Advertisement picture generation method and device, storage medium and program product

Abstract

The embodiment of the invention provides an advertisement picture generation method and device, a storage medium and a program product. In the method, advertisement demand information of a user can be acquired; generating structured data corresponding to the plurality of page layer layout modes by utilizing an advertisement template generation model; determining a candidate background picture matched with the advertisement demand information from an advertisement material library, and correcting the candidate background picture by using a picture-to-picture model to obtain a background picture corresponding to the background picture layer; generating a foreground picture matched with the advertisement demand information by using a text graph model; the target foreground picture corresponds to at least one foreground picture layer; running a resource document of the target page to load the target background picture and the target foreground picture to the target page according to the target structured data to obtain a plurality of candidate advertisement pictures; and determining a target advertisement picture from the plurality of candidate advertisement pictures. In this way, the advertisement picture better meeting the advertisement requirement of the user can be efficiently generated.

Description

Technical Field

[0001] The present application relates to the field of image generation technology, and in particular to an advertising picture generation method, device, storage medium and program product. Background Art

[0002] With the rapid development of e-commerce, merchants have an increasing demand for efficient and personalized advertising. The traditional advertising design process is time-consuming and costly, and it is difficult to meet the rapidly changing market and consumer needs. In the prior art, the generation of advertising images is too dependent on manual design, resulting in low efficiency in generating advertising images, which is difficult to meet the needs of merchants. Summary of the invention

[0003] Multiple aspects of the present application provide an advertising image generation method, device, storage medium and program product for more efficiently generating advertising images that better meet user advertising needs.

[0004] The embodiment of the present application provides an advertisement image generation method, comprising: in response to an advertisement generation instruction of a user, obtaining advertisement demand information of the user; the advertisement demand information is information required for the user to request to provide advertisement generation service for its target object; using an advertisement template generation model, generating target structured data corresponding to multiple page layer layout modes in a resource document of a target page; each page layer layout mode is used to describe a layout mode between a background layer and at least one foreground layer; determining a candidate background image matching the advertisement demand information from an advertisement material library, and modifying the candidate background image using a graph-generated graph model to obtain a target background image corresponding to the background layer; in the case that at least one target foreground image matching the advertisement demand information does not exist in the advertisement material library, generating at least one target foreground image matching the advertisement demand information using a graph-generated graph model; the at least one target foreground image corresponds to the at least one foreground layer; running the resource document of the target page, so as to load the target background image and the at least one target foreground image to the target page according to the target structured data in different page layer layout modes, so as to obtain multiple candidate advertisement images; in response to the user's selection operation on the multiple candidate advertisement images, determining a target advertisement image from the multiple candidate advertisement images.

[0005] Further optionally, the advertisement template generation model is used to generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page, including: using the advertisement template generation model to parse the description information of a preset domain feature language to obtain the layer layout information of the target page; under the constraints of the layer layout information of the target page, generating target structured data corresponding to the multiple page layer layout modes in the resource document.

[0006] Further optionally, under the constraint of the layer layout information of the target page, target structured data corresponding to the multiple page layer layout modes are generated in the resource document, including: under the constraint of the layer layout information of the target page, initial structured data corresponding to the multiple page layer layout modes are generated; based on the initial structured data corresponding to the multiple page layer layout modes, multiple page layers corresponding to each of the multiple page layer layout modes are displayed in a graphics processing platform for layer layout adjustment; in response to a page layer adjustment operation, the initial structured data corresponding to the multiple page layer layout modes are adjusted to obtain target structured data corresponding to the multiple page layer layout modes; and the target structured data corresponding to the multiple page layer layout modes are added to the resource document of the target page.

[0007] Further optionally, it also includes: determining a target website and / or target database to be traversed that matches the user category based on the user category information to which the user belongs; using a web crawler to capture images in the target website and / or the target database; using an image recognition model to identify target images that match the advertising demand information from images in the target website and / or the target database, and adding the target images as background images to the advertising material library.

[0008] Further optionally, the candidate background image is corrected using a graph-generated image model to obtain a target background image corresponding to the background layer, including: using the graph-generated image model to generate feature correction information; performing feature extraction on the candidate background image to obtain a feature vector of the candidate background image; fusing the feature vector with the feature correction information to obtain a fused feature vector; and generating the target background image based on the fused feature vector.

[0009] Further optionally, the graph-generated image model includes: a feature extractor, a prompt word generator, a text encoder, a diffusion generator and an image decoder; the candidate background image is corrected by using the graph-generated image model to obtain a target background image corresponding to the background layer, including: using the feature extractor to extract a feature vector of the candidate background image; using the prompt word generator to generate a plurality of scene prompt words according to the advertising demand information; using the text encoder to perform vector embedding according to the plurality of scene prompt words to obtain semantic vectors corresponding to the plurality of scene prompt words; inputting the semantic vectors corresponding to the plurality of scene prompt words into the diffusion generator for denoising processing to obtain feature vectors corresponding to the plurality of scene prompt words; using the image decoder to generate a target background image according to the feature vectors corresponding to the plurality of scene prompt words and the feature vectors of the candidate background image.

[0010] Further optionally, it also includes: generating a corresponding demand description text according to the advertising demand information; using a text graph model to generate at least one target foreground image matching the advertising demand information, including: inputting the demand description text into a feature extraction layer in the text graph model, extracting features from the demand description text to obtain text features; inputting the text features into an image generation layer in the text graph model, and generating at least one target foreground image matching the advertising demand information according to the text features.

[0011] Further optionally, the resource document of the target page is run to load the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes to obtain multiple candidate advertising images, including: using a headless browser module that simulates a browser environment to run the resource document of the target page, and loading the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes to obtain multiple advertising pages, each advertising page including an advertising area generated by laying out the target background image and the at least one target foreground image according to a corresponding page layer layout mode; taking a screenshot of the advertising area in any advertising page to obtain any candidate advertising image.

[0012] An embodiment of the present application also provides an electronic device, comprising: a memory and a processor; wherein the memory is used to: store one or more computer instructions; the processor is used to execute the one or more computer instructions to: execute the steps in the advertising image generation method.

[0013] The embodiment of the present application also provides a computer-readable storage medium, which, when the computer program is executed by a processor, enables the processor to implement the steps in the advertising image generation method.

[0014] The embodiment of the present application also provides a computer program product, including a computer program / instruction. When the computer program / instruction is executed by a processor, the processor is enabled to implement the steps in the advertising image generation method.

[0015] In this embodiment, the advertising template generation model can be used to generate a richer page layer layout method, and the image-generated image model and the text-generated image model can be used respectively to more accurately generate background images and foreground images that meet the user's advertising needs; thereby, based on the page layer layout method, background images and foreground images, advertising images that better meet the user's advertising needs can be generated more efficiently. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 A schematic diagram of a flow chart of a method for generating an advertisement image provided by an exemplary embodiment of the present application;

[0018] Figure 2 A schematic diagram of a demand acquisition page provided for an exemplary embodiment of the present application;

[0019] Figure 3 An architectural diagram of an advertising image generation platform provided for an exemplary embodiment of the present application;

[0020] Figure 4 A schematic diagram of a flow chart of a method for generating an advertisement image in a practical scenario provided by an exemplary embodiment of the present application;

[0021] Figure 5 Another schematic diagram of a flow chart of a method for generating an advertisement image in a practical scenario provided by an exemplary embodiment of the present application;

[0022] Figure 6 A schematic diagram of an interface of a graphics processing platform provided by an exemplary embodiment of the present application;

[0023] Figure 7 A schematic diagram of an electronic device provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0025] 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 portals for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) are in compliance with relevant laws and standards.

[0026] With the rapid development of e-commerce, merchants have a growing demand for efficient and personalized advertising. The traditional advertising design process is time-consuming and costly, and it is difficult to meet the rapidly changing market and consumer needs. At present, the generation of advertising images is too dependent on manual design, resulting in low efficiency in generating advertising images and failing to meet the needs of merchants.

[0027] In view of the technical problems existing in the above-mentioned prior art, a solution is provided in some embodiments of the present application. The technical solutions provided in each embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0028] Figure 1 This is a flowchart of a method for generating an advertisement image provided by an exemplary embodiment of the present application. This embodiment can be executed by a terminal device, such as a smart phone, a tablet computer, a computer, etc., or by any type of server, and this embodiment is not limited. Figure 1 As shown, the method includes:

[0029] Step 11: Respond to the user's advertisement generation instruction and obtain the user's advertisement demand information; the advertisement demand information is the information required by the user to request to provide advertisement generation service for its target object.

[0030] Step 12: Generate a model using an advertisement template to generate target structured data corresponding to multiple page layer layout modes in a resource document of a target page; each page layer layout mode is used to describe a layout mode between a background layer and at least one foreground layer.

[0031] Step 13: Determine a candidate background image that matches the advertising demand information from the advertising material library, and use the graph-to-graph model to modify the candidate background image to obtain a target background image corresponding to the background layer.

[0032] Step 14: if there is no at least one target foreground image matching the advertising demand information in the advertising material library, generate at least one target foreground image matching the advertising demand information using the text graph model; the at least one target foreground image corresponds to at least one foreground layer.

[0033] Step 15: Run the resource document of the target page to load the target background image and at least one target foreground image into the target page according to the target structured data and in different page layer layout modes to obtain a plurality of candidate advertisement images.

[0034] Step 16: In response to the user's selection operation on a plurality of candidate advertising images, a target advertising image is determined from the plurality of candidate advertising images.

[0035] In this embodiment, the user may issue an advertisement generation instruction by voice, manual click, eye movement, or other triggering methods, which are not limited in this embodiment. Correspondingly, the user's advertisement generation instruction may be responded to to obtain the user's advertisement demand information, which is the information required by the user to request to provide advertisement generation services for its target object, for example, it may be the relevant information required to generate an advertisement image. The advertisement demand information may include, but is not limited to: description information and introduction information of the target object, and information on the requirements for the advertisement background and foreground, etc.

[0036] For example, the advertising demand information may include the following descriptive information: store label "warehouse", store name "A", city "Beijing" and industry "cleaning", and also includes the following introduction information: "This small shop integrates life and work, carrying dreams and passion", and also includes the following requirements for the advertising background and foreground: the background needs high contrast and the foreground needs a striking effect.

[0037] Afterwards, the advertisement template generation model can be used to generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page. The advertisement template generation model can be any natural language processing model, and the natural language processing model can be a machine learning model or a deep learning model for understanding and generating human language, which is not limited in this embodiment. The advertisement template generation model has been pre-trained and has the ability to generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page.

[0038] The target page can be any page, such as a web page, an application page, or a mini-program page, etc., which is not limited in this embodiment. The resource document of the target page is used to construct the target page, which can be understood as a collection of information resources related to the target page. For example, when the target page is a web page, the resource document of the target page can be an HTML (HyperText Markup Language) document.

[0039] The advertisement template generation model can generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page. Each page layer layout mode is used to describe a layout mode between a background layer and at least one foreground layer. Structured data refers to data with a clear format and organization form, which can be Json (JavaScript Object Notation, JS key-value pair data) data or other data, which is not limited in this embodiment.

[0040] Take the target page as a web page and the resource document as an HTML document as an example: the advertisement template generation model can generate Json data corresponding to two page layer layout modes in the backend, that is, in the HTML document of the web page. In one of the page layer layout modes, the spacing between the background layer and at least one foreground layer is large; in the other page layer layout mode, the spacing between the background layer and at least one foreground layer is small.

[0041] The page layer in this embodiment may include a background layer and at least one foreground layer. The background layer may be used to load a target background image, and the at least one foreground layer may be used to load at least one target foreground image. In this embodiment, the target background image and the target foreground image are generated in different ways, which will be described below in different cases:

[0042] How to generate the target background image:

[0043] Determine a candidate background image that matches the advertising demand information from the advertising material library, and use the graph-generated graph model to correct the candidate background image to obtain a target background image corresponding to the background layer. The advertising material library may store multiple background images and multiple foreground images. The graph-generated graph model is used to convert an input image into an output image with different attributes or styles. In this way, the graph-generated graph model can be used to correct the candidate background image to a target background image that better meets the user's advertising needs.

[0044] The target foreground image is generated as follows:

[0045] In the case that there is no at least one target foreground image matching the advertising demand information in the advertising material library, the text-generated graph model is used to generate at least one target foreground image matching the advertising demand information. The at least one target foreground image corresponds to at least one foreground layer. The graph-generated graph model is used to generate a corresponding image according to given text information. In this way, the text-generated graph model can be used to generate a target foreground image that better meets the user's advertising demand according to the advertising demand information.

[0046] It should be noted that the target background image may be generated first and then the target foreground image, or the target foreground image may be generated first and then the target background image, or the target background image and the target foreground image may be generated at the same time, which is not limited in this embodiment.

[0047] Based on the above steps, the resource document of the target page can be run to load the target background image and at least one target foreground image to the target page according to the target structured data in different page layer layout modes to obtain multiple candidate advertising images. Among them, since there are target structured data corresponding to multiple page layer layout modes in the resource document of the target page, and the target foreground image corresponds to the foreground layer, and the target background image corresponds to the background layer, therefore, when the resource document is run, the target structured data corresponding to each page layer layout mode can obtain the corresponding target foreground image and target background image, and record the target foreground image and target background image on the target page according to the page layer layout mode. Based on this, multiple candidate advertising images can be generated on the target page for users to choose. The layout mode between the target background image and at least one target foreground image is different in different candidate advertising images.

[0048] The user may select a candidate advertisement picture from a plurality of candidate advertisement pictures. Correspondingly, in response to the user's selection operation on the plurality of candidate advertisement pictures, a target advertisement picture may be determined from the plurality of candidate advertisement pictures.

[0049] In this embodiment, the advertising template generation model can be used to generate a richer page layer layout method, and the image-generated image model and the text-generated image model can be used respectively to more accurately generate background images and foreground images that meet the user's advertising needs; thereby, based on the page layer layout method, background images and foreground images, advertising images that better meet the user's advertising needs can be generated more efficiently.

[0050] In some optional embodiments, step 11 of the aforementioned embodiment, "responding to the user's ad generation instruction and obtaining the user's ad demand information", can be implemented based on the following methods: providing a demand acquisition page to the user for the user to fill in multiple demand items for ad generation; responding to the user's information input operation, extracting information from the demand acquisition page to generate ad demand information based on the extracted user input information. The multiple demand items can be used to obtain at least one of the description information, introduction information, and ad background and foreground requirement information of the target object. Figure 2 As shown, the demand acquisition page displays demand items such as store name, city, industry, store label, and store introduction for users to fill in. In this way, the user's advertising demand information can be obtained more accurately through the demand acquisition page containing multiple demand items displayed to the user.

[0051] In some optional embodiments, step 12 in the aforementioned embodiment, "using the advertisement template generation model to generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page", can be implemented based on the following steps:

[0052] Step 121: Utilize the advertisement template generation model to parse the description information of the preset domain-specific feature language (DSFL) to obtain the layer layout information of the target page.

[0053] It should be noted that DSL is a programming language or specification language specially designed for a specific application domain, allowing developers to express solutions in a language that is closer to the problem domain. Based on the description information of the domain feature language, the advertisement template generation model generates more concise structured data. For example, assuming that the structured data generated by the advertisement template generation model directly without using the description information of the domain feature language may be 500 lines, then when the description information of the domain feature language is used, the structured data generated by the advertisement template generation model can be far less than 500 lines, which significantly improves the efficiency of generating structured data of the advertisement template generation model.

[0054] The description information of the domain feature language can be used to instruct the advertisement template generation model how to generate the required structured data. In other words, the description information of the domain feature language can be used to instruct the advertisement template generation model on the layout rules between the various layers of the target page, such as the position and identification rules of the layers.

[0055] The advertisement template generation model can parse the description information to obtain the layer layout information of the target page, and the layer layout information can be used to indicate the layout rules between the layers of the target page. For example, the layer layout information can be used to indicate the layout range of each layer on the target page, for example, a layer can be located at any position in a certain area of ​​the target page, but cannot exceed the area.

[0056] Step 122: Under the constraint of the layer layout information of the target page, generate target structured data corresponding to multiple page layer layout modes in the resource document. Among them, the target structured data corresponding to multiple page layer layout modes can be randomly generated in the resource document under the constraint of the layer layout information. In other words, since the layer layout information can indicate the layout range of each layer on the target page, multiple random positions of each layer on the target page can be determined in a random manner under the constraint of the layout range of each layer on the target page, so that multiple page layer layout modes can be obtained, and then the target structured data corresponding to multiple page layer layout modes can be generated.

[0057] Through the above steps 121 and 122, the advertisement template generation model can be used to more efficiently generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page based on the description information of the domain feature language.

[0058] Optionally, when generating target structured data corresponding to multiple page layer layout modes in a resource document, a fine-tuning function for the page layer layout mode may be provided to the user or other relevant personnel to generate an advertising template that better meets the user's advertising needs. Specifically, the above step 122 may be implemented based on the following steps R1 to R5:

[0059] Step R1: Under the constraints of the layer layout information of the target page, generate initial structured data corresponding to multiple page layer layout modes. The initial structured data refers to the structured data generated by the advertisement template generation model and not modified.

[0060] Step R2: Based on the initial structured data corresponding to the multiple page layer layout modes, the multiple page layers corresponding to the multiple page layer layout modes are displayed in the graphics processing platform for layer layout adjustment. The graphics processing platform can be any platform for layer layout adjustment, and this embodiment does not limit it. In other words, the initial structured data generated by the advertisement template generation model can be sent to the graphics processing platform for layer layout adjustment of different advertisement templates.

[0061] Step R3, in response to the page layer adjustment operation, the initial structured data corresponding to the multiple page layer layout modes are adjusted to obtain the target structured data corresponding to the multiple page layer layout modes. For example, in response to the position movement operation for a certain layer, the structured data corresponding to the new target position of the layer can be generated in the resource document to replace the structured data corresponding to the original target position of the layer.

[0062] Step R4, adding the target structured data corresponding to the layout modes of the multiple page layers to the resource document of the target page. According to the correspondence between the preset target structured data and the target position in the resource document, the target structured data corresponding to the layout modes of the multiple page layers can be added to the target position in the resource document of the target page.

[0063] Through the above steps R1-R4, the fine-tuning function for the page layer layout method can be provided to users or other relevant personnel, thereby generating an advertising template that better meets the user's advertising needs.

[0064] In some optional embodiments, the advertisement template generation model may be trained based on the following method:

[0065] Sample description information and sample structured data corresponding to multiple sample page layer layout modes are obtained; the sample description information and the sample structured data corresponding to multiple sample page layer layout modes are input into an advertisement template generation network, and under the supervision of the sample structured data corresponding to multiple sample page layer layout modes, the sample description information is used to train the advertisement template generation network with the goal of converging the loss function of the preset advertisement template generation network to a specified range, so as to obtain an advertisement template generation model. The loss function can be used to characterize the error between the structured data corresponding to multiple page layer layout modes generated by the advertisement template generation network according to the sample description information and the sample structured data corresponding to multiple sample page layer layout modes.

[0066] In this way, the advertisement template generation model can be trained more efficiently, so that the advertisement template generation model has the ability to generate structured data.

[0067] In some optional embodiments, before determining the candidate background images matching the advertisement demand information from the advertisement material library, background images that better meet the user's advertisement demand may be pre-acquired and filled into the advertisement material library in the following manner:

[0068] Method 1: According to the user category information to which the user belongs, determine the target website to be traversed that matches the user category; use a web crawler to grab pictures from the target website; use an image recognition model to identify target pictures that match the advertising demand information from the pictures in the target website, and add the target pictures as background pictures to the advertising material library.

[0069] Method 2: According to the user category information to which the user belongs, determine the target database to be traversed that matches the user category; use a web crawler to grab pictures in the target database; use an image recognition model to identify target pictures that match the advertising demand information from the pictures in the target database, and add the target pictures as background pictures to the advertising material library.

[0070] Method 3: According to the user category information to which the user belongs, determine the target website and target database to be traversed that match the user category; use a web crawler to capture images in the target website and target database; use an image recognition model to identify target images that match the advertising demand information from the images in the target website and target database, and add the target images as background images to the advertising material library.

[0071] The user category information of the user can be used to indicate at least one of the industry, occupation, region and gender of the user. According to the user category information of the user, a plurality of websites and databases can be preliminarily screened to determine a data source that is more suitable for the user, that is, at least one of a target website and a target database.

[0072] Among them, a web crawler is an automated program that can be used to automatically browse websites and databases on the Internet and download content from them.

[0073] Among them, when the image recognition model identifies the target image that matches the advertising demand information from the images in the target website and the target database, the image and the advertising demand information in the target website and the target database can be input into the feature extraction layer to extract the semantic features of the image and the semantic features of the advertising demand information, and the semantic features of the image and the semantic features of the advertising demand information can be input into the recognition layer to identify the target image that matches the advertising demand information based on the feature similarity between the semantic features of the image and the semantic features of the advertising demand information.

[0074] The image recognition model can be trained in the following manner: obtain sample images captured by a web crawler, sample advertising demand information, and target images that match the sample advertising demand information; input the sample images and sample advertising demand information into the image recognition network, and under the supervision of the target image that matches the sample advertising demand information, train the image recognition network with the sample images and sample advertising demand information with the goal of converging the loss function of the preset image recognition network to a specified range, thereby obtaining the image recognition model. The loss function can be used to characterize: the error between the sample image that matches the sample advertising demand information and the target image determined by the image recognition network from the sample images captured by the web crawler.

[0075] Through the above methods 1 to 3, based on the web crawler and image recognition model, it is possible to more efficiently obtain background images that better meet the user's advertising needs and fill them into the advertising material library.

[0076] Based on this, the background image obtained above can be determined as a candidate background image matching the advertisement demand information in a subsequent process.

[0077] In some optional embodiments, step 13 in the above embodiment, "correcting the candidate background image using the graph-generated graph model to obtain a target background image corresponding to the background layer" can be implemented based on the following implementation methods:

[0078] In the first embodiment, the following steps K1-K4 may be included:

[0079] Step K1, using the image-generated image model to generate feature correction information. The feature correction information can be used to describe the feature offset for the candidate background image. The feature offset can be randomly determined to generate feature correction information corresponding to the feature offset, or a preset default feature offset can be obtained to generate corresponding feature correction information according to the default feature offset.

[0080] Step K2: extract features from the candidate background image to obtain a feature vector of the candidate background image. The candidate background image may be input into a feature extraction layer in the graph generation model to extract the feature vector of the candidate background image.

[0081] Step K3: Fusing the feature vector with the feature correction information to obtain a fused feature vector. The feature vector and the feature correction information may be input into a fusion layer in the graph generation model to perform vector shifting on the feature vector using the feature correction information to obtain a fused feature vector.

[0082] Step K4: Generate a target background image based on the fused feature vector. The fused feature vector may be input into a fully connected layer in the graph-generated graph model to map the fused feature vector into the target background image.

[0083] Through the above implementation mode 1, the candidate background image can be modified more efficiently using the graph-generated graph model to obtain the target background image corresponding to the background layer.

[0084] In the second implementation, the graph model may include: a feature extractor, a prompt word generator, a text encoder, a diffusion generator, and an image decoder. Based on this, the second implementation can be implemented based on the following steps L1-L6:

[0085] Step L1: Use a feature extractor to extract a feature vector of the candidate background image.

[0086] Step L2: Generate multiple scene prompt words according to the advertising demand information using a prompt word generator. The scene prompt words are used to describe the scene where the target object is located. That is, the prompt word generator can determine the scene where the target object is located based on the advertising demand information and generate scene prompt words corresponding to the scene. For example, if the advertising demand information includes a store name, a store introduction, etc., the target object may be in an e-commerce scene, and the scene prompt words can be used to describe the e-commerce scene. There can be one or more scene prompt words, which are not limited in this embodiment.

[0087] Step L3: Using a text encoder, perform vector embedding according to the multiple scene prompt words to obtain semantic vectors corresponding to the multiple scene prompt words. Vector embedding refers to the operation of converting multiple scene prompt words into continuous numerical vectors.

[0088] Step L4, input the semantic vectors corresponding to the multiple scene prompt words into the diffusion generator for denoising, so as to obtain the feature vectors corresponding to the multiple scene prompt words. The diffusion generator is an image generator based on the diffusion process, which has the following capabilities: under the guidance of the input generation conditions, it can continuously perform denoising from a random noise vector, and finally obtain a feature vector that meets the generation condition requirements. The semantic vectors corresponding to the multiple scene prompt words can be used as the generation condition, and under the guidance of the generation condition, denoising is continuously performed from a random noise vector, and finally the feature vectors corresponding to the multiple scene prompt words that meet the generation condition requirements are obtained.

[0089] In this embodiment, the number of times the random noise vector is denoised is not limited, for example, it can be 30-50 times. In each denoising process, semantic vectors corresponding to multiple scene prompt words are continuously injected into the noise vector to obtain a feature vector with rich semantic information, which can also be called a latent space vector.

[0090] Step L5: Generate a target background image using an image decoder based on the feature vectors corresponding to the multiple scene prompt words and the feature vectors of the candidate background images, wherein the image decoder has been pre-trained to learn the mapping relationship between the feature vectors corresponding to the multiple scene prompt words and the target background image.

[0091] Through the above implementation mode 2, a relatively rich plurality of scene prompt words can be generated based on the prompt word generator, and denoising processing can be performed based on the diffusion generator to more accurately obtain feature vectors corresponding to the plurality of scene prompt words, thereby combining the scene in which the target object is located to more accurately generate a richer target background image.

[0092] Optionally, the graph-generated graph model can be trained in the following manner: obtain a sample background image and a sample target background image; input the sample background image into the image recognition network, and under the supervision of the sample target background image, train the graph-generated graph network with the goal of converging the loss function of the preset graph-generated graph network to a specified range, thereby obtaining the graph-generated graph model. The loss function can be used to characterize the error between the target background image obtained by correcting the sample background image with the graph-generated graph network and the sample target background image.

[0093] In some optional embodiments, corresponding demand description text may also be generated based on the advertising demand information. Among them, any sentence or paragraph generation method may be used to merge the advertising demand information into the demand description text corresponding to the advertising demand information. Among them, the demand description text is used to characterize the user's advertising demand in the form of a sentence or paragraph. In other words, the demand description text can be understood as an "advertising slogan". Based on this, step 14 in the aforementioned embodiment "using the text graph model to generate at least one target foreground image matching the advertising demand information" can be implemented based on the following steps 141-142:

[0094] Step 141: Input the requirement description text into the feature extraction layer in the text graph model, and perform feature extraction on the requirement description text to obtain text features.

[0095] Step 142: Input the text features into the image generation layer in the text-generated graph model, and generate at least one target foreground image matching the advertising demand information according to the text features.

[0096] In some optional embodiments, step 15 in the aforementioned embodiment, "running the resource document of the target page to load the target background image and at least one target foreground image into the target page according to the target structured data in different page layer layout modes to obtain multiple candidate advertising images", can be implemented based on the following steps 151-152:

[0097] Step 151: Run the resource document of the target page using a headless browser module that simulates the browser environment. A headless browser is a browser without a graphical user interface that can run in the background and perform various tasks, such as web page crawling, automated testing, etc., and can be used to simulate a real browser environment.

[0098] Step 152: According to the target structured data, a target background image and at least one target foreground image are loaded into the target page in accordance with different page layer layouts to obtain multiple advertisement pages.

[0099] That is, when the resource document of the target page is run, the target structured data in the resource document can be deserialized, so that the target background image and at least one target foreground image can be loaded into the target page according to different page layer layout modes, so as to obtain multiple advertisement pages. Each advertisement page includes: an advertisement area generated by laying out the target background image and at least one target foreground image according to the corresponding page layer layout mode.

[0100] Among them, the target background image and at least one target foreground image can be loaded into the target page according to any page layer layout method to obtain an advertising page, and a partial area (i.e., the advertising area) in the advertising page can display: the layout effect of laying out the target background image and at least one target foreground image according to any page layer layout method.

[0101] Step 153: Take a screenshot of the advertisement area in any advertisement page to obtain any candidate advertisement image.

[0102] In this way, multiple candidate advertising images can be generated more efficiently based on the headless browser module for users to choose from. The above method based on the headless browser module simulating the browser environment is only an exemplary description and does not limit the specific implementation of step 15. For example, instead of using a headless browser, multiple advertising pages can be directly displayed to the user based on the resource document, so that the user can download, save, and manually download the image.

[0103] The above-mentioned advertisement image generation method will be further explained below in combination with actual application scenarios.

[0104] In actual application scenarios, the advertising image generation method can be executed by an advertising image generation platform. The advertising image generation platform can be installed on a terminal device, or installed on a server, or can be distributed on a terminal device and a server, which is not limited in this embodiment. Figure 3The architecture shown in the figure may include image capture service, AI (Artificial Intelligence) image generation service, advertising slogan AI production service of language large model, headless browser interception service, advertising slogan and AI graphic matching service, image advertising slogan fine-tuning service and DSL export service. Among them, AI image generation service and advertising slogan AI production service of language large model belong to AI multimodal service, advertising slogan and AI graphic matching service, image advertising slogan fine-tuning page and DSL export service belong to Shanling platform.

[0105] Among them, the image crawling service is used to: use a web crawler to crawl images from the target website and / or target database. The AI ​​image generation service is used to: use the image generation model to correct the candidate background image to obtain the target background image corresponding to the background layer. The advertising slogan AI production service of the language large model is used to: use the text generation model to generate at least one target foreground image that matches the advertising demand information. The headless browser interception service is used to: run the resource document of the target page to load the target background image and at least one target foreground image to the target page according to the target structured data in different page layer layout modes to obtain multiple candidate advertising images. The advertising slogan and AI graphic matching service is used to: use the advertising template generation model to generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page. The image advertising slogan fine-tuning service is used to: display multiple page layers corresponding to multiple page layer layout modes in the graphics processing platform according to the initial structured data corresponding to the multiple page layer layout modes for layer layout adjustment. The DSL export service is used to: use the advertisement template generation model to parse the description information of the preset domain feature language to obtain the layer layout information of the target page; and add the target structured data corresponding to multiple page layer layout methods to the resource document of the target page.

[0106] like Figure 4 As shown, the example in which the user is a merchant and the executing entity is an advertising image generation platform is used for explanation: the merchant can activate membership rights, thereby starting the advertising image generation service and amortizing it (i.e. consuming the number of advertising image generation times). After that, it can be determined whether the advertising demand information provided by the user is complete. If it is incomplete, the user is prompted to supplement the content. If it is complete, the advertising image generation method is triggered to execute a series of operations such as image AI synthesis to obtain multiple candidate advertising images for the user to select.

[0107] like Figure 5As shown, the user can initiate a request for image generation and fill in the advertising demand information, such as UID (User Identification), city, industry and store data. After that, the advertising image generation platform can match the image library. The image library here is the material library mentioned above. The image is generated based on the background image obtained from the image library, and the template is displayed to the user based on the Shanling platform (i.e., the graphics processing platform in the aforementioned embodiment) for template optimization, thereby generating multiple candidate advertising images based on the optimized template for the user to select images. Among them, the process of using the Shanling platform for template optimization is as follows: Figure 6 Optionally, a crawler service may be used to obtain relevant sample images through the industry to which the user belongs confirmed offline, and fill the image library with the images.

[0108] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 11 to 16 can be device A; for another example, the execution subject of steps 11 to 12 can be device A, and the execution subject of steps 13 to 16 can be device B; and so on.

[0109] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel, and the sequence numbers of the operations, such as 12, 13, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel.

[0110] It should be noted that the descriptions such as “first” and “second” in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit “first” and “second” to different types.

[0111] Figure 7 is a schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present application, and the electronic device is applicable to the advertising picture generation method provided by the aforementioned embodiment, such as Figure 7 As shown, the electronic device may include: a memory 701 , a processor 702 and a communication component 703 .

[0112] The memory 701 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, contact data, phone book data, messages, pictures, videos, etc.

[0113] In some exemplary embodiments, the processor 702 is coupled to the memory 701 and is used to execute the computer program in the memory 701, so as to: respond to the user's advertisement generation instruction and obtain the user's advertisement demand information; the advertisement demand information is the information required for the user to request to provide advertisement generation services for its target object; generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page by using the advertisement template generation model; each page layer layout mode is used to describe a layout mode between a background layer and at least one foreground layer; determine the candidate background image matching the advertisement demand information from the advertisement material library, and modify the candidate background image by using the graph generation model , obtain a target background image corresponding to the background layer; when there is no at least one target foreground image matching the advertising demand information in the advertising material library, use the Wensheng graph model to generate at least one target foreground image matching the advertising demand information; the at least one target foreground image corresponds to the at least one foreground layer; run the resource document of the target page to load the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes, so as to obtain multiple candidate advertising images; in response to the user's selection operation on the multiple candidate advertising images, determine the target advertising image from the multiple candidate advertising images.

[0114] Further optionally, when the processor 702 uses the advertisement template generation model to generate target structured data corresponding to multiple page layer layout methods in the resource document of the target page, it is specifically used to: use the advertisement template generation model to parse the description information of the preset domain feature language to obtain the layer layout information of the target page; and generate the target structured data corresponding to the multiple page layer layout methods in the resource document under the constraints of the layer layout information of the target page.

[0115] Further optionally, when the processor 702 generates target structured data corresponding to the multiple page layer layout modes in the resource document under the constraint of the layer layout information of the target page, it is specifically used to: generate initial structured data corresponding to the multiple page layer layout modes under the constraint of the layer layout information of the target page; display multiple page layers corresponding to each of the multiple page layer layout modes in a graphics processing platform according to the initial structured data corresponding to the multiple page layer layout modes for layer layout adjustment; in response to a page layer adjustment operation, adjust the initial structured data corresponding to the multiple page layer layout modes to obtain target structured data corresponding to the multiple page layer layout modes; and add the target structured data corresponding to the multiple page layer layout modes to the resource document of the target page.

[0116] Further optionally, the processor 702 is also used to: determine the target website and / or target database to be traversed that matches the user category based on the user category information to which the user belongs; use a web crawler to capture pictures in the target website and / or the target database; use an image recognition model to identify target pictures that match the advertising demand information from pictures in the target website and / or the target database, and add the target pictures as background pictures to the advertising material library.

[0117] Further optionally, when the processor 702 uses the graph-generated image model to correct the candidate background image to obtain a target background image corresponding to the background layer, it is specifically used to: use the graph-generated image model to generate feature correction information; perform feature extraction on the candidate background image to obtain a feature vector of the candidate background image; fuse the feature vector with the feature correction information to obtain a fused feature vector; and generate the target background image based on the fused feature vector.

[0118] Further optionally, the graph-generated image model includes: a feature extractor, a prompt word generator, a text encoder, a diffusion generator and an image decoder; when the processor 702 uses the graph-generated image model to correct the candidate background image to obtain a target background image corresponding to the background layer, it is specifically used to: use the feature extractor to extract the feature vector of the candidate background image; use the prompt word generator to generate multiple scene prompt words according to the advertising demand information; use the text encoder to perform vector embedding according to the multiple scene prompt words to obtain semantic vectors corresponding to the multiple scene prompt words; input the semantic vectors corresponding to the multiple scene prompt words into the diffusion generator for denoising processing to obtain the feature vectors corresponding to the multiple scene prompt words; use the image decoder to generate the target background image according to the feature vectors corresponding to the multiple scene prompt words and the feature vectors of the candidate background image.

[0119] Further optionally, the processor 702 is also used to: generate a corresponding demand description text according to the advertising demand information; when the processor 702 uses the Wensheng graph model to generate at least one target foreground image that matches the advertising demand information, it is specifically used to: input the demand description text into the feature extraction layer in the Wensheng graph model, and extract features from the demand description text to obtain text features; input the text features into the image generation layer in the Wensheng graph model, and generate at least one target foreground image that matches the advertising demand information according to the text features.

[0120] Further optionally, the processor 702 runs the resource document of the target page to load the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes to obtain multiple candidate advertising images. Specifically, it is used to: use a headless browser module that simulates a browser environment to run the resource document of the target page, and load the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes to obtain multiple advertising pages, each advertising page including an advertising area generated by laying out the target background image and the at least one target foreground image according to a corresponding page layer layout mode; take a screenshot of the advertising area in any advertising page to obtain any candidate advertising image.

[0121] Further, if Figure 7 As shown, the electronic device also includes: a display 704, a power supply component 705, an audio component 706 and other components. Figure 7 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 7 Components shown.

[0122] The embodiment of the present application also provides a computer-readable storage medium, which, when the computer program is executed by a processor, enables the processor to implement the steps in the advertising image generation method.

[0123] The embodiment of the present application also provides a computer program product, including a computer program / instruction. When the computer program / instruction is executed by a processor, the steps in the advertising image generation method are executed.

[0124] In this embodiment, the user's advertisement generation instruction can be responded to to obtain the user's advertisement demand information; the advertisement demand information is the information required by the user to request to provide advertisement generation service for its target object; the target structured data corresponding to multiple page layer layout modes are generated in the resource document of the target page by using the advertisement template generation model; each page layer layout mode is used to describe a layout mode between the background layer and at least one foreground layer; the candidate background image matching the advertisement demand information is determined from the advertisement material library, and the candidate background image is corrected by using the graph generation model to obtain the target background image corresponding to the background layer; in the case that there is no at least one target foreground image matching the advertisement demand information in the advertisement material library, at least one target foreground image matching the advertisement demand information is generated by using the graph generation model; at least one target foreground image corresponds to at least one foreground layer; the resource document of the target page is run to load the target background image and at least one target foreground image to the target page according to the target structured data according to different page layer layout modes to obtain multiple candidate advertisement images; in response to the user's selection operation on multiple candidate advertisement images, the target advertisement image is determined from the multiple candidate advertisement images. In this way, advertisement images that better meet the user's advertisement needs can be generated more efficiently.

[0125] The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0126] The above-mentioned communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.

[0127] The above-mentioned display includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0128] The power supply assembly provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply assembly is located.

[0129] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (Microphone, MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.

[0130] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-readable storage media (including but not limited to disk storage, compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM), optical storage, etc.) containing computer-usable program code.

[0131] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0132] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0134] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input / output interface, network interface and memory.

[0135] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0136] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0137] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0138] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for generating an advertisement picture, characterized in that: include: Responding to an advertisement generation instruction from a user, obtaining advertisement demand information from the user; The advertisement demand information is the information required by the user to request to provide advertisement generation service for its target object; Generate a target structured data corresponding to a plurality of page layer layout modes in a resource document of a target page by using an advertisement template generation model; each page layer layout mode is used to describe a layout mode between a background layer and at least one foreground layer; Determine a candidate background image that matches the advertisement demand information from an advertisement material library, and modify the candidate background image using a graph-to-graph model to obtain a target background image corresponding to the background layer; In the case that at least one target foreground image matching the advertisement demand information does not exist in the advertisement material library, generating at least one target foreground image matching the advertisement demand information by using a text graph model; The at least one target foreground picture corresponds to the at least one foreground layer; Running the resource document of the target page to load the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes, so as to obtain a plurality of candidate advertisement images; In response to the user's selection operation on the multiple candidate advertising images, a target advertising image is determined from the multiple candidate advertising images.

2. The method according to claim 1, characterized in that The advertisement template generation model is used to generate target structured data corresponding to multiple page layer layout modes in the resource document of the target page, including: The advertisement template generation model is used to parse the description information of the preset domain feature language to obtain the layer layout information of the target page; Under the constraint of the layer layout information of the target page, target structured data corresponding to the multiple page layer layout modes are generated in the resource document.

3. The method according to claim 2, characterized in that Under the constraint of the layer layout information of the target page, generating target structured data corresponding to the multiple page layer layout modes in the resource document includes: Under the constraint of the layer layout information of the target page, generating initial structured data corresponding to the layer layout modes of the multiple pages; According to the initial structured data corresponding to the plurality of page layer layout modes, displaying the plurality of page layers corresponding to the plurality of page layer layout modes in the graphics processing platform for layer layout adjustment; In response to the page layer adjustment operation, the initial structured data corresponding to the plurality of page layer layout modes are adjusted to obtain target structured data corresponding to the plurality of page layer layout modes; The target structured data corresponding to the multiple page layer layout modes are added to the resource document of the target page.

4. The method according to claim 1, characterized in that: Also includes: According to the user category information to which the user belongs, determining the target website and / or target database to be traversed that matches the user category; Using a web crawler to crawl the images in the target website and / or the target database; By using the picture recognition model, a target picture matching the advertisement demand information is identified from pictures in the target website and / or the target database, and the target picture is added to the advertisement material library as a background picture.

5. The method according to claim 1, characterized in that: The candidate background image is modified by using the graph-generated graph model to obtain a target background image corresponding to the background layer, including: Using the graph-to-graph model, generating feature correction information; Performing feature extraction on the candidate background image to obtain a feature vector of the candidate background image; Fusing the feature vector with the feature correction information to obtain a fused feature vector; The target background image is generated according to the fused feature vector.

6. The method according to claim 1, characterized in that The graph-generated graph model includes: a feature extractor, a prompt word generator, a text encoder, a diffusion generator and an image decoder; The candidate background image is modified by using the graph-generated graph model to obtain a target background image corresponding to the background layer, including: Using a feature extractor, extracting a feature vector of the candidate background image; Using the prompt word generator to generate a plurality of scene prompt words according to the advertising demand information; Using the text encoder, performing vector embedding according to the multiple scene prompt words to obtain semantic vectors corresponding to the multiple scene prompt words; Inputting the semantic vectors corresponding to the multiple scene prompt words into the diffusion generator for denoising to obtain feature vectors corresponding to the multiple scene prompt words; The target background picture is generated by using the image decoder according to the feature vectors corresponding to the multiple scene prompt words and the feature vectors of the candidate background pictures.

7. The method according to claim 1, characterized in that Also includes: Generate a corresponding demand description text according to the advertising demand information; Generating at least one target foreground image matching the advertisement demand information by using the text graph model includes: Inputting the requirement description text into the feature extraction layer in the text graph model, and performing feature extraction on the requirement description text to obtain text features; The text features are input into the image generation layer in the text graph model, and at least one target foreground image matching the advertising demand information is generated according to the text features.

8. The method according to any one of claims 1 to 7, characterized in that: Running the resource document of the target page to load the target background image and the at least one target foreground image into the target page according to the target structured data in different page layer layout modes to obtain a plurality of candidate advertisement images, including: Using a headless browser module that simulates a browser environment to run the resource document of the target page, and according to the target structured data, loading the target background image and the at least one target foreground image into the target page in accordance with different page layer layout modes, so as to obtain a plurality of advertisement pages, each advertisement page including an advertisement area generated by laying out the target background image and the at least one target foreground image in accordance with a corresponding page layer layout mode; Take a screenshot of the advertisement area in any advertisement page to obtain any candidate advertisement image.

9. An electronic device, characterized in that: include: A memory and a processor; wherein the memory is used to: store one or more computer instructions; The processor is used to execute the one or more computer instructions to: perform the steps in the method according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: When the computer program is executed by a processor, the processor is enabled to implement the steps in the method according to any one of claims 1 to 8.

11. A computer program product, characterized in that The method comprises a computer program / instruction, which, when executed by a processor, enables the processor to implement the steps of the method according to any one of claims 1 to 8.

Images