Method, device and electronic device for publishing video content and commodity object information

Abstract

The embodiment of the present application discloses a method, device and electronic device for publishing video content and commodity object information, the method comprising: providing an operation option for obtaining recommended tag information in a video content publishing interface; after receiving a request through the operation option, obtaining a recommended tag according to the associated video content; wherein the recommended tag is obtained by using the video content as input information of a prediction model, the prediction model is used to extract features of the video content, and determine the recommended tag of the target video content by performing a similarity comparison with the features of the image template corresponding to the multiple tags; the recommended tag is displayed, and an operation option for selecting the recommended tag is provided; after receiving a publishing request, the video content and the selected recommended tag are submitted to the server for publishing. Through the embodiment of the present application, more effective tag information can be associated with the video content.

Description

Technical Field

[0001] The present application relates to the technical field of video tag prediction, and in particular to a method, device and electronic device for publishing video content and commodity object information. Background Art

[0002] In the video content publishing system, in order to facilitate the distribution or recommendation of video content, tag information can be added to specific video content. Such tags can usually express the video content in a descriptive text form. When distributing or recommending video content to users, it can be done based on such tag information. For example, if a user is interested in videos related to "professional basketball games", or when a user initiates a search using keywords such as "professional basketball games", video content with relevant tags can be provided to the user, and so on.

[0003] Among them, since there are some video contents published by publisher users in the system, for such video contents, users can usually add corresponding tags when publishing video contents. However, due to the lack of professionalism of publisher users, some videos lack effective video tag information. Or, some users may not understand the impact of tag information on the exposure rate of video content, so they often add unfamiliar tags or give up adding tags. The unfamiliarity or lack of tags increases the difficulty of matching and recommending video content with user needs in the future, which makes it difficult for the published video content to be exposed.

[0004] Therefore, how to associate more effective tag information with video content becomes a technical problem that needs to be solved by those skilled in the art. Summary of the invention

[0005] The present application provides a method, device and electronic device for publishing video content and commodity object information, which can associate more effective tag information with video content.

[0006] This application provides the following solutions:

[0007] A method for publishing video content, comprising:

[0008] Providing an operation option for obtaining recommended tag information in the video content publishing interface;

[0009] After receiving the request through the operation option, obtaining the recommended tags according to the associated video content; wherein the recommended tags are obtained by taking the video content as input information of a prediction model, the prediction model is used to extract features of the video content, and determine the recommended tags of the target video content by performing similarity comparison with the features of the image templates corresponding to the multiple tags;

[0010] Displaying the recommended tags and providing operation options for selecting the recommended tags;

[0011] After receiving the publishing request, the video content and the selected recommendation tags are submitted to the server for publishing.

[0012] A method for publishing commodity object information, comprising:

[0013] Provide an operation option for obtaining recommended tag information in the product object publishing interface;

[0014] After receiving the request through the operation option, obtaining a recommended tag according to the video content associated with the commodity object, wherein the recommended tag is obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content and determine the recommended tag of the target video content by performing a similarity comparison with the features of the image templates corresponding to the multiple tags;

[0015] Displaying the recommended tags provided by the server and providing operation options for selecting the recommended tags;

[0016] After receiving the publishing request, the information of the product object and the selected recommendation tags are submitted to the server for publishing.

[0017] A method for generating a video content label, comprising:

[0018] Obtaining a prediction model, wherein the prediction model stores template image feature information for determining features of image templates corresponding to a plurality of tags in a tag library;

[0019] Determining target video content to be predicted;

[0020] At least one recommended tag for the target video content is determined based on the prediction model, wherein the prediction model is used to extract features of the target video content and determine the recommended tag for the target video content by performing similarity comparison with features of image templates corresponding to the multiple tags.

[0021] A method for training a prediction model, comprising:

[0022] Obtaining a training sample, wherein the training sample includes video content and corresponding label information;

[0023] Training the prediction model using training data samples;

[0024] In each round of iterative training, feature information of the video content is extracted, and template image features corresponding to multiple labels in the label library are obtained. According to the similarity between the features of the video content and the template image features corresponding to the labels, the probability that the video content belongs to the categories corresponding to the multiple labels is determined, and according to the label information in the sample, the parameters in the prediction model are updated.

[0025] A device for publishing video content, comprising:

[0026] A first operation option providing unit, configured to provide an operation option for obtaining recommended tag information in a video content publishing interface;

[0027] A first recommended tag acquisition unit is used to acquire recommended tags according to the associated video content after receiving a request through the operation option; wherein the recommended tags are obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content and determine the recommended tags of the target video content by performing similarity comparison with the features of the image templates corresponding to the multiple tags;

[0028] A first recommended tag display unit, used to display the recommended tags and provide operation options for selecting the recommended tags;

[0029] The first publishing unit is used to submit the video content and the selected recommendation tags to the server for publishing after receiving the publishing request.

[0030] A device for publishing commodity object information, comprising:

[0031] A second operation option providing unit, configured to provide an operation option for obtaining recommended tag information in the commodity object publishing interface;

[0032] a second recommendation tag acquisition unit, configured to acquire a recommendation tag according to the video content associated with the commodity object after receiving a request through the operation option, wherein the recommendation tag is obtained by taking the video content as input information of a prediction model, the prediction model is used to extract features of the video content, and determine the recommendation tag of the target video content by performing a similarity comparison with features of image templates corresponding to the multiple tags;

[0033] A second recommendation tag display unit, used to display the recommendation tags provided by the server and provide operation options for selecting the recommendation tags;

[0034] The second publishing unit is used to submit the information of the commodity object and the selected recommendation tag to the server for publishing after receiving the publishing request.

[0035] A device for generating a video content tag, comprising:

[0036] A prediction model acquisition unit, used to acquire a prediction model, wherein the prediction model stores template image feature information, so as to determine features of image templates corresponding to a plurality of tags in a tag library;

[0037] A video content determination unit, used to determine the target video content to be predicted;

[0038] A recommended tag determination unit is used to determine at least one recommended tag for the target video content based on the prediction model, wherein the prediction model is used to extract features of the target video content and determine the recommended tag for the target video content by performing similarity comparison with features of image templates corresponding to the multiple tags.

[0039] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0040] Through the embodiments of the present application, the target video content can be used as the input of the prediction model to predict the probability that the target video content belongs to the categories corresponding to multiple tags, and the prediction results can be used to determine the recommended tags. In this way, the user can select specific tags for the video content to be released from the recommended tags, so as to improve the quality of the tags, which is more conducive to effectively improving the effective exposure rate of the target video content in the specific content distribution or recommendation process. Among them, in the specific prediction model, features can be extracted from the video content, and the similarity can be compared with the features of the template image expressed by multiple tags in the tag library, and the probability that the video content belongs to the categories corresponding to multiple tags can be predicted based on the comparison results. In this way, there is no need to learn the association between the video content and multiple tags through the prediction model, which is convenient for reducing the learning difficulty of the prediction model and also makes it easier for the prediction model to obtain prediction results.

[0041] In order to determine the features of the template image expressed by the label during the prediction process, the correspondence between multiple labels and the features of the template image expressed by the label can be saved in the "dictionary" in advance. Or, in another way, the correspondence between multiple word segments and the features of the template image expressed by the word segmentation can also be saved, and the mapping relationship information between the word segmentation and the label can be saved at the same time. In this way, when making a prediction, the features of the template image expressed by the label can be determined through the features of the template image expressed by the word segmentation and the above-mentioned mapping relationship information. This can help reduce the size of the "dictionary" and facilitate storage and query calculations.

[0042] In addition, the features of the template image expressed by the specific segmentation and the mapping relationship between the segmentation and the label can be gradually learned during the training process of the prediction model, thereby making it easier for the prediction model to obtain accurate prediction results.

[0043] 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

[0044] 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.

[0045] Figure 1 It is a schematic diagram of the system architecture provided by the embodiment of the present application;

[0046] Figure 2 is a flow chart of the first method provided in an embodiment of the present application;

[0047] Figure 3 is a flow chart of the second method provided in an embodiment of the present application;

[0048] Figure 4 is a flowchart of the third method provided in an embodiment of the present application;

[0049] Figure 5 is a flowchart of the fourth method provided in an embodiment of the present application;

[0050] Figure 6 is a schematic diagram of a first device provided in an embodiment of the present application;

[0051] Figure 7 is a schematic diagram of the second device provided in the embodiment of the present application

[0052] Figure 8 is a schematic diagram of a third device provided in an embodiment of the present application

[0053] Fig. 9 It is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] 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.

[0055] In the embodiment of the present application, in order to associate more effective tag information with video content, the video content can be analyzed by an intelligent recognition algorithm (for example, a prediction model such as an artificial neural network), features can be extracted from the video content, and the probability of the video content belonging to the categories corresponding to multiple tags can be predicted, and then the recommended tags corresponding to the video content can be automatically determined based on the prediction results. Such recommended tags can be provided to publisher users, so that when users publish videos, they can select tags that are specifically associated with the video content from the recommended tags, thereby improving the effectiveness of the tags associated with the video content published by the user.

[0056] Specifically, when using the prediction model for prediction, it is first necessary to train the prediction model and provide some training samples so that the prediction model can learn specific classification knowledge from them. Then, the trained prediction model is used to predict the probability that the specific video content belongs to the category corresponding to the specific label.

[0057] If the traditional model training method is used, the model structure can be constructed, including the number of layers of the artificial neural network, etc.; in addition, specific training samples can be collected, which may include multiple video contents (for example, video contents that have been published in the system, etc.) and associated label data (which may be labels added by the publisher user, etc.). Afterwards, the model will be trained using the above training samples. After the algorithm converges, the model parameters will be fixed to obtain a trained model. Then, by inputting specific video content into the prediction model, the probability that the video content belongs to the category corresponding to each label can be predicted.

[0058] However, for a model specifically used for classification, the difficulty of model training and prediction is greatly related to the number of categories that need to be divided. The more categories that need to be divided, the greater the difficulty. In the scenario of the embodiment of the present application, the fields involved in various video contents are different. For example, the video in the commodity object information service system may usually be a video related to a specific commodity object. The labels of such video content may usually include category labels, attribute labels (for example, the material, style, color, etc. of clothing commodities), sales-related labels (for example, whether it is free shipping, etc.), etc.; and there are many categories of commodity objects, and the attribute characteristics and sales attributes of the individual commodity objects associated in the specific video are even more diverse. If it is a system such as an ordinary video publishing platform, the number of labels required for the video content is even larger, usually hundreds of thousands or even higher. That is to say, if the model is directly learned and predicted in a traditional way, the number of categories that need to be divided is very large, and the specific model needs to learn the association between various video contents and the categories corresponding to the specific labels. Therefore, it is very difficult and even difficult to achieve.

[0059] Therefore, in the embodiment of the present application, in order to reduce the difficulty of learning and predicting the model, various labels that may be associated with the video can be collected in advance to form a label set, and the features of the template image expressed by the label can be determined, and the corresponding relationship between the label and the features of the template image expressed by the label can be saved in the form of a "dictionary" or the like. In this way, the extracted features in the video content can be compared with the features of the template image expressed by the specific label, and the probability that the video content belongs to the category to which the corresponding label belongs can be determined based on the specific similarity.

[0060] That is to say, for a specific label, it can usually be used to express a certain template image. For example, if a label is "red high heels", then the label "red high heels" actually expresses a template image. If the specific image content in a video content is also related to "red high heels", then the video content and the template image corresponding to the label "red high heels" will have a relatively high similarity. Therefore, a "dictionary" can be established in advance. The "dictionary" can save the correspondence between the specific label and the features of the template image expressed by the label. In this way, for the prediction model, the features of the video can be first extracted from the specific video content, and then compared with the features of the template image expressed by the specific label. According to the similarity between the two, the probability that the video content belongs to the category represented by the label can be determined.

[0061] In this way, for the prediction model, we only need to focus on how to extract the specific features of the video, without having to learn the associations between various video contents and the categories corresponding to specific labels. Therefore, the difficulty of learning and prediction can be reduced.

[0062] Among them, the feature information of the template image expressed by the specific label is usually expressed by a high-dimensional vector, for example, it can be 2048 dimensions or even higher. Therefore, when the number of labels is very large, the storage space required for the "dictionary" may be very large, and it may also increase the difficulty of calculation.

[0063] To this end, in a preferred embodiment of the present application, the tags in the tag set can also be first segmented and de-duplicated, and the mapping relationship information between the tags and the segmented words can be established (which can exist in the form of a matrix, etc.). The specific "dictionary" can save the corresponding relationship between multiple segmented words and the features of the template image they express, wherein the feature of each segmented word can be represented by a multidimensional vector. As for the features of the template image expressed by the tag, it can be obtained by combining multiple segmented word features according to the features of the segmented words and the mapping relationship matrix. For example, the features of the template image expressed by "red" and the features of the template image expressed by "high heels" can be combined into the features of the template image expressed by the label "red high heels", and so on. After the tags in the tag library are segmented and de-duplicated, the number of segmented words will be much smaller than the number of tags. In this way, the "dictionary" only needs to save the features of the template image expressed by the segmented words. Therefore, the scale of the "dictionary" will be greatly reduced, which is convenient for storage and query.

[0064] Among them, the features of the template image expressed by the word segmentation and the mapping relationship information between the label and the analysis can be determined in advance by other means, or can be determined by learning in the process of training the prediction model, etc.

[0065] From the perspective of system architecture, the solution provided by the embodiments of the present application can be applied to systems related to video content publishing or commodity object information services. In such systems, a client and a server may be included. The client can be used to interact with users, and users can submit publishing requests through a specific client. Among them, the object published by the user may be directly the video content, or it may be commodity object information associated with the video content (for example, the video content provided for the "header image" of the commodity object details page, etc.). In the process of publishing such information, the solution provided by the embodiments of the present application can be used to directly use the video content as the input of the prediction model, and determine the probability that the video content belongs to the category corresponding to the multiple tags by comparing the similarity with the template image features expressed by the multiple tags. In a specific implementation method, such as Figure 1As shown, the "dictionary" and the mapping relationship information between the word segmentation and the complete label can be saved in the model. The dictionary can store the features of the template image expressed by the word segmentation, which can be a 2048-dimensional vector (or other lengths). Among them, each word segmentation corresponds to a feature vector, so the dictionary can store "the number of word segmentations × 2048-dimensional vector"; when making specific predictions, the features of the template image expressed by the label can be obtained based on the information stored in the dictionary and the mapping relationship information between the word segmentation and the complete label. Each complete label can also correspond to a feature vector, so after the conversion is complete, the "number of complete labels × 2048-dimensional vector" can be obtained. After that, the 2048-dimensional vector of the video content extracted from the CNN (convolutional neural network) structure can be multiplied with the feature vector corresponding to each complete label to obtain the probability that the video content belongs to the category corresponding to the specific label.

[0066] The process of training the prediction model is mainly completed on the server side. The process of making predictions through the prediction model can be completed on the server side, or, if the performance of the terminal device associated with the client meets the requirements, the specific prediction model can be sent to the client side, so that the prediction process can also be completed directly on the client side.

[0067] The specific implementation scheme provided in the embodiments of the present application is described in detail below.

[0068] Embodiment 1

[0069] First, from the perspective of the server, this embodiment provides a method for generating video content tags. Figure 2 , the method may specifically include:

[0070] S201: Acquire a prediction model, wherein the prediction model stores template image feature information for determining features of image templates corresponding to a plurality of tags in a tag library;

[0071] In specific implementation, the prediction model can be obtained by training with training samples in advance (the training process will be introduced in detail later). There can be multiple algorithms, structures, etc. of the prediction model. For example, in one way, it can be implemented by an artificial neural network model, etc. In the embodiment of the present application, in the process of training the prediction model, the association between the video content and various tags is not directly learned, but the probability of the video content belonging to the category corresponding to each tag is determined by extracting the video content features and comparing the similarity with the features of the template image represented by the tag.

[0072] As mentioned above, the template image feature information stored in the prediction model can be directly the template image feature information corresponding to each label. However, since this requires a large amount of storage space and search cost, in a preferred embodiment, the labels can be segmented and deduplicated in advance, and the template image feature information expressed by multiple segmentations, as well as the mapping relationship information between the segmentations and the labels, can be stored in the model. In this way, when predicting the target video content, the features of the template image expressed by the label can be determined based on the template image features expressed by the segmentation and the mapping relationship information.

[0073] For example, in a certain tag library, the number of tags may be hundreds of thousands or even millions, and the features corresponding to the tags are usually represented by high-dimensional vectors. Therefore, if the features of the template image expressed by each tag are saved, a large amount of storage space will be required, and there will also be high calculation or query difficulty.

[0074] Therefore, in a preferred embodiment of the present application, the tags in the tag set can also be first segmented and deduplicated, and the mapping relationship information between the tags and the segmented words can be established (which can exist in the form of a matrix, etc.). The specific "dictionary" can save the correspondence between multiple segmented words and the features of the template image they express, wherein the feature of each segmented word can be represented by a multidimensional vector. As for the features of the template image expressed by the label, it can be obtained by combining multiple segmented word features based on the features of the segmented words and the mapping relationship matrix. For example, the features of the template image expressed by "red" and the features of the template image expressed by "high heels" can be combined into the features of the template image expressed by the label "red high heels", and so on.

[0075] For example, if a label is "red high heels", then the two participles "red" and "high heels" can be obtained through word segmentation. Through word segmentation, a label can be divided into two or more participles. On the surface, the number of words after word segmentation increases, but after deduplication, the actual number of participles is much smaller than the number of labels. For example, another label is "red dress", which can be divided into "red" and "dress" through word segmentation. There are also "red windbreaker", "red sweater", "red sweater", etc. After word segmentation and deduplication, the participles obtained are "red", "high heels", "dress", "windbreaker", "sweater", "sweater". In addition, specific labels may also include "black high heels", "black dress", etc. Since the participles "high heels" and "dress" have appeared before, after word segmentation and deduplication, only "black" will become a new participle, and so on. Therefore, by segmenting and removing duplicates from the tags in the tag library, the number of segmented words will be much smaller than the number of tags. In this way, the "dictionary" only needs to save the features of the template image expressed by the segmented words. Therefore, the size of the "dictionary" will be greatly reduced, making it easier to store and query.

[0076] Among them, the features of the template image stored in the "dictionary" can be determined before the model training, or, in a preferred manner, can be learned during the model training process. In this case, when training the model, the initial value of the feature of the template image expressed by the segmentation can be set to a random value, and the specific model can also include a part for learning and training the features of the template image expressed by the segmentation. In this way, after the model training is completed, not only the parameters in the model can be learned, such as the weight matrix in the artificial neural network model, but also the feature information of the template image expressed by each segmentation can be learned, and then it can be saved as the aforementioned "dictionary". In the subsequent prediction process through the model, the features corresponding to the segmentation stored in the "dictionary" can be directly used for prediction.

[0077] In addition, the mapping relationship information between the segmentation words and the label can also be pre-set, or can also be learned during the model training process. That is, when the model training starts, the mapping relationship can be set to an initial value (usually represented by a matrix). For example, if a label consists of two segmentation words, the values ​​corresponding to the two segmentation words in the matrix are 1, and the other segmentation words are 0. Normalization can be performed afterwards, etc. In the training process, the weights of specific segmentation words in the mapping relationship information can be gradually updated to make them more in line with the actual situation. For example, for the label "red high heels", the model believes that "red" is not as important as "high heels". After the training, the weight of "high heels" may be 0.8, and "red" is only 0.2, and so on. In this way, in the mapping relationship, multiple different segmentation words corresponding to the same label are associated with different weights. Such different weights can be used to represent the different importance of different segmentation words when the label expresses the template image, thereby facilitating more accurate prediction.

[0078] S202: Determine the target video content to be predicted;

[0079] After obtaining a specific prediction model, the specific video content can be predicted to determine the probability that the video content belongs to the category corresponding to the specific label.

[0080] The specific target video content may be different in different application scenarios. For example, in a video content publishing platform system, upon receiving a request for publishing video content submitted by a client, the video content to be published may be determined as the target video content to be predicted. In specific implementation, an operation option for obtaining a recommended tag may be provided in the video publishing interface. In the process of publishing video content, the video content to be published may be submitted to the server for prediction, or may be predicted locally on the client, and so on.

[0081] Alternatively, in the product object information service system, if the merchant user includes video content when publishing product object information, then when receiving a request to publish product object information submitted by the client, the video content associated with the product object to be published can be determined as the target video content to be predicted, and so on.

[0082] S203: Determine at least one recommended tag to which the target video content belongs based on the prediction model, wherein the prediction model is used to extract features of the target video content and determine the recommended tag of the target video content by performing similarity comparison with features of image templates corresponding to the multiple tags.

[0083] After determining the specific target video content, it can be used as the input information of the prediction model (specifically, each frame of the target video, etc.), and input into the specific prediction model to determine the probability information that the target video content belongs to the category corresponding to the multiple labels. Specifically, when the prediction model is predicting, it can extract features of the target video content, and determine the probability that the target video content belongs to the category corresponding to the multiple labels by performing similarity comparison with the features of the image template expressed by the multiple labels. If the "dictionary" stores the features of the template image expressed by the segmentation, the features of the image template expressed by the label can be first determined based on the features of the template image expressed by the segmentation and the mapping relationship information between the label and the segmentation. Specifically, the features of the template image expressed by the segmentation can exist in the form of a feature vector, and the mapping relationship between the label and the segmentation can exist in the form of a matrix. By multiplying the feature vector of the template image expressed by the segmentation with the mapping relationship matrix, the corresponding feature vector can be determined as the feature information of the image template expressed by the label.

[0084] In addition, in a specific implementation, the features of the target video content may also exist in the form of a vector, and may have the same dimensional distribution as the feature vector of the image template expressed by the label (the same number of dimensions, and the same attributes corresponding to each dimension). Therefore, when performing a similarity comparison with the features of the image template expressed by the multiple labels, the probability that the target video content belongs to the corresponding label can be determined by calculating the inner product of the feature vector of the target video content and the feature vector of the image template expressed by the label (dot multiplication of the two feature vectors to obtain a real number).

[0085] After determining the probability that the target video content belongs to the categories corresponding to the multiple tags, the tags whose probabilities meet the target conditions can be determined as recommended tags for the target video content. There can be multiple specific target conditions, for example, the probability can be higher than a certain target threshold, or the number of targets ranked first in probability, etc. Specifically, for a target video content, multiple recommended tags can be determined.

[0086] In specific implementation, if the target video content is associated with a commodity object, the attribute information of the commodity object associated with the target video content can also be combined with the output result of the prediction model to determine the recommended tag information. For example, a comprehensive judgment can be made based on the attribute information, location information, time information, etc. of the merchant user associated with the commodity object to determine a more suitable recommended tag, etc.

[0087] After determining the recommended tag, it can be directly associated with the specific target video content for subsequent recommendation or distribution. Alternatively, in another way, the specific recommended tag can be returned to the client for the user to select. For example, in a video content publishing system, the recommended tag can be returned to the client, and after the user makes a selection, the video content to be published can be associated with the tag information from the recommended tag before publishing. Alternatively, in a commodity object information system, the recommended tag can also be returned to the client, and after the user makes a selection, the commodity object to be published can be associated with the tag information from the recommended tag before publishing.

[0088] In summary, in an embodiment of the present application, the target video content can be used as the input of the prediction model to predict the probability that the target video content belongs to the categories corresponding to multiple tags, and the prediction results can be used to determine the recommended tags. In this way, the user can select specific tags for the video content to be released from the recommended tags, so as to improve the quality of the tags, which is more conducive to effectively improving the effective exposure rate of the target video content in the specific content distribution or recommendation process. Among them, in the specific prediction model, features can be extracted from the video content, and the similarity can be compared with the features of the template image expressed by multiple tags in the tag library, and the probability that the video content belongs to the categories corresponding to multiple tags can be predicted based on the comparison results. In this way, there is no need to learn the association between the video content and multiple tags through the prediction model, which is convenient for reducing the learning difficulty of the prediction model and also makes it easier for the prediction model to obtain prediction results.

[0089] In order to determine the features of the template image expressed by the label during the prediction process, the correspondence between multiple labels and the features of the template image expressed by the label can be saved in the "dictionary" in advance. Or, in another way, the correspondence between multiple word segments and the features of the template image expressed by the word segmentation can also be saved, and the mapping relationship information between the word segmentation and the label can be saved at the same time. In this way, when making a prediction, the features of the template image expressed by the label can be determined through the features of the template image expressed by the word segmentation and the above-mentioned mapping relationship information. This can help reduce the size of the "dictionary" and facilitate storage and query calculations.

[0090] In addition, the features of the template image expressed by the specific segmentation and the mapping relationship between the segmentation and the label can be gradually learned during the training process of the prediction model, thereby making it easier for the prediction model to obtain accurate prediction results.

[0091] Embodiment 2

[0092] The second embodiment corresponds to the first embodiment and provides a method for publishing video content from the perspective of a client in a video content publishing system. Figure 3, the method may include:

[0093] S301: providing an operation option for obtaining recommended tag information in the video content publishing interface;

[0094] S302: After receiving the request through the operation option, obtain the recommended tags according to the associated video content; wherein the recommended tags are obtained by taking the video content as input information of a prediction model, the prediction model is used to extract features of the video content, and determine the recommended tags of the target video content by performing similarity comparison with the features of the image templates corresponding to the multiple tags;

[0095] S303: Displaying the recommended tags and providing operation options for selecting the recommended tags;

[0096] S304: After receiving the publishing request, the video content and the selected recommendation tags are submitted to the server for publishing.

[0097] Embodiment 3

[0098] The third embodiment is also corresponding to the first embodiment, and provides a method for publishing commodity object information from the perspective of the client in the commodity object information service system, see Figure 4 , the method may include:

[0099] S401: providing an operation option for obtaining recommended tag information in the product object publishing interface;

[0100] S402: After receiving the request through the operation option, obtain the recommended tags according to the video content associated with the commodity object, wherein the recommended tags are obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content and determine the recommended tags of the target video content by performing similarity comparison with the features of the image templates corresponding to the multiple tags;

[0101] S403: Displaying the recommended tags and providing operation options for selecting the recommended tags;

[0102] S404: After receiving the publishing request, the information of the product object and the selected recommendation tag are submitted to the server for publishing.

[0103] Specifically, in a more specific application scenario, the commodity objects may specifically include: catering or fresh food commodity objects that are delivered to a user-specified address at a user-specified time within a target area around an associated physical store.

[0104] At this time, in the specific implementation, the basic information associated with the product object can also be obtained and automatically filled in the corresponding input control for publication, wherein the basic information may include associated physical store information, location information, inventory information, and / or expiration date information, etc.

[0105] Embodiment 4

[0106] This fourth embodiment also provides a method for training a prediction model. Figure 5 , the method may specifically include:

[0107] S501: Acquire a training sample, where the training sample includes video content and corresponding label information;

[0108] S502: Training the prediction model using training data samples;

[0109] In each round of iterative training, feature information of the video content is extracted, and template image features corresponding to multiple labels in the label library are obtained. According to the similarity between the features of the video content and the template image features corresponding to the multiple labels, the probability that the video content belongs to the categories corresponding to the multiple labels is determined, and according to the label information in the sample, the parameters in the prediction model are updated.

[0110] In specific implementation, the multiple tags in the tag library can also be segmented and deduplicated to obtain a segmentation set, and a mapping relationship between the tags and the segmentations can be established; at this time, the template image features expressed by the multiple tags can be determined based on the template image features expressed by the segmentations and the mapping relationship between the tags and the segmentations.

[0111] In a specific implementation, the template image feature expressed by the word segmentation may be a random value in an initial state, and is updated during the iterative training of the prediction model. In addition, the mapping relationship between the label and the word segmentation may be updated during the iterative training of the prediction model, wherein in the updated mapping relationship, multiple different word segmentations corresponding to the same label are associated with different weights to represent the different importance of different word segmentations when the label expresses the template image.

[0112] Specifically, when obtaining training samples, multiple video contents and corresponding initial label information are obtained. For example, the video contents may be published in the system and associated with label information, or the video contents may be collected in advance through other channels, and the corresponding label information may be added, etc. Of course, since the specific label information may be added by the user, it may not be accurate enough. To this end, in the specific implementation, the sample data may also be "cleaned" to improve the quality of the training samples. There may be many specific "cleaning" methods. For example, in one case, since the specific video content may be associated with commodity object information, the video content may be clustered according to the commodity object information associated with the video content. For example, the video content associated with the same or the same commodity object may be clustered, etc. Afterwards, the initial labels corresponding to the same video content may be counted, and the low-frequency initial labels therein may be filtered out. In other words, the same video content can usually be described using roughly the same labels. If a label that is obviously different from other similar video content appears in the label of a certain video content, it can be filtered out to avoid affecting the accuracy of the training results.

[0113] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0114] Corresponding to the first embodiment, the present application embodiment also provides a device for generating a video content tag, see Figure 6 , the device may include:

[0115] The prediction model acquisition unit 601 is used to acquire a prediction model, wherein the prediction model stores template image feature information, so as to determine features of image templates corresponding to a plurality of tags in the tag library;

[0116] A video content determination unit 602, configured to determine a target video content to be predicted;

[0117] The recommended tag determination unit 603 is used to determine at least one recommended tag of the target video content according to the prediction model, wherein the prediction model is used to extract features of the target video content and determine the recommended tag of the target video content by performing similarity comparison with features of image templates corresponding to the multiple tags.

[0118] The template image feature information stored in the prediction model includes: template image feature information corresponding to a plurality of word segments, wherein the word segments are determined after word segmentation and deduplication of a plurality of tags in the tag library, and the prediction model also stores mapping relationship information between word segments and tags;

[0119] The prediction model is specifically used to determine the features of the template image corresponding to the label according to the template image features corresponding to the word segments and the mapping relationship information when predicting the target video content.

[0120] The template image feature information expressed by the multiple word segments is determined in the process of training the prediction model.

[0121] The mapping relationship information between the participles and the labels is determined in the process of training the prediction model. In the mapping relationship, multiple different participles corresponding to the same label are associated with different weights, which are used to represent the different importance of different participles when the label expresses the template image.

[0122] The features of the target video content and the features of the image template expressed by the label are feature vectors with the same dimensional distribution;

[0123] The prediction model is specifically used to determine the probability that the target video content belongs to the corresponding label by calculating the inner product of the feature vector of the target video content and the feature vector of the image template expressed by the label.

[0124] Specifically, the video content determination unit may be used to:

[0125] When a request for publishing video content submitted by a client is received, the video content to be published is determined as the target video content to be predicted.

[0126] At this time, the device may further include:

[0127] The first recommendation tag providing unit is used to provide the recommendation tag to the client, so as to associate tag information with the video content to be published from the recommendation tag and then publish it.

[0128] Alternatively, the video content determination unit may be specifically configured to:

[0129] When a request for publishing the commodity object information submitted by the client is received, the video content associated with the commodity object to be published is determined as the target video content to be predicted.

[0130] At this time, the device may further include:

[0131] The second recommended tag providing unit is used to provide the recommended tags to the client, so that the commodity object to be published can be published after being associated with tag information from the recommended tags.

[0132] In addition, the device may also include:

[0133] A combining unit is used to combine the attribute information of the commodity object associated with the target video content with the output result of the prediction model to determine the recommended tag information.

[0134] Corresponding to the second embodiment, the present application embodiment also provides a device for publishing video content, see Figure 7 , the device may include:

[0135] A first operation option providing unit 701 is used to provide an operation option for obtaining recommended tag information in the video content publishing interface;

[0136] A first recommended tag acquisition unit 702 is used to acquire recommended tags according to the associated video content after receiving a request through the operation option; wherein the recommended tags are obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content and determine the recommended tags of the target video content by performing similarity comparison with the features of the image templates corresponding to the multiple tags;

[0137] A first recommended tag display unit 703, used to display the recommended tags and provide operation options for selecting the recommended tags;

[0138] The first publishing unit 704 is used to submit the video content and the selected recommendation tags to the server for publishing after receiving the publishing request.

[0139] Corresponding to the third embodiment, the present application embodiment also provides a device for publishing commodity object information, see Figure 8 , the device may include:

[0140] The second operation option providing unit 801 is used to provide an operation option for obtaining recommended tag information in the product object publishing interface;

[0141] A second recommended tag acquisition unit 802 is used to acquire recommended tags according to the video content associated with the commodity object after receiving a request through the operation option, wherein the recommended tags are obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content and determine the recommended tags of the target video content by performing similarity comparison with the features of the image templates corresponding to the multiple tags;

[0142] The second recommendation tag display unit 803 is used to display the recommendation tags provided by the server and provide operation options for selecting the recommendation tags;

[0143] The second publishing unit 804 is used to submit the information of the commodity object and the selected recommendation tag to the server for publishing after receiving the publishing request.

[0144] The commodity objects include: catering or fresh food commodity objects that are delivered to the user-specified address at the user-specified time within the target area around the associated physical store.

[0145] In addition, the device may also include:

[0146] The intelligent filling unit is used to obtain basic information associated with the commodity object and automatically fill it into the corresponding input control for publishing. The basic information includes associated physical store information, location information, inventory information, and / or validity period information.

[0147] 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.

[0148] And an electronic device, comprising:

[0149] one or more processors; and

[0150] 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.

[0151] in, Fig. 9 The architecture of the electronic device is shown as an example, which may include a processor 910, a video display adapter 911, a disk drive 912, an input / output interface 913, a network interface 914, and a memory 920. The processor 910, the video display adapter 911, the disk drive 912, the input / output interface 913, the network interface 914, and the memory 920 may be communicatively connected via a communication bus 930.

[0152] Among them, the processor 910 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solution provided in this application.

[0153] The memory 920 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 920 can store an operating system 921 for controlling the operation of the electronic device 900, and a basic input and output system (BIOS) for controlling the low-level operation of the electronic device 900. In addition, a web browser 923, a data storage management system 924, and a video information processing system 925, etc. can also be stored. The above-mentioned video information processing system 925 can be an application program that specifically implements the operations of the aforementioned steps in the embodiment of the present application. In short, when the technical solution provided by the present application is implemented by software or firmware, the relevant program code is stored in the memory 920 and is called and executed by the processor 910.

[0154] The input / output interface 913 is used to connect the input / output module to realize information input and output. The input / output module can be configured in the device as a component (not shown in the figure), or it can be externally connected to the device to provide corresponding functions. The input device may include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device may include a display, a speaker, a vibrator, an indicator light, etc.

[0155] The network interface 914 is used to connect to a communication module (not shown) to realize communication interaction between the device and other devices. The communication module can realize communication through a wired mode (such as USB, network cable, etc.) or a wireless mode (such as mobile network, WIFI, Bluetooth, etc.).

[0156] The bus 930 comprises a pathway for transmitting information between the various components of the device (eg, the processor 910, the video display adapter 911, the disk drive 912, the input / output interface 913, the network interface 914, and the memory 920).

[0157] It should be noted that, although the above device only shows a processor 910, a video display adapter 911, a disk drive 912, an input / output interface 913, a network interface 914, a memory 920, a bus 930, etc., in the specific implementation process, the device may also include other components necessary for normal operation. In addition, it can be understood by those skilled in the art that the above device may also only include components necessary for implementing the solution of the present application, and does not necessarily include all the components shown in the figure.

[0158] 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.

[0159] 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.

[0160] The above is a detailed introduction to the method, device and electronic device for publishing video content and commodity object information provided by this application. This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this 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 this application.

Claims

1. A method for publishing video content, characterized in that: include: Providing an operation option for obtaining recommended tag information in the video content publishing interface; After receiving the request through the operation option, the recommended tag is obtained according to the associated video content; wherein the video content is used as the input information of the prediction model to obtain the recommended tag, and the prediction model is used to extract features of the video content, and determine the recommended tag of the video content by performing similarity comparison with the features of the template image corresponding to multiple tags in the tag library; the template image feature information stored in the prediction model includes: template image feature information corresponding to multiple word segments, wherein the word segmentation is determined after word segmentation and deduplication of multiple tags in the tag library, and the prediction model also stores the mapping relationship information between the word segmentation and the tag; the prediction model is specifically used to: when predicting the video content, determine the features of the template image corresponding to the tag according to the template image features corresponding to the word segmentation and the mapping relationship information; Displaying the recommended tags and providing operation options for selecting the recommended tags; After receiving the publishing request, the video content and the selected recommendation tags are submitted to the server for publishing.

2. The method according to claim 1, characterized in that The template image feature information expressed by the multiple word segments is determined in the process of training the prediction model.

3. The method according to claim 1, characterized in that The mapping relationship information between the participles and the labels is determined in the process of training the prediction model. In the mapping relationship, multiple different participles corresponding to the same label are associated with different weights, which are used to represent the different importance of different participles when the label expresses the template image.

4. A method for publishing commodity object information, characterized in that: include: Provide an operation option for obtaining recommended tag information in the product object publishing interface; After receiving the request through the operation option, a recommended tag is obtained according to the video content associated with the commodity object, wherein the recommended tag is obtained by taking the video content as the input information of the prediction model, and the prediction model is used to extract features of the video content, and determine the recommended tag of the video content by performing a similarity comparison with the features of the template image corresponding to the multiple tags in the tag library; the template image feature information stored in the prediction model includes: template image feature information corresponding to multiple word segments, wherein the word segmentation is determined after word segmentation and deduplication of multiple tags in the tag library, and the prediction model also stores the mapping relationship information between word segmentation and tag; the prediction model is specifically used to: when predicting the video content, determine the features of the template image corresponding to the tag according to the template image features corresponding to the word segmentation and the mapping relationship information; Displaying the recommended tags provided by the server and providing operation options for selecting the recommended tags; After receiving the publishing request, the information of the product object and the selected recommendation tags are submitted to the server for publishing.

5. The method according to claim 4, characterized in that The template image feature information expressed by the multiple word segments is determined in the process of training the prediction model.

6. The method according to claim 4, characterized in that The mapping relationship information between the participles and the labels is determined in the process of training the prediction model. In the mapping relationship, multiple different participles corresponding to the same label are associated with different weights, which are used to represent the different importance of different participles when the label expresses the template image.

7. The method according to claim 4, characterized in that Also includes: The commodity objects include: catering or fresh food commodity objects that are delivered to the user-specified address at the user-specified time within the target area around the associated physical store.

8. The method according to claim 7, characterized in that Also includes: The basic information associated with the commodity object is obtained and automatically filled into the corresponding input control for publishing, wherein the basic information includes the associated physical store information, location information, inventory information, and / or validity period information.

9. A method for generating a video content tag, characterized in that: include: Acquire a prediction model, wherein the prediction model stores feature information of the template image, so as to determine features of the template image corresponding to the multiple tags in the tag library; The template image feature information stored in the prediction model includes: template image feature information corresponding to a plurality of word segments, wherein the word segments are determined after word segmentation and deduplication of a plurality of tags in a tag library, and the prediction model also stores mapping relationship information between word segments and tags; the prediction model is specifically used to: when predicting the video content, determine the features of the template image corresponding to the tag according to the template image features corresponding to the word segments and the mapping relationship information; Determining target video content to be predicted; At least one recommended tag for the target video content is determined based on the prediction model, wherein the prediction model is used to extract features of the target video content and determine the recommended tag for the target video content by performing similarity comparison with features of a template image corresponding to the multiple tags.

10. The method according to claim 9, characterized in that The template image feature information expressed by the multiple word segments is determined in the process of training the prediction model.

11. The method according to claim 9, characterized in that The mapping relationship information between the participles and the labels is determined in the process of training the prediction model. In the mapping relationship, multiple different participles corresponding to the same label are associated with different weights, which are used to represent the different importance of different participles when the label expresses the template image.

12. The method according to claim 9, characterized in that The features of the target video content and the features of the template image expressed by the label are feature vectors with the same dimensional distribution; The prediction model is specifically used to determine the probability that the target video content belongs to the corresponding label by calculating the inner product of the feature vector of the target video content and the feature vector of the template image expressed by the label.

13. The method according to claim 9, characterized in that The step of determining the target video content to be predicted includes: When a request for publishing video content submitted by a client is received, the video content to be published is determined as the target video content to be predicted.

14. The method according to claim 13, characterized in that Also includes: The recommended tags are provided to the client so that the video content to be published is published after being associated with tag information from the recommended tags.

15. The method according to claim 9, characterized in that The step of determining the target video content to be predicted includes: When a request for publishing the commodity object information submitted by the client is received, the video content associated with the commodity object to be published is determined as the target video content to be predicted.

16. The method according to claim 13, characterized in that Also includes: The recommended tags are provided to the client so that the commodity object to be published is associated with tag information from the recommended tags and then published.

17. The method according to any one of claims 9 to 16, characterized in that: Also includes: The attribute information of the commodity object associated with the target video content is combined with the output result of the prediction model to determine the recommended tag information.

18. A method for training a prediction model, characterized in that: include: Obtaining a training sample, wherein the training sample includes video content and corresponding label information; Training the prediction model using training data samples; Among them, in each round of iterative training process, the feature information of the video content is extracted, and the template image features corresponding to multiple tags in the tag library are obtained, and the probability that the video content belongs to the category corresponding to the multiple tags is determined according to the similarity between the features of the video content and the template image features corresponding to the tags, and the parameters in the prediction model are updated according to the tag information in the sample; the template image feature information stored in the prediction model includes: template image feature information corresponding to multiple word segments, wherein the word segmentation is determined after word segmentation and deduplication of multiple tags in the tag library, and the prediction model also stores mapping relationship information between word segmentations and tags; the prediction model is specifically used for: when predicting the video content, according to the template image features corresponding to the word segmentations and the mapping relationship information, determining the features of the template image corresponding to the tag.

19. The method according to claim 18, characterized in that Also includes: Segment and remove duplicates of multiple tags in the tag library to obtain a segmentation set, and establish a mapping relationship between the tags and the segmentations; The step of obtaining the template image features respectively expressed by the multiple tags in the tag library includes: The template image features respectively expressed by the multiple labels are determined according to the template image features expressed by the segmentation words and the mapping relationship between the label and the segmentation words.

20. The method according to claim 19, characterized in that The template image features expressed by the word segmentation are random values ​​in an initial state and are updated during the iterative training of the prediction model.

21. The method according to claim 19, characterized in that Also includes: During the iterative training of the prediction model, the mapping relationship between the label and the participle is updated, wherein in the updated mapping relationship, multiple different participles corresponding to the same label are associated with different weights, which are used to represent the different importance of different participles when the label expresses the template image.

22. The method according to claim 18, characterized in that The obtaining of training samples comprises: Obtain multiple video contents and corresponding initial label information; Clustering the video content according to the commodity object information associated with the video content; The initial labels corresponding to the same type of video content are counted and the low-frequency initial labels are filtered out.

23. A device for publishing video content, characterized in that: include: A first operation option providing unit, configured to provide an operation option for obtaining recommended tag information in a video content publishing interface; A first recommendation tag acquisition unit is used to acquire a recommendation tag according to the associated video content after receiving a request through the operation option; wherein the recommendation tag is obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content, and determine the recommendation tag of the video content by performing a similarity comparison with features of template images corresponding to multiple tags in a tag library; the template image feature information stored in the prediction model includes: template image feature information corresponding to multiple word segments, wherein the word segmentation is determined after word segmentation and deduplication of multiple tags in the tag library, and the prediction model also stores mapping relationship information between word segments and tags; the prediction model is specifically used to: when predicting the video content, determine the features of the template image corresponding to the tag according to the template image features corresponding to the word segmentation and the mapping relationship information; A first recommended tag display unit, used to display the recommended tags and provide operation options for selecting the recommended tags; The first publishing unit is used to submit the video content and the selected recommendation tags to the server for publishing after receiving the publishing request.

24. A device for publishing commodity object information, characterized in that: include: A second operation option providing unit, configured to provide an operation option for obtaining recommended tag information in the commodity object publishing interface; A second recommendation tag acquisition unit is used to acquire a recommendation tag according to the video content associated with the commodity object after receiving a request through the operation option, wherein the recommendation tag is obtained by taking the video content as input information of a prediction model, and the prediction model is used to extract features of the video content, and determine the recommendation tag of the video content by performing a similarity comparison with features of template images corresponding to multiple tags in a tag library; the template image feature information stored in the prediction model includes: template image feature information corresponding to multiple word segments, wherein the word segmentation is determined after word segmentation and deduplication of multiple tags in the tag library, and the prediction model also stores mapping relationship information between word segments and tags; the prediction model is specifically used to: when predicting the video content, determine the features of the template image corresponding to the tag according to the template image features corresponding to the word segmentation and the mapping relationship information; A second recommendation tag display unit, used to display the recommendation tags provided by the server and provide operation options for selecting the recommendation tags; The second publishing unit is used to submit the information of the commodity object and the selected recommendation tag to the server for publishing after receiving the publishing request.

25. A device for generating a video content tag, characterized in that: include: A prediction model acquisition unit, used to acquire a prediction model, wherein the prediction model stores feature information of the template image, so as to determine features of the template image corresponding to the multiple tags in the tag library; The template image feature information stored in the prediction model includes: template image feature information corresponding to a plurality of word segments, wherein the word segments are determined after word segmentation and deduplication of a plurality of tags in a tag library, and the prediction model also stores mapping relationship information between word segments and tags; the prediction model is specifically used to: when predicting the video content, determine the features of the template image corresponding to the tag according to the template image features corresponding to the word segments and the mapping relationship information; A video content determination unit, used to determine the target video content to be predicted; A recommended tag determination unit is used to determine at least one recommended tag for the target video content based on the prediction model, wherein the prediction model is used to extract features of the target video content and determine the recommended tag for the target video content by performing a similarity comparison with features of a template image corresponding to the multiple tags.

26. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 22 are implemented.

27. An electronic device, 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 the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of claims 1 to 22.

Images