A method, device, storage medium and equipment for generating H5 scenes

Abstract

The present application discloses a method, device, storage medium and equipment for generating H5 scenes, obtaining demand information input by users, inputting marketing pictures into a picture entity recognition model, and obtaining recognition results output by the picture entity recognition model. The marketing theme information is semantically understood to obtain copy features, and the copy features are input into an industry prediction model to obtain prediction results output by the industry prediction model. From the various pictures shown in the picture library, a picture containing an entity is selected as the target picture. From the various components shown in the component library, a component with the same industry type as the industry type of the copy features is selected as the target component. Marketing pictures, marketing theme information, target pictures and target components are collected to construct material features. The material features are input into a scene generation model to obtain multiple H5 scenes output by the scene generation model. It can be seen that the method shown in this application can effectively reduce the difficulty of making H5 scenes.

Description

Technical Field

[0001] The present application relates to the field of advertising design, and in particular to a method, device, storage medium and equipment for generating an H5 scene. Background Art

[0002] H5 scene refers to an interactive web application based on HTML5 technology. In addition to pictures and texts, H5 scene can also add factors such as sound, motion effects, video and interactive elements, so as to make the interactive visual effect of the advertisement better and the interactive sensibility stronger.

[0003] At present, the existing production process of H5 scenes is all manual operation. Specifically, users use editors to create templates and modify the content in the templates according to their own required materials to obtain H5 scenes. However, the editor operation is complicated, including nearly 1,000 adjustable contents such as templates, pages, layers, combinations, components, etc. The high cost of getting started has greatly hindered the application and development of H5 scenes. Summary of the invention

[0004] The present application provides a method, apparatus, storage medium and device for generating an H5 scene, aiming to reduce the difficulty of producing the H5 scene.

[0005] In order to achieve the above objectives, this application provides the following technical solutions:

[0006] A method for generating an H5 scene, comprising:

[0007] Obtaining demand information input by a user; the demand information includes marketing pictures and marketing theme information required for the advertising business;

[0008] Input the marketing picture into a picture entity recognition model to obtain a recognition result output by the picture entity recognition model; the picture entity recognition model is pre-trained based on taking a sample picture as input and taking entities manually identified in the sample picture as training targets; the recognition result includes entities contained in the marketing picture;

[0009] The marketing theme information is semantically understood to obtain text features, and the text features are input into an industry prediction model to obtain a prediction result output by the industry prediction model; the industry prediction model is pre-trained based on taking sample text features as input and taking the industry type of the sample text features manually identified as a training target; the prediction result includes the industry type of the text features;

[0010] Selecting, from various pictures shown in a preset picture library, the picture containing the entity contained in the marketing picture as the target picture;

[0011] From various components shown in the preset component library, select the component whose industry type is the same as the industry type of the copy feature as the target component;

[0012] Gather the marketing picture, the marketing theme information, the target picture and the target component to construct a material feature;

[0013] The material features are input into a scene generation model to obtain multiple H5 scenes output by the scene generation model; the scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

[0014] Optionally, selecting the picture containing the entity contained in the marketing picture as the target picture from the pictures shown in the preset picture library includes:

[0015] Performing color recognition on the marketing image to obtain the main color and matching color of the marketing image;

[0016] The main color, the matching color and the marketing picture are input into a color matching model to obtain a first set of pictures to be selected output by the color matching model; wherein the color matching model is pre-trained based on taking sample colors and sample pictures as input, manually scoring pictures in a preset picture library, and selecting pictures with scores greater than a preset score as training targets; the first set of pictures to be selected includes a plurality of first pictures to be selected, and the scores of the first pictures to be selected are greater than the preset score;

[0017] From each of the first candidate pictures, the first candidate picture that includes the entity included in the marketing picture is selected as the target picture.

[0018] Optionally, selecting, from each of the first candidate images, the first candidate image that includes the entity included in the marketing image as the target image includes:

[0019] Measuring the size of the marketing picture to obtain the size of the marketing picture;

[0020] From each of the first pictures to be selected, select the first pictures to be selected that have the same size as the marketing picture as the second picture to be selected;

[0021] From each of the second candidate pictures, the second candidate picture that includes the entity included in the marketing picture is selected as the target picture.

[0022] Optionally, the component library includes a music component library, a text component library and a motion effect component library;

[0023] The step of selecting, from among the components shown in the preset component library, the component having the same industry type as the industry type of the copy feature as the target component includes:

[0024] From the various music components shown in the music component library, select the music component whose industry type is the same as the industry type of the copy feature as the target music component;

[0025] Selecting, from the various copywriting components shown in the copywriting component library, the copywriting component having the same industry type as the industry type of the copywriting feature as the target copywriting component;

[0026] From the various motion effect components shown in the motion effect component library, select the motion effect component whose industry type is the same as the industry type of the copy feature as the target motion effect component;

[0027] The target music component, the target text component and the target motion effect component are identified as target components.

[0028] Optionally, also include:

[0029] Perform feature extraction on each of the H5 scenes respectively to obtain data features of each of the H5 scenes;

[0030] For each H5 scene, based on the data features of the H5 scene, a multi-dimensional score of the H5 scene is obtained, and the arithmetic mean of the multi-dimensional scores is calculated to obtain the hit index of the H5 scene;

[0031] From each of the H5 scenes, select the H5 scene whose value of the hit index is greater than a preset threshold as the target H5 scene;

[0032] The target H5 scene is presented to the user through a preset interface.

[0033] Optionally, the data features include text features, picture features, music features, motion effect features, historical features and hot spot features;

[0034] The method of obtaining a multi-dimensional score of the H5 scene based on the data features of the H5 scene, and calculating the arithmetic mean of the multi-dimensional scores to obtain the hit index of the H5 scene includes:

[0035] The text features, the picture features and the music features of the H5 scene are input into a theme fit evaluation model to obtain a theme fit score output by the theme fit evaluation model; the theme fit evaluation model is pre-trained based on taking the text features, the picture features and the music features of the sample H5 scene as input, and manually scoring the sample H5 scene, and obtaining the theme fit score as a training target;

[0036] The motion effect features of the H5 scene are input into a visual motion effect evaluation model to obtain a visual motion effect score output by the visual motion effect evaluation model; the visual motion effect evaluation model is pre-trained based on taking the motion effect features of the sample H5 scene as input and manually scoring the sample H5 scene, and the obtained visual motion effect score is used as a training target;

[0037] The text features, the picture features, the music features and the motion effect features of the H5 scene are input into a content innovation evaluation model to obtain a content innovation score output by the content innovation evaluation model; the content innovation evaluation model is based on taking the text features, the picture features, the music features and the motion effect features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained content innovation score is used as a training target, and is pre-trained;

[0038] The hot spot features of the H5 scene are input into the hot spot follow-up evaluation model to obtain the hot spot follow-up score output by the hot spot follow-up evaluation model; the hot spot follow-up evaluation model is based on taking the hot spot features of the Haloxylon ammodendron sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained hot spot follow-up score is used as a training target, and is pre-trained;

[0039] The historical features and the hotspot features of the H5 scene are input into a communication efficiency evaluation model to obtain a communication efficiency score output by the communication efficiency evaluation model; the communication efficiency evaluation model is based on the historical features and the hotspot features of the sample H5 scene as input, and artificially scores the sample H5 scene, and the obtained communication efficiency score is used as a training target, and is pre-trained;

[0040] The historical features of the H5 scene are input into the historical analogy evaluation model to obtain the historical analogy score output by the historical analogy evaluation model; the historical analogy evaluation model is pre-trained based on the historical features of the sample H5 scene as input and the sample H5 scene is manually scored, and the obtained historical analogy score is used as a training target;

[0041] The arithmetic mean of the theme fit score, the visual animation score, the content innovation score, the hot spot follow-up score, the communication efficiency score and the historical analogy score is calculated to obtain the hit index of the H5 scene.

[0042] A device for generating an H5 scene, comprising:

[0043] An information acquisition unit, used to acquire demand information input by a user; the demand information includes marketing pictures and marketing theme information required for the advertising business;

[0044] An image recognition unit is used to input the marketing image into an image entity recognition model to obtain a recognition result output by the image entity recognition model; the image entity recognition model is pre-trained based on taking a sample image as input and taking entities artificially identified in the sample image as training targets; the recognition result includes entities contained in the marketing image;

[0045] An industry prediction unit is used to perform semantic understanding on the marketing theme information, obtain text features, and input the text features into an industry prediction model to obtain a prediction result output by the industry prediction model; the industry prediction model is pre-trained based on taking sample text features as input and taking the industry type of the sample text features manually identified as a training target; the prediction result includes the industry type of the text features;

[0046] A picture selection unit, configured to select, from various pictures shown in a preset picture library, the picture containing the entity contained in the marketing picture as a target picture;

[0047] A component selection unit, used to select, from various components shown in a preset component library, the component whose industry type is the same as the industry type of the copy feature as a target component;

[0048] A feature construction unit, used to collect the marketing picture, the marketing theme information, the target picture and the target component to construct a material feature;

[0049] A scene generation unit is used to input the material features into a scene generation model to obtain multiple H5 scenes output by the scene generation model; the scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

[0050] Optionally, also include:

[0051] Scene screening unit;

[0052] The scene screening unit is used for:

[0053] Perform feature extraction on each of the H5 scenes respectively to obtain data features of each of the H5 scenes;

[0054] For each H5 scene, based on the data features of the H5 scene, a multi-dimensional score of the H5 scene is obtained, and the arithmetic mean of the multiple scores is calculated to obtain the hit index of the H5 scene;

[0055] From each of the H5 scenes, select the H5 scene whose value of the hit index is greater than a preset threshold as the target H5 scene;

[0056] The target H5 scene is presented to the user through a preset interface.

[0057] A computer-readable storage medium, the computer-readable storage medium comprising a stored program, wherein the program executes the method for generating an H5 scene.

[0058] A device for generating an H5 scene, comprising: a processor, a memory and a bus; the processor and the memory are connected via the bus;

[0059] The memory is used to store the program, and the processor is used to run the program, wherein the method for generating the H5 scene is executed when the program is run.

[0060] The technical solution provided by the present application obtains the demand information input by the user, inputs the marketing picture into the picture entity recognition model, and obtains the recognition result output by the picture entity recognition model. The marketing theme information is semantically understood to obtain the copywriting features, and the copywriting features are input into the industry prediction model to obtain the prediction results output by the industry prediction model. From the various pictures shown in the preset picture library, the picture containing the entity contained in the marketing picture is selected as the target picture. From the various components shown in the preset component library, the component with the same industry type as the industry type of the copywriting feature is selected as the target component. The marketing picture, marketing theme information, target picture and target component are collected to construct the material features. The material features are input into the scene generation model to obtain multiple H5 scenes output by the scene generation model. Compared with the prior art, the method shown in the present application does not need to use an editor to create a template, and does not require the user to modify the content in the template. The entire H5 scene production process does not require user participation. The user only needs to enter the marketing pictures and marketing theme information required for the advertising business to obtain the H5 scene. It can be seen that by using the method shown in this embodiment, the difficulty of producing H5 scenes can be effectively reduced, and the efficiency of producing H5 scenes can also be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] 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 or the description of the prior art 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.

[0062] Figure 1a A schematic diagram of a method for generating an H5 scene provided in an embodiment of the present application;

[0063] Figure 1b A schematic diagram of a method for generating an H5 scene provided in an embodiment of the present application;

[0064] Figure 1c A schematic diagram of a method for generating an H5 scene provided in an embodiment of the present application;

[0065] Figure 2 A schematic diagram of another method for generating an H5 scene provided in an embodiment of the present application;

[0066] Figure 3 A schematic diagram of the architecture of an apparatus for generating an H5 scene provided in an embodiment of the present application. DETAILED DESCRIPTION

[0067] 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. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0068] like Figure 1a , Figure 1b and Figure 1c As shown, it is a schematic diagram of a method for generating an H5 scene provided in an embodiment of the present application, comprising the following steps:

[0069] S101: Obtaining demand information input by the user.

[0070] Among them, the demand information includes marketing pictures and marketing theme information required for the advertising business. Specifically, taking the advertising business of store opening as an example, the demand information includes opening pictures (ie marketing pictures), store name, opening time and store address (ie marketing theme information).

[0071] S102: Perform color recognition on the marketing image to obtain the main color and matching color of the marketing image.

[0072] Among them, the open source image color extraction model can be used to perform color recognition on marketing images. Specifically, the marketing image is input into the image color extraction model to obtain the main color and color ratio of the marketing image output by the image color extraction model.

[0073] S103: Measure the size of the marketing image to obtain the size of the marketing image.

[0074] S104: Input the marketing picture into the picture entity recognition model to obtain the recognition result output by the picture entity recognition model.

[0075] The image entity recognition model is pre-trained based on taking sample images as input and artificially identified entities in the sample images as training targets. In the embodiment of the present application, the recognition result includes the entities contained in the marketing image. Taking the advertising business of store opening as an example, the entity contained in the marketing image is the store.

[0076] S105: Perform semantic understanding on the marketing theme information to obtain text features, and input the text features into the industry prediction model to obtain the prediction results output by the industry prediction model.

[0077] The industry prediction model is pre-trained based on taking sample copy features as input and taking the industry type of the sample copy features manually identified as training targets. In the embodiment of the present application, the prediction result includes the industry type of the copy features.

[0078] It should be noted that the specific implementation process of semantic understanding is common knowledge familiar to those skilled in the art and will not be repeated here. Taking the opening time recorded in the marketing theme information as an example, the semantic understanding of the opening time can obtain the holidays and hot marketing events in the near future. The holidays and hot marketing events in the near future are all copywriting features obtained by semantic understanding of the marketing theme.

[0079] Specifically, taking the text feature as a store address as an example, the store address is input into the industry prediction model to obtain a prediction result output by the industry prediction model, and the prediction result indicates that the industry type of the store address is a map.

[0080] S106: Input the main color, the matching color and the marketing picture into the color matching model to obtain a first set of pictures to be selected output by the color matching model.

[0081] The color matching model is pre-trained based on taking sample colors and sample images as input, manually scoring images in a preset image library, and selecting images with scores greater than a preset score as training targets. The first set of images to be selected includes multiple first images to be selected, and the scores of the first images to be selected are greater than the preset scores.

[0082] S107: From each first candidate picture, select a first candidate picture whose size is the same as that of the marketing picture as the second candidate picture.

[0083] S108: Selecting, from each second candidate picture, a second candidate picture that includes the entity included in the marketing picture as the target picture.

[0084] S109: Selecting a music component having the same industry type as the industry type of the text feature from various music components shown in the preset music component library as a target music component.

[0085] S110: Selecting a copy component whose industry type is the same as the industry type of the copy feature from various copy components shown in a preset copy component library as a target copy component.

[0086] S111: From various motion effect components shown in a preset motion effect component library, select a motion effect component whose industry type is the same as the industry type of the copy feature as a target motion effect component.

[0087] S112: Collect user feature information, marketing pictures, marketing themes, target pictures, target music components, target text components, and target motion effect components to construct material features.

[0088] S113: Inputting the material features into the scene generation model to obtain multiple H5 scenes output by the scene generation model.

[0089] Among them, the scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

[0090] S114: Extract features of each H5 scene respectively to obtain data features of each H5 scene.

[0091] Among them, data features include but are not limited to: text features, picture features, music features, motion effect features, historical features, and hot spot features.

[0092] S115: For each H5 scene, the text features, image features, and music features of the H5 scene are input into the theme fit evaluation model to obtain an evaluation result output by the theme fit evaluation model.

[0093] The theme fit evaluation model is based on taking the text features, image features and music features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained theme fit score is used as the training target, and is pre-trained. In the embodiment of the present application, the evaluation result output by the theme fit evaluation model includes the theme fit score of the H5 scene.

[0094] S116: Input the dynamic effect features of the H5 scene into the visual dynamic effect evaluation model to obtain the evaluation result output by the visual dynamic effect evaluation model.

[0095] The visual motion effect evaluation model is based on taking the motion effect features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained visual motion effect score is used as the training target, and is pre-trained. In the embodiment of the present application, the evaluation result output by the visual motion effect evaluation model includes the visual motion effect score of the H5 scene.

[0096] S117: Input the text features, picture features, music features, and motion effect features of the H5 scene into the content innovation evaluation model to obtain the evaluation results output by the content innovation evaluation model.

[0097] The content innovation evaluation model is based on taking the text features, picture features, music features and motion effect features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained content innovation score is used as the training target, and is pre-trained. In the embodiment of the present application, the evaluation results output by the content innovation evaluation model include the content innovation score of the H5 scene.

[0098] S118: Input the hot spot features of the H5 scene into the hot spot follow-up evaluation model to obtain the evaluation result output by the hot spot follow-up evaluation model.

[0099] The hotspot follow-up evaluation model is pre-trained based on taking the hotspot features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained hotspot follow-up score is used as the training target. In the embodiment of the present application, the evaluation result output by the hotspot follow-up evaluation model includes the hotspot follow-up score of the H5 scene.

[0100] S119: Input the historical features and hot spot features of the H5 scene into the communication efficiency evaluation model to obtain the evaluation results output by the communication efficiency evaluation model.

[0101] The communication efficiency evaluation model is based on the historical characteristics and hot spot characteristics of the sample H5 scene as input, and the sample H5 scene is scored manually, and the obtained communication efficiency score is used as the training target, and is pre-trained. In the embodiment of the present application, the evaluation result output by the communication efficiency evaluation model includes the communication efficiency score of the H5 scene.

[0102] S120: Input the historical features of the H5 scene into the historical analogy evaluation model to obtain the evaluation results output by the historical analogy evaluation model.

[0103] The historical analogy evaluation model is based on taking the historical features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained historical analogy score is used as the training target, and is pre-trained. In the embodiment of the present application, the evaluation result output by the historical analogy evaluation model includes the historical analogy score of the H5 scene.

[0104] S121: Calculate the arithmetic mean of the theme fit score, visual animation score, content innovation score, hot spot follow-up score, communication efficiency score and historical analogy score to obtain the explosive index of the H5 scene.

[0105] S122: From each H5 scene, select an H5 scene whose hit index value is greater than a preset threshold as the target H5 scene.

[0106] S123: Displaying the target H5 scene to the user through a preset interface.

[0107] In summary, compared with the prior art, the method shown in this embodiment does not require the use of an editor to create a template, nor does it require the user to modify the content in the template. The entire H5 scene production process does not require user participation. The user only needs to enter the marketing pictures and marketing theme information required for the advertising business to obtain the H5 scene. It can be seen that the method shown in this embodiment can effectively reduce the difficulty of H5 scene production and improve the efficiency of H5 scene production.

[0108] It should be noted that S122 mentioned in the above embodiment is an optional implementation of the method for generating H5 scenes described in this application. In addition, S123 mentioned in the above embodiment is also an optional implementation of the method for generating H5 scenes described in this application. To this end, the process mentioned in the above embodiment can be summarized as follows: Figure 2 The method shown.

[0109] S201: Obtaining demand information input by the user.

[0110] Among them, demand information includes marketing pictures and marketing theme information required for advertising business.

[0111] S202: Input the marketing picture into the picture entity recognition model to obtain the recognition result output by the picture entity recognition model.

[0112] Among them, the image entity recognition model is pre-trained based on taking sample images as input and the entities manually identified in the sample images as training targets; the recognition results include the entities contained in the marketing images.

[0113] S203: Perform semantic understanding on the marketing theme information to obtain text features, and input the text features into the industry prediction model to obtain the prediction results output by the industry prediction model.

[0114] Among them, the industry prediction model is pre-trained based on taking sample copy features as input and the industry type of the manually identified sample copy features as the training target; the prediction result includes the industry type of the copy features.

[0115] S204: Selecting a picture containing an entity contained in the marketing picture from pictures shown in a preset picture library as a target picture.

[0116] S205: Selecting a component whose industry type is the same as the industry type of the copy feature from among the components shown in the preset component library as a target component.

[0117] S206: Gather marketing images, marketing theme information, target images and target components to construct material features.

[0118] S207: Inputting the material features into the scene generation model to obtain multiple H5 scenes output by the scene generation model.

[0119] Among them, the scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

[0120] In summary, compared with the prior art, the method shown in this embodiment does not require the use of an editor to create a template, nor does it require the user to modify the content in the template. The entire H5 scene production process does not require user participation. The user only needs to enter the marketing pictures and marketing theme information required for the advertising business to obtain the H5 scene. It can be seen that the method shown in this embodiment can effectively reduce the difficulty of H5 scene production and improve the efficiency of H5 scene production.

[0121] Corresponding to the method for generating an H5 scene provided in the above embodiment, the present application also provides a device for generating an H5 scene.

[0122] like Figure 3 As shown, it is a schematic diagram of the architecture of a device for generating an H5 scene provided in an embodiment of the present application, including:

[0123] The information acquisition unit 100 is used to acquire demand information input by the user; the demand information includes marketing pictures and marketing theme information required for the advertising business.

[0124] The image recognition unit 200 is used to input the marketing image into the image entity recognition model to obtain the recognition result output by the image entity recognition model; the image entity recognition model is pre-trained based on taking the sample image as input and the entities manually identified in the sample image as training targets; the recognition result includes the entities contained in the marketing image.

[0125] The industry prediction unit 300 is used to perform semantic understanding on the marketing theme information, obtain the copy features, and input the copy features into the industry prediction model to obtain the prediction results output by the industry prediction model; the industry prediction model is based on the sample copy features as input and the industry type of the manually identified sample copy features as the training target, and is pre-trained; the prediction results include the industry type of the copy features.

[0126] The picture selection unit 400 is used to select a picture containing an entity contained in the marketing picture from various pictures shown in a preset picture library as a target picture.

[0127] Among them, the picture selection unit 400 is specifically used for: performing color recognition on the marketing picture to obtain the main color and matching color of the marketing picture; inputting the main color, matching color and marketing picture into the color matching model to obtain a first set of pictures to be selected output by the color matching model; wherein the color matching model is based on taking sample colors and sample pictures as input, manually scoring pictures in a preset picture library, and selecting pictures with scores greater than preset scores as training targets, and pre-training; the first set of pictures to be selected includes multiple first pictures to be selected, and the scores of the first pictures to be selected are greater than the preset scores; from each first picture to be selected, a first picture to be selected that contains an entity contained in the marketing picture is selected as the target picture.

[0128] The picture selection unit 400 is used to select a first candidate picture containing an entity contained in a marketing picture from each first candidate picture shown in a picture library as a target picture, including: measuring the size of the marketing picture to obtain the size of the marketing picture; selecting a first candidate picture having the same size as the marketing picture from each first candidate picture as a second candidate picture; and selecting a second candidate picture containing the entity contained in the marketing picture as a target picture from each second candidate picture.

[0129] The component selection unit 500 is used to select a component with the same industry type as the industry type of the copy feature from various components shown in the preset component library as a target component.

[0130] Among them, the component library includes a music component library, a copy component library and a motion effect component library, and the component selection unit 500 is specifically used to: select a music component whose industry type is the same as the industry type of the copy feature from each music component shown in the music component library as a target music component; select a copy component whose industry type is the same as the industry type of the copy feature from each copy component shown in the copy component library as a target copy component; select a motion effect component whose industry type is the same as the industry type of the copy feature from each motion effect component shown in the motion effect component library as a target motion effect component; and identify the target music component, target copy component and target motion effect component as target components.

[0131] The feature construction unit 600 is used to collect marketing pictures, marketing theme information, target pictures and target components to construct material features.

[0132] The scene generation unit 700 is used to input material features into the scene generation model to obtain multiple H5 scenes output by the scene generation model; the scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

[0133] The scene screening unit 800 is used to: extract features from each H5 scene respectively to obtain data features of each H5 scene; for each H5 scene, obtain a multi-dimensional score of the H5 scene based on the data features of the H5 scene, and calculate the arithmetic mean between the multi-dimensional scores to obtain the hit index of the H5 scene; from each H5 scene, select an H5 scene whose hit index value is greater than a preset threshold as the target H5 scene; and display the target H5 scene to the user through a preset interface.

[0134] The data features include text features, picture features, music features, dynamic features, historical features and hot features; the scene screening unit 800 is used to obtain the multi-dimensional score of the H5 scene based on the data features of the H5 scene, and calculate the arithmetic mean between the multi-dimensional scores to obtain the hot index of the H5 scene, including the following steps:

[0135] The text features, picture features and music features of the H5 scene are input into the theme fit evaluation model to obtain the theme fit score output by the theme fit evaluation model; the theme fit evaluation model is based on taking the text features, picture features and music features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained theme fit score is used as the training target, and is pre-trained;

[0136] The motion effect features of the H5 scene are input into the visual motion effect evaluation model to obtain the visual motion effect score output by the visual motion effect evaluation model; the visual motion effect evaluation model is based on taking the motion effect features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained visual motion effect score is used as the training target, and is pre-trained;

[0137] The text features, picture features, music features and motion effect features of the H5 scene are input into the content innovation evaluation model to obtain the content innovation score output by the content innovation evaluation model; the content innovation evaluation model is based on the text features, picture features, music features and motion effect features of the sample H5 scene as input, and the sample H5 scene is scored manually, and the obtained content innovation score is used as the training target, and is pre-trained;

[0138] The hotspot features of the H5 scene are input into the hotspot follow-up evaluation model to obtain the hotspot follow-up score output by the hotspot follow-up evaluation model; the hotspot follow-up evaluation model is based on taking the hotspot features of the Haloxylon ammodendron sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained hotspot follow-up score is used as the training target, and is pre-trained;

[0139] The historical characteristics and hot spot characteristics of the H5 scene are input into the communication efficiency evaluation model to obtain the communication efficiency score output by the communication efficiency evaluation model; the communication efficiency evaluation model is based on the historical characteristics and hot spot characteristics of the sample H5 scene as input, and the sample H5 scene is manually scored, and the obtained communication efficiency score is used as the training target, and is pre-trained;

[0140] The historical features of the H5 scene are input into the historical analogy evaluation model to obtain the historical analogy score output by the historical analogy evaluation model; the historical analogy evaluation model is based on the historical features of the sample H5 scene as input, and the sample H5 scene is manually scored, and the obtained historical analogy score is used as the training target, and is pre-trained;

[0141] Calculate the arithmetic mean of the theme fit score, visual animation score, content innovation score, hot spot follow-up score, communication efficiency score and historical analogy score to obtain the explosive index of the H5 scene.

[0142] In summary, compared with the prior art, the method shown in this embodiment does not require the use of an editor to create a template, nor does it require the user to modify the content in the template. The entire H5 scene production process does not require user participation. The user only needs to enter the marketing pictures and marketing theme information required for the advertising business to obtain the H5 scene. It can be seen that the method shown in this embodiment can effectively reduce the difficulty of H5 scene production and improve the efficiency of H5 scene production.

[0143] The present application also provides a computer-readable storage medium, which includes a stored program, wherein the program executes the method for generating an H5 scene provided by the present application.

[0144] The present application also provides a device for generating an H5 scene, including: a processor, a memory, and a bus. The processor and the memory are connected via a bus, the memory is used to store a program, and the processor is used to run the program, wherein the method for generating an H5 scene provided by the present application is executed when the program is run.

[0145] If the functions described in the method of the embodiment of the present application are implemented in the form of software functional units and sold or used as independent products, they can be stored in a storage medium readable by a computing device. Based on this understanding, the part of the embodiment of the present application that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to enable a computing device (which can be a personal computer, server, mobile computing device or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0146] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0147] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating an H5 scene, characterized in that: include: Obtain the demand information input by the user; The demand information includes marketing pictures and marketing theme information required for the advertising business; Input the marketing picture into a picture entity recognition model to obtain a recognition result output by the picture entity recognition model; the picture entity recognition model is pre-trained based on taking a sample picture as input and taking entities manually identified in the sample picture as training targets; the recognition result includes entities contained in the marketing picture; The marketing theme information is semantically understood to obtain text features, and the text features are input into an industry prediction model to obtain a prediction result output by the industry prediction model; the industry prediction model is pre-trained based on taking sample text features as input and taking the industry type of the sample text features manually identified as a training target; the prediction result includes the industry type of the text features; Selecting, from various pictures shown in a preset picture library, the picture containing the entity contained in the marketing picture as the target picture; From various components shown in the preset component library, select the component whose industry type is the same as the industry type of the copy feature as the target component; Gather the marketing picture, the marketing theme information, the target picture and the target component to construct a material feature; Inputting the material features into a scene generation model to obtain multiple H5 scenes output by the scene generation model; The scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

2. The method according to claim 1, characterized in that The selecting, from the pictures shown in the preset picture library, the picture containing the entity contained in the marketing picture as the target picture includes: Performing color recognition on the marketing image to obtain the main color and matching color of the marketing image; The main color, the matching color and the marketing picture are input into a color matching model to obtain a first set of pictures to be selected output by the color matching model; wherein the color matching model is pre-trained based on taking sample colors and sample pictures as input, manually scoring pictures in a preset picture library, and selecting pictures with scores greater than a preset score as training targets; the first set of pictures to be selected includes a plurality of first pictures to be selected, and the scores of the first pictures to be selected are greater than the preset score; From each of the first candidate pictures, the first candidate picture that includes the entity included in the marketing picture is selected as the target picture.

3. The method according to claim 2, characterized in that The step of selecting, from each of the first candidate images, the first candidate image that includes the entity included in the marketing image as the target image includes: Measuring the size of the marketing picture to obtain the size of the marketing picture; From each of the first candidate pictures, select the first candidate picture having the same size as the marketing picture as the second candidate picture; From each of the second candidate pictures, the second candidate picture that includes the entity included in the marketing picture is selected as the target picture.

4. The method according to claim 1, characterized in that The component library includes a music component library, a text component library and a dynamic effect component library; The step of selecting, from among the components shown in the preset component library, the component having the same industry type as the industry type of the copy feature as the target component includes: From the various music components shown in the music component library, select the music component whose industry type is the same as the industry type of the copy feature as the target music component; Selecting, from the various copywriting components shown in the copywriting component library, the copywriting component having the same industry type as the industry type of the copywriting feature as the target copywriting component; From the various motion effect components shown in the motion effect component library, select the motion effect component whose industry type is the same as the industry type of the copy feature as the target motion effect component; The target music component, the target text component and the target motion effect component are identified as target components.

5. The method according to claim 1, characterized in that: Also includes: Perform feature extraction on each of the H5 scenes respectively to obtain data features of each of the H5 scenes; For each H5 scene, based on the data features of the H5 scene, a multi-dimensional score of the H5 scene is obtained, and the arithmetic mean of the multi-dimensional scores is calculated to obtain the hit index of the H5 scene; From each of the H5 scenes, select the H5 scene whose value of the hit index is greater than a preset threshold as the target H5 scene; The target H5 scene is presented to the user through a preset interface.

6. The method according to claim 5, characterized in that The data features include text features, picture features, music features, motion effect features, historical features and hot spot features; The method of obtaining a multi-dimensional score of the H5 scene based on the data features of the H5 scene, and calculating the arithmetic mean of the multi-dimensional scores to obtain the hit index of the H5 scene includes: The text features, the picture features and the music features of the H5 scene are input into a theme fit evaluation model to obtain a theme fit score output by the theme fit evaluation model; the theme fit evaluation model is pre-trained based on taking the text features, the picture features and the music features of the sample H5 scene as input, and manually scoring the sample H5 scene, and obtaining the theme fit score as a training target; The motion effect features of the H5 scene are input into a visual motion effect evaluation model to obtain a visual motion effect score output by the visual motion effect evaluation model; the visual motion effect evaluation model is pre-trained based on taking the motion effect features of the sample H5 scene as input and manually scoring the sample H5 scene, and the obtained visual motion effect score is used as a training target; The text features, the picture features, the music features and the motion effect features of the H5 scene are input into a content innovation evaluation model to obtain a content innovation score output by the content innovation evaluation model; the content innovation evaluation model is based on taking the text features, the picture features, the music features and the motion effect features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained content innovation score is used as a training target, and is pre-trained; The hot spot features of the H5 scene are input into the hot spot follow-up evaluation model to obtain the hot spot follow-up score output by the hot spot follow-up evaluation model; the hot spot follow-up evaluation model is based on taking the hot spot features of the sample H5 scene as input, and manually scoring the sample H5 scene, and the obtained hot spot follow-up score is used as a training target, and is pre-trained; The historical features and the hotspot features of the H5 scene are input into a communication efficiency evaluation model to obtain a communication efficiency score output by the communication efficiency evaluation model; the communication efficiency evaluation model is based on the historical features and the hotspot features of the sample H5 scene as input, and artificially scores the sample H5 scene, and the obtained communication efficiency score is used as a training target, and is pre-trained; The historical features of the H5 scene are input into the historical analogy evaluation model to obtain the historical analogy score output by the historical analogy evaluation model; the historical analogy evaluation model is pre-trained based on the historical features of the sample H5 scene as input and the sample H5 scene is manually scored, and the obtained historical analogy score is used as a training target; The arithmetic mean of the theme fit score, the visual animation score, the content innovation score, the hot spot follow-up score, the communication efficiency score and the historical analogy score is calculated to obtain the hit index of the H5 scene.

7. A device for generating H5 scenes, characterized in that: include: An information acquisition unit, used to acquire demand information input by a user; The demand information includes marketing pictures and marketing theme information required for the advertising business; An image recognition unit is used to input the marketing image into an image entity recognition model to obtain a recognition result output by the image entity recognition model; the image entity recognition model is pre-trained based on taking a sample image as input and taking entities artificially identified in the sample image as training targets; the recognition result includes entities contained in the marketing image; An industry prediction unit is used to perform semantic understanding on the marketing theme information, obtain text features, and input the text features into an industry prediction model to obtain a prediction result output by the industry prediction model; the industry prediction model is pre-trained based on taking sample text features as input and taking the industry type of the sample text features manually identified as a training target; the prediction result includes the industry type of the text features; A picture selection unit, configured to select, from various pictures shown in a preset picture library, the picture containing the entity contained in the marketing picture as a target picture; A component selection unit, used to select, from various components shown in a preset component library, the component whose industry type is the same as the industry type of the copy feature as a target component; A feature construction unit, used to collect the marketing picture, the marketing theme information, the target picture and the target component to construct a material feature; A scene generation unit, used for inputting the material features into a scene generation model to obtain a plurality of H5 scenes output by the scene generation model; The scene generation model is pre-trained based on taking sample material features as input and taking sample H5 scenes corresponding to the sample material features as training targets.

8. The device according to claim 7, characterized in that Also includes: Scene screening unit; The scene screening unit is used for: Perform feature extraction on each of the H5 scenes respectively to obtain data features of each of the H5 scenes; For each H5 scene, based on the data features of the H5 scene, a multi-dimensional score of the H5 scene is obtained, and the arithmetic mean of the multiple scores is calculated to obtain the hit index of the H5 scene; From each of the H5 scenes, select the H5 scene whose value of the hit index is greater than a preset threshold as the target H5 scene; The target H5 scene is presented to the user through a preset interface.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method for generating an H5 scene as described in any one of claims 1-6.

10. A device for generating H5 scenes, characterized in that: include: processor, memory, and bus; The processor is connected to the memory via the bus; The memory is used to store programs, and the processor is used to run programs, wherein the program executes the method for generating an H5 scene described in any one of claims 1 to 6 when running.

Images