the physical store can be discovered.

[0146] This objective phenomenon can be reflected as a store browsing route that conforms to the shopping habits and shopping preferences of most users. Specifically, during the model training process, the browsing route of the physical store can be generated by combining the browsing routes of sample users in the physical store determined by the location information and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information. This browsing route conforms to the shopping habits of most consumer users and is universal and universal. In this way, when providing intelligent shopping guide services to consumer users through the model, consumer users can be guided to shop along the browsing route, allowing users to efficiently and conveniently view the products provided by the physical store, which helps to improve user shopping efficiency and product transaction conversion rate.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0148] Corresponding to the aforementioned method embodiment, the embodiment of the present application also provides a device for constructing a product information provision model, wherein the product information provision model is deployed in augmented reality AR glasses provided by a physical store, and shopping guide services are provided to consumers who come to the store through the AR glasses. Figure 4 , the device may include:

[0149] The sample obtaining unit 401 is used to obtain a plurality of training samples, each of which includes a plurality of sets of training data sequentially collected during the whole process of the sample user wearing the AR glasses according to the browsing situation of the sample user in the physical store, and each set of training data includes the location information of the sample user, the identification information of the viewed commodity, and the interest level information of the sample user in the viewed commodity;

[0150] A model building unit 402, used to build an initial model for providing shopping guide services;

[0151] The model training unit 403 is used to use multiple groups of training data collected in the order included in the training samples as the input of the initial model, perform model training on the initial model, and obtain the product information providing model, so that the product information providing model has the ability to provide browsing route recommendation services for the products displayed in the physical store, and the browsing route is generated based on the browsing movement of the sample user in the physical store determined by the location information, and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information.

[0152] The interest level information includes at least one of the following: the length of time the sample user stays at the viewed product, the number of times the sample user performs voice interaction with the viewed product, the number of times the sample user performs physical interaction with the viewed product, and whether an operation behavior indicating purchase intention is performed.

[0153] The model training unit can be specifically used to: in the process of training the initial model using the training samples, obtain the popular products of the physical store determined according to the interest level information as the products that meet the user's general shopping preferences, and then generate a browsing route of the physical store based on the browsing movements of the sample users in the physical store determined according to the location information collected in the sequence.

[0154] The operation behavior indicating the purchase intention is that the sample user performs a purchase operation on the viewed product.

[0155] The model training unit can be specifically used to: in the process of training the initial model using the training samples, obtain the product association between multiple purchased products, and use the products with product association as the products that meet the user's general shopping preferences, so as to generate a browsing route of the physical store based on the products with product association, the popular products and the browsing route.

[0156] The model training unit is further used to: based on the product association, enable the product information providing model to have the ability to identify matching products associated with the product currently purchased by the consumer user and provide product information of the matching products.

[0157] Among them, the model training unit is specifically used to: enable the product information providing model to have the ability to provide the identification information of the popular products and / or the identification information of the products with product association to the manager user associated with the physical store, so that the manager user can adjust the product display method of the physical store.

[0158] The sample user performs a purchase operation on the viewed product, and the purchase operation is realized after a shopping guide information is pushed to the sample user, the shopping guide information is pre-configured by the physical store according to the user's focus associated with the viewed product, and the type of the shopping guide information includes at least product knowledge answer information and / or product store promotion information;

[0159] The training data also includes: the push rounds for pushing shopping guide information to the sample users, and the types and contents of the shopping guide information pushed in different push rounds;

[0160] The model training unit is further specifically used to: use the push rounds and the types and contents of the shopping guide information pushed in different push rounds as inputs of the initial model, perform model training on the initial model, and enable the product information push model to obtain the degree of influence of different types of shopping guide information on shopping decisions, and determine the shopping guide information push method corresponding to the target product currently viewed by the consumer user according to the degree of influence. The shopping guide information push method includes the shopping guide information type associated with the target product and the push priority of different types of shopping guide information.

[0161] The training data also includes: the consulting questions input by the sample users during voice interaction with the viewed products,

[0162] The model training unit is further specifically used to: use the consulting question as the input of the initial model to perform model training on the initial model, and also enable the product information providing model to have the ability to determine the user's focus points associated with the viewed product based on the consulting question.

[0163] Corresponding to the aforementioned method embodiment, the embodiment of the present application further provides a device for obtaining model training data, which is applied to augmented reality AR glasses provided by physical stores, and the device may include:

[0164] A data collection unit is used to sequentially collect multiple sets of training data related to the viewed products during the whole process of the sample user wearing the AR glasses to browse the displayed products in the physical store, and the multiple sets of training data are used as a training sample for model training to obtain a product information provision model for providing shopping guide services to in-store consumer users;

[0165] The training data includes: location information of the sample user, identification information of the viewed commodity, and information on the interest level of the sample user in the viewed commodity;

[0166] The location information is obtained by positioning the AR glasses through multiple sensors deployed in the physical store, and the location information in the multiple sets of training data collected sequentially is used to represent the browsing route of the sample user in the physical store;

[0167] The identification information of the viewed product and the interest level information are determined based on information collected by the AR glasses, where the information collected by the AR glasses includes image information within the field of view of the AR glasses and / or voice interaction information between the AR glasses and the sample user.

[0168] Corresponding to the aforementioned method embodiment, the present application embodiment also provides a product information providing device, which is applied to augmented reality AR glasses provided by physical stores, and provides shopping guide services to in-store consumers based on the product information providing model deployed in the AR glasses. Figure 5 , the device may include:

[0169] The location obtaining unit 501 is used to obtain the current target location of the consumer user when it is determined that the consumer user has a store navigation requirement;

[0170] The model calling unit 502 is used to call the product information providing model, wherein the model is trained based on a plurality of training samples to obtain the ability to provide browsing route recommendation services for the products displayed in the physical store; wherein each training sample includes a plurality of sets of training data sequentially collected during the whole process of the sample user wearing the AR glasses according to the browsing situation of the sample user in the physical store, and each set of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product;

[0171] The navigation route obtaining unit 503 is used to take the target position as a model input, and obtain a target navigation route output by the model based on the browsing route, wherein the target navigation route takes the target position as a starting point, and is used to guide the consumer user to browse the commodities displayed on the target navigation route through the AR glasses worn by the consumer user; wherein the browsing route is generated based on the browsing movement of the sample user in the physical store determined by the position information, and the commodities in the physical store that can meet the user's general shopping preferences determined by the interest level information.

[0172] After the consumer user enters the store and wears the AR glasses, it is determined that the consumer user has a store navigation need.

[0173] The location acquisition unit may be specifically configured to: obtain the starting point of the browsing route generated by the model training as the target location; the browsing route is generated based on the browsing movement of the sample user in the physical store determined by the location information, and the commodities in the physical store that can satisfy the user's general shopping preferences determined by the interest level information;

[0174] The navigation route obtaining unit may be specifically configured to: after inputting the starting point of the browsing route into the model, obtain a target navigation route output by the model, wherein the target navigation route is the browsing route.

[0175] When receiving a navigation request submitted by the consumer user through the AR glasses worn during product browsing, it is determined that the consumer user has a store navigation demand, then

[0176] The position obtaining unit may be specifically configured to: obtain the target position by positioning the AR glasses;

[0177] The navigation route obtaining unit may be specifically configured to obtain a target navigation route output by the model after the target location is input into the model, and the target navigation route may be part or all of the browsing route.

[0178] In addition, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods in the aforementioned method embodiments are implemented.

[0179] And an AR glasses, comprising:

[0180] one or more processors; and

[0181] A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of the aforementioned method embodiments.

[0182] In addition, an embodiment of the present application also provides a computer program product, including a computer program / computer executable instructions, which, when executed by a processor in the AR glasses, implement the steps of any one of the methods described in the aforementioned method embodiments.

[0183] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly contributed to the prior art in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0184] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0185] The above is a detailed introduction to the scheme provided by the present application. The principles and implementation methods of the present application are described in detail using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present application.

Claims

1. A method for constructing a commodity information providing model, characterized in that: The commodity information providing model is deployed in augmented reality (AR) glasses provided by a physical store, and shopping guide services are provided to consumers who visit the store through the AR glasses. The method includes: Obtain multiple training samples, each training sample includes multiple sets of training data collected sequentially during the whole process of the sample user wearing the AR glasses according to the browsing situation of the sample user in the physical store, and each set of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product; Build an initial model for providing shopping guide services; A plurality of sets of training data collected in the order included in the training samples are used as inputs of the initial model, and the initial model is trained to obtain the product information providing model, so that the product information providing model has the ability to provide browsing route recommendation services for the products displayed in the physical store, and the browsing route is generated based on the browsing movements of the sample users in the physical store determined by the location information, and the products in the physical store that can meet the user's general shopping preferences determined by the interest level information.

2. The method according to claim 1, characterized in that The interest level information includes at least one of: the length of time the sample user stays at the viewed product, the number of times the sample user performs voice interaction with the viewed product, the number of times the sample user performs body interaction with the viewed product, and whether an operation behavior indicating purchase intention is performed. In the process of training the initial model using the training samples, the popular products of the physical store determined according to the interest level information are obtained as the products that meet the user's general shopping preferences, and then the browsing route of the physical store is generated based on the browsing movements of the sample users in the physical store determined according to the location information collected in the sequence.

3. The method according to claim 2, characterized in that The operation behavior indicating purchase intention is that the sample user performs a purchase operation on the viewed product. In the process of training the initial model using the training samples, the product associations between multiple purchased products are also obtained, and the products with product associations are used as the products that meet the user's general shopping preferences, so as to generate a browsing route of the physical store based on the products with product associations, the popular products and the browsing routes.

4. The method according to claim 3, characterized in that Based on the commodity association, the commodity information providing model is also enabled to have the ability to identify matching commodities associated with the commodity currently purchased by the consumer user and provide commodity information of the matching commodities.

5. The method according to claim 3, characterized in that: The product information providing model also has the ability to provide the identification information of the popular products and / or the identification information of the products with product association to the manager user associated with the physical store, so that the manager user can adjust the product display method of the physical store.

6. The method according to any one of claims 1 to 5, characterized in that: The sample user performs a purchase operation on the viewed product, and the purchase operation is realized after a shopping guide information is pushed to the sample user, the shopping guide information is pre-configured by the physical store according to the user's focus associated with the viewed product, and the type of the shopping guide information at least includes product knowledge answer information and / or product store promotion information; The training data also includes: the push rounds for pushing shopping guide information to the sample users, and the types and contents of the shopping guide information pushed in different push rounds; The method also includes: using the push rounds and the types and contents of the shopping guide information pushed in different push rounds as inputs of the initial model, performing model training on the initial model, and enabling the product information push model to obtain the degree of influence of different types of shopping guide information on shopping decisions, and determine the shopping guide information push method corresponding to the target product currently viewed by the consumer user according to the degree of influence, wherein the shopping guide information push method includes the type of shopping guide information associated with the target product and the push priority of different types of shopping guide information.

7. The method according to claim 6, characterized in that The training data also includes: consulting questions input by the sample user during voice interaction with the viewed product, The user attention points associated with the viewed product are obtained in the following manner: The consulting question is used as the input of the initial model to perform model training on the initial model, and the product information providing model is also enabled to have the ability to determine the user's focus points associated with the viewed product based on the consulting question.

8. A method for obtaining model training data, characterized in that: Applied to augmented reality (AR) glasses provided by physical stores, the method includes: During the whole process of the sample user wearing the AR glasses and browsing the displayed commodities in the physical store, multiple sets of training data related to the viewed commodities are sequentially collected, and the multiple sets of training data are used as a training sample for model training to obtain a commodity information provision model for providing shopping guide services to in-store consumer users; The training data includes: location information of the sample user, identification information of the viewed commodity, and information on the interest level of the sample user in the viewed commodity; The location information is obtained by positioning the AR glasses through multiple sensors deployed in the physical store, and the location information in the multiple sets of training data collected sequentially is used to represent the browsing route of the sample user in the physical store; The identification information of the viewed product and the interest level information are determined based on information collected by the AR glasses, where the information collected by the AR glasses includes image information within the field of view of the AR glasses and / or voice interaction information between the AR glasses and the sample user.

9. A method for providing commodity information, characterized in that: Applied to augmented reality (AR) glasses provided by physical stores, a shopping guide service is provided to in-store consumers based on a product information provision model deployed in the AR glasses. The method includes: When it is determined that the consumer user has a store navigation requirement, obtaining the current target location of the consumer user; The product information providing model is called, and the model is trained based on multiple training samples to obtain the ability to provide browsing route recommendation services for the products displayed in the physical store; wherein each training sample includes multiple sets of training data collected sequentially according to the browsing situation of the sample user in the physical store during the whole process of the sample user wearing the AR glasses, and each set of training data includes the location information of the sample user, the identification information of the viewed product, and the interest level information of the sample user in the viewed product; The target location is used as a model input to obtain a target navigation route output by the model based on the browsing route, wherein the target navigation route takes the target location as a starting point and is used to guide the consumer user to browse the merchandise displayed on the target navigation route through the AR glasses worn by the consumer user; wherein the browsing route is generated based on the browsing movements of the sample user in the physical store determined by the location information, and the merchandise in the physical store that can satisfy the user's general shopping preferences determined by the interest level information.

10. The method according to claim 9, characterized in that After the consumer user enters the store and wears the AR glasses, it is determined that the consumer user has a store navigation need, then The obtaining of the target location where the consumer user is currently located includes: obtaining the starting point of the browsing route generated by the model training as the target location; Then, after the starting point of the browsing route is input into the model, the target navigation route output by the model is obtained, and the target navigation route is the browsing route.

11. The method according to claim 9, characterized in that When receiving a navigation request submitted by the consumer user through the AR glasses worn during the product browsing process, it is determined that the consumer user has a store navigation demand, then The obtaining of the target location where the consumer user is currently located comprises: obtaining the target location by positioning the AR glasses; Then, after the target location is input into the model, a target navigation route output by the model is obtained, and the target navigation route is part or all of the browsing route.

12. An AR glasses, characterized in that: include: one or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions, wherein when the program instructions are read and executed by the one or more processors, the steps of the method according to any one of claims 1 to 11 are executed.

Images