CN117857894A - Live cover processing method and device, electronic equipment and computer storage medium - Google Patents

Abstract

The embodiment of the application discloses a live cover processing method, a live cover processing device, electronic equipment and a computer storage medium; in the embodiment of the application, acquiring live broadcast information of a target live broadcast room; according to the live broadcast information, carrying out scene recognition on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room; selecting a preset cover generation strategy corresponding to the live broadcast scene from at least one preset cover generation strategy to obtain a target cover generation strategy; and generating a live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy. According to the embodiment of the application, the quality of the live cover can be improved.

Description

Live cover processing method and device, electronic equipment and computer storage medium

Technical Field

The application relates to the technical field of data processing, in particular to a live cover processing method, a live cover processing device, electronic equipment and a computer storage medium.

Background

With the development of science and technology, live broadcasting is increasingly popular with users. Typically, there will be a corresponding cover for each living room, and a cover for a good living room may attract more users into the living room.

Currently, the cover of the live broadcast room is usually uploaded by the host, or a screenshot of the live broadcast room is adopted as the live broadcast cover. However, the association degree between the live cover which is uploaded independently and the live room is possibly low, the exposure conversion rate of the live room is reduced, the live room screenshot is taken as the live cover, the characteristics cannot be highlighted, and the exposure conversion rate of the live room is also reduced.

Disclosure of Invention

The embodiment of the application provides a live cover processing method, a live cover processing device, electronic equipment and a computer storage medium, which can solve the technical problem of low exposure conversion rate of a live room caused by a live cover.

The embodiment of the application provides a live cover processing method, which comprises the following steps:

acquiring live broadcast information of a target live broadcast room;

according to the live broadcast information, carrying out scene recognition on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room;

selecting a preset cover generation strategy corresponding to the live broadcast scene from at least one preset cover generation strategy to obtain a target cover generation strategy;

and generating a live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

The embodiment of the application provides a live cover processing method, which is applied to a host side and comprises the following steps:

Displaying a target live broadcasting room, wherein the target live broadcasting room comprises a comment display area;

displaying a live broadcast cover corresponding to the target live broadcast room on the comment display area, wherein the live broadcast cover is determined based on a scene identification result of the target live broadcast room;

and responding to the confirmation operation of the live cover, and sending the live cover to a spectator terminal for display.

Accordingly, an embodiment of the present application provides a live cover processing device, including:

the acquisition module is used for acquiring live broadcast information of the target live broadcast room;

the identification module is used for carrying out scene identification on the target live broadcasting room according to the live broadcasting information to obtain a live broadcasting scene of the target live broadcasting room;

the selecting module is used for selecting a preset cover generation strategy corresponding to the live broadcast scene from at least one preset cover generation strategy to obtain a target cover generation strategy;

and the generation module is used for generating the live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

Accordingly, an embodiment of the present application provides a live cover processing device, including:

the first display module is used for displaying a target live broadcast room, wherein the target live broadcast room comprises a comment display area;

The second display module is used for displaying a live broadcast cover corresponding to the target live broadcast room on the comment display area, wherein the live broadcast cover is determined based on a scene identification result of the target live broadcast room;

and the operation response module is used for responding to the confirmation operation of the live cover and sending the live cover to the audience terminal for display.

In addition, the embodiment of the application also provides electronic equipment, which comprises a processor and a memory, wherein the memory stores a computer program, and the processor is used for running the computer program in the memory to realize the live cover processing method provided by the embodiment of the application.

In addition, the embodiment of the application further provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program is suitable for being loaded by a processor to execute any live cover processing method provided by the embodiment of the application.

In addition, the embodiment of the application also provides a computer program product, which comprises a computer program, and the computer program realizes any live cover processing method provided by the embodiment of the application when being executed by a processor.

In the embodiment of the application, acquiring live broadcast information of a target live broadcast room; according to the live broadcast information, carrying out scene identification on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room; selecting a preset cover generation strategy corresponding to the live scene from at least one preset cover generation strategy to obtain a target cover generation strategy; and generating the live cover of the target live broadcasting room according to the live broadcasting information through the target cover generation strategy, so that when live broadcasting scenes are different, the cover generation strategy is different, the quality and the stability of the live cover are improved, and the exposure conversion rate of the live broadcasting room is improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic view of a live cover processing procedure according to an embodiment of the present application;

fig. 2 is a flow chart of a live cover processing method according to an embodiment of the present application;

FIG. 3 is a schematic diagram illustrating a flow of another live cover processing method according to an embodiment of the present disclosure;

FIG. 4 is a schematic diagram of a comment display area provided by an embodiment of the present application;

FIG. 5 is another schematic view of a comment display area provided by an embodiment of the present application;

FIG. 6 is a schematic diagram of a live cover provided by an embodiment of the present application;

FIG. 7 is a schematic illustration of another live cover provided in an embodiment of the present application;

FIG. 8 is a schematic diagram of an association control provided by an embodiment of the present application;

FIG. 9 is a schematic diagram of candidate notes provided by an embodiment of the present application;

FIG. 10 is a schematic diagram of a target note provided by an embodiment of the present application;

FIG. 11 is a schematic diagram of a toggle control provided by an embodiment of the present application;

FIG. 12 is a flow chart of a training process of a model provided by an embodiment of the present application;

FIG. 13 is a flowchart illustrating another method for processing live covers according to an embodiment of the present disclosure;

FIG. 14 is a schematic structural view of a model provided in an embodiment of the present application;

FIG. 15 is a schematic diagram of another method for processing live covers according to an embodiment of the present application;

FIG. 16 is a schematic diagram of another method for processing live covers according to an embodiment of the present application;

fig. 17 is a schematic diagram of interaction among a anchor side, a server, and a viewer side according to an embodiment of the present application;

FIG. 18 is a schematic structural diagram of a live cover processing device according to an embodiment of the present disclosure;

FIG. 19 is a schematic diagram of another apparatus for processing live covers according to an embodiment of the present disclosure;

fig. 20 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.

The embodiment of the application provides a live cover processing method, a live cover processing device, electronic equipment and a computer storage medium. The live cover processing device can be integrated in electronic equipment, and the electronic equipment can be a server or a terminal and other equipment.

The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, network acceleration services (Content Delivery Network, CDN), basic cloud computing services such as big data and an artificial intelligent platform.

The terminal may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal may include at least one of a head office and a viewer office, and the terminal and the server may be directly or indirectly connected through wired or wireless communication, which is not limited herein.

For example, as shown in fig. 1, the main broadcasting end acquires the live broadcast information of the target live broadcast room, and sends the live broadcast information to the server. The method comprises the steps that a server carries out scene identification on a target live broadcasting room according to live broadcasting information to obtain live broadcasting scenes of the target live broadcasting room, a preset cover generation strategy corresponding to the live broadcasting scenes is selected from at least one preset cover generation strategy to obtain a target cover generation strategy, and the live broadcasting covers of the target live broadcasting room are generated according to the live broadcasting information through the target cover generation strategy and are sent to audience terminals to be displayed.

In addition, "plurality" in the embodiments of the present application means two or more. "first" and "second" and the like in the embodiments of the present application are used for distinguishing descriptions and are not to be construed as implying relative importance.

The following will describe in detail. The following description of the embodiments is not intended to limit the preferred embodiments.

In this embodiment, description will be made from the viewpoint of a live cover processing apparatus, and for convenience, a live cover processing method of the present application will be described in detail below by integrating the live cover processing apparatus into a terminal, that is, by using the terminal as an execution subject.

Referring to fig. 2, fig. 2 is a flowchart illustrating a live cover processing method according to an embodiment of the present application. The live cover processing method can comprise the following steps:

s201, acquiring live broadcast information of a target live broadcast room.

The target live broadcast room may be a live broadcast room that is being live broadcast, or may be a live broadcast room that has been created but not yet opened, which is not limited herein.

Live information refers to information related to a target live room, which may include at least one of static information and dynamic information. The static information can be, for example, personal information of a host broadcast of the target living broadcast room, and the dynamic information can be, for example, at least one of a living broadcast picture of the target living broadcast room, a theme of the target living broadcast room, comment information of the target living broadcast room, commodity information in the target living broadcast room and a document of the target living broadcast room.

The mode of the live broadcast information can be set according to actual conditions, for example, the mode of the live broadcast information can include at least one of a text mode and a visual mode, when the mode of the live broadcast information is the text mode, the live broadcast information can be at least one of personal information of a main broadcast of a target live broadcast room, a theme of the target live broadcast room, comment information of the target live broadcast room, commodity information in the target live broadcast room and a file of the target live broadcast room, and when the mode of the live broadcast information is the visual mode, the live broadcast information can be a live broadcast picture of the target live broadcast room, the live broadcast picture can be a historical live broadcast picture or a current live broadcast picture, and the embodiment of the application is not limited herein.

Optionally, the terminal may acquire the live broadcast information of the target live broadcast room in response to the triggering operation of the recommendation control, or the terminal may also periodically acquire the live broadcast information of the target live broadcast room. When the live broadcast information of the target live broadcast room is periodically acquired, the periods of different target live broadcast rooms may be the same or different, and the embodiment of the present application is not limited herein.

The manner in which the terminal obtains the live broadcast information of the target live broadcast room can be set according to actual situations, and embodiments of the present application are not limited herein.

S202, carrying out scene recognition on the target live broadcasting room according to the live broadcasting information to obtain a live broadcasting scene of the target live broadcasting room.

The live broadcast scene refers to a category of the target live broadcast room, for example, the live broadcast scene may be at least one of a self-learning live broadcast scene, an outdoor sports live broadcast scene, an indoor body-building live broadcast scene, a live broadcast with goods scene, an interactive live broadcast scene, a chat live broadcast scene, a live broadcast scene including a face, and a live broadcast scene not including a face.

The terminal can acquire a preset mapping table, wherein the preset mapping table comprises candidate live broadcast information and candidate live broadcast scenes, and the terminal can determine the candidate live broadcast scenes corresponding to the candidate live broadcast information matched with the live broadcast information in the preset mapping table as live broadcast scenes of the target live broadcast room.

Or the terminal can also input the live broadcast information into the trained scene recognition model to recognize the scene, so as to obtain the live broadcast scene of the target live broadcast room.

The manner of identifying the scene of the target live broadcast room according to the live broadcast information can be selected according to actual situations, and the embodiment of the application is not limited herein.

In some embodiments, when the live information includes a live screen of the target live room and text information associated with the target live room, performing scene recognition on the target live room according to the live information to obtain a live scene of the target live room, including:

extracting features of the live broadcast pictures to obtain visual mode features corresponding to the target live broadcast room;

extracting the characteristics of the text information to obtain the text modal characteristics corresponding to the target live broadcasting room;

fusion processing is carried out on the video modality features and the text modality features to obtain multi-modality features corresponding to the target live broadcasting room;

and carrying out scene recognition on the target live broadcasting room according to the multi-mode characteristics to obtain a live broadcasting scene of the target live broadcasting room.

The text information associated with the target live broadcasting room may be set according to actual situations, for example, the text information associated with the target live broadcasting room may include at least one of personal information of a host broadcasting of the target live broadcasting room, a theme of the target live broadcasting room, comment information of the target live broadcasting room, commodity information in the target live broadcasting room, and a document of the target live broadcasting room, and embodiments of the present application are not limited herein.

The live broadcast picture of the target live broadcast room can comprise at least one frame, the live broadcast picture of the target live broadcast room can be a historical live broadcast picture of the target live broadcast room, and also can be a current live broadcast picture of the target live broadcast room, and the embodiment of the application is not limited herein.

In the embodiment of the application, the feature extraction is carried out on the live broadcast picture to obtain the corresponding visual mode feature of the target live broadcast room; extracting the characteristics of the text information to obtain the text modal characteristics corresponding to the target live broadcasting room; fusion processing is carried out on the video modality features and the text modality features to obtain multi-modality features corresponding to the target live broadcasting room; and according to the multi-mode characteristics, scene identification is carried out on the target live broadcasting room, so as to obtain the live broadcasting scene of the target live broadcasting room, realize the identification of the live broadcasting scene of the target live broadcasting room through the multi-mode, and improve the identification accuracy of the live broadcasting scene.

In some embodiments, the process of extracting features of the live broadcast picture to obtain the corresponding visual mode features of the target live broadcast room may include:

extracting features of the live broadcast pictures to obtain first visual mode features corresponding to the target live broadcast room;

extracting features of the images uploaded by the anchor to obtain second visual mode features corresponding to the target live broadcasting room;

And carrying out fusion processing on the first visual mode characteristics and the second visual mode characteristics to obtain the visual mode characteristics corresponding to the target live broadcasting room.

Because the image uploaded by the anchor may contain the requirement of the anchor on the live cover and has unique characteristics of the anchor, in the embodiment of the application, the feature extraction is performed on the live image to obtain the first visual mode feature corresponding to the target live room; extracting features of the images uploaded by the anchor to obtain second visual mode features corresponding to the target live broadcasting room; and the first visual mode characteristics and the second visual mode characteristics are fused to obtain the visual mode characteristics corresponding to the target live broadcasting room, so that the relevance between the live broadcasting cover of the target live broadcasting room and the target live broadcasting room is high, and the requirement of a host can be met.

In some embodiments, the target live broadcast room includes at least one, and the scene recognition is performed on the target live broadcast room according to the live broadcast information to obtain a live broadcast scene of the target live broadcast room, including:

generating a cover generation task according to the live broadcast information;

sending the cover generating task to a task queue for sorting;

acquiring a target cover generation task from a task queue;

And carrying out scene recognition on the target live broadcasting room according to the target cover generation task to obtain a live broadcasting scene of the target live broadcasting room.

The target cover generation task may be the same as or different from the cover generation task.

In the embodiment of the application, the target live broadcast room comprises at least one, and a cover generation task is generated according to live broadcast information; sending the cover generating task to a task queue for sorting; acquiring a target cover generation task from a task queue; and carrying out scene recognition on the target live broadcasting room according to the target cover generating task to obtain the live broadcasting scene of the target live broadcasting room, thereby ensuring the recognition order of the live broadcasting scene of the target live broadcasting room.

Optionally, the manner in which the cover generation tasks are sent to the task queue for ordering may be set according to actual situations, for example, each cover generation task in the task queue may be ordered according to the feature information of the target living broadcast room, or each cover generation task in the task queue may be ordered according to the time of the received cover generation task (the earlier the time of the received cover generation task, the earlier the order of the cover generation task in the task queue, the later the time of the received cover generation task, and the later the order of the cover generation task in the task queue), which is not limited herein.

When sequencing each cover generation task in the task queue according to the characteristic information of the target live broadcasting room, sending the cover generation task to the task queue for sequencing, wherein the sequencing comprises the following steps:

transmitting the cover generating task to a task queue;

acquiring characteristic information of a target living broadcast room corresponding to each cover generation task in a task queue;

and sequencing all cover generating tasks in the task queue according to the characteristic information.

The feature information of the target live broadcasting room can refer to the features of the target live broadcasting room. The feature information of the target living broadcast room may be set according to actual situations, for example, the feature information of the target living broadcast room may include at least one of the number of viewers entering the target living broadcast room, the amount of goods sold in the target living broadcast room, the amount of carried goods in the target living broadcast room, the number of fans of the anchor in the target living broadcast room, the number of fans of the account corresponding to the target living broadcast room, and the living broadcast level of the anchor in the target living broadcast room.

The manner of ordering the cover generation tasks in the task queue according to the feature information may be set according to actual situations, for example, when the feature information of the target living broadcast room is the number of fans of the main broadcasting in the target living broadcast room, the more fans are, the fewer fans are, the order of the cover generation tasks corresponding to the target living broadcast room in the task queue is, the more the order of the cover generation tasks corresponding to the target living broadcast room in the task queue is, when the feature information of the target living broadcast room is the number of audience entering the target living broadcast room, the fewer the number of audience entering the target living broadcast room, the more the audience entering the target living broadcast room is, and the more the order of the cover generation tasks corresponding to the target living broadcast room in the task queue is, which is not limited in this embodiment.

In the embodiment of the application, a cover generation task is sent to a task queue; acquiring characteristic information of a target living broadcast room corresponding to each cover generation task in a task queue; sequencing all cover generation tasks in the task queue according to the characteristic information, determining the order of generating the live covers of the target live broadcasting room according to the characteristic information of the target live broadcasting room, and improving the rationality of generating the live covers of the target live broadcasting room so as to further improve the exposure conversion rate of all the target live broadcasting rooms in the platform.

S203, selecting a preset cover generation strategy corresponding to the live scene from at least one preset cover generation strategy to obtain a target cover generation strategy.

The terminal can store at least one preset cover generation strategy and candidate live broadcast scenes in an associated mode, and then uses the preset cover generation strategy associated with the candidate live broadcast scenes matched with the live broadcast scenes as the preset cover generation strategy corresponding to the live broadcast scenes.

The number of at least one preset cover generation policy may be set according to practical situations, and embodiments of the present application are not limited herein.

The form of the at least one preset cover generation policy may be set according to practical situations, for example, the at least one preset cover generation policy may be at least one preset cover generation function or at least one preset cover generation model, which is not limited herein.

The type of the at least one preset cover generation model may be set according to actual situations, and the types of different preset cover generation models may be the same or different, for example, the at least one preset cover generation model may be at least one of a convolutional neural network model (Convolutional Neural Networks, CNN), a diffusion model (diffusion model), and a recurrent neural network (Recurrent Neural Network, RNN).

S204, generating a live cover of the target live room according to the live information through a target cover generation strategy.

When the live broadcast information includes multiple information, the terminal may generate a live broadcast cover of the target live broadcast room according to the multiple information in the live broadcast information through the target cover generation policy, or when the live broadcast information includes a live broadcast picture, the terminal may also generate the live broadcast cover of the target live broadcast room according to the live broadcast picture in the live broadcast information through the target cover generation policy.

In the related art, the cover is generated according to only one cover generation strategy, which results in better quality of the cover generated in some cases and worse quality of the cover generated in some cases, so that the stability of the quality of the cover generated is worse.

In the embodiment of the application, at least one preset cover generation strategy is preset, scene identification is performed on a target live broadcasting room according to live broadcasting information, a live broadcasting scene of the target live broadcasting room is obtained, a preset cover generation strategy corresponding to the live broadcasting scene is selected from the at least one preset cover generation strategy, the target cover generation strategy is obtained, the live broadcasting cover of the target live broadcasting room is generated according to the live broadcasting information through the target cover generation strategy, the live broadcasting covers of different live broadcasting scenes can be generated according to different cover generation strategies, the quality of the live broadcasting cover is improved, and the stability of the quality of the live broadcasting cover is ensured.

In some embodiments, generating, by the target cover generation policy, a live cover of the target live room according to the live information includes:

generating a first candidate live broadcast cover of the target live broadcast room according to live broadcast information through a target cover generation strategy;

acquiring face information of a target live broadcasting room;

and according to the face information, adjusting the face area in the first candidate live broadcast cover to obtain the live broadcast cover of the target live broadcast room.

The face information of the target live broadcasting room can be extracted through live broadcasting pictures in the live broadcasting information, or the terminal can also extract the face information of the target live broadcasting room in the process of carrying out scene identification on the target live broadcasting room according to the live broadcasting information to obtain the live broadcasting scene of the target live broadcasting room.

After the terminal obtains the first candidate live broadcast cover, the terminal can detect whether the first candidate live broadcast cover has a face area, if so, the terminal calculates the face similarity according to the face information in the face area and the face information of the target live broadcast room, if the face similarity does not meet the preset similarity, the terminal adjusts the face area in the first candidate live broadcast cover according to the face information to obtain the live broadcast cover of the target live broadcast room, and if the face similarity meets the preset similarity, the first candidate live broadcast cover is determined to be the live broadcast cover of the target live broadcast room, so that the face in the live broadcast cover of the target live broadcast room is matched with the face of the target live broadcast room.

In some embodiments, generating, by the target cover generation policy, a live cover of the target live room according to the live information includes:

generating a second candidate live broadcast cover of the target live broadcast room according to the live broadcast information through a target cover generation strategy;

determining the quality fraction corresponding to the second candidate live cover;

and if the quality score meets the preset score condition, determining the second candidate live broadcast cover as the live broadcast cover of the target live broadcast room.

At this time, after determining the quality score corresponding to the second candidate live cover, the method further includes:

If the quality score does not meet the preset score condition, updating the live broadcast information of the target live broadcast room to obtain updated live broadcast information;

and returning the updated live broadcast information as live broadcast information to execute the step of identifying the scene of the target live broadcast room according to the live broadcast information to obtain the live broadcast scene of the target live broadcast room.

In the embodiment of the application, a second candidate live broadcast cover of the target live broadcast room is generated according to live broadcast information through a target cover generation strategy; determining the quality fraction corresponding to the second candidate live cover; if the quality score meets the preset score condition, determining the second candidate live broadcast cover as the live broadcast cover of the target live broadcast room, and if the quality score does not meet the preset score condition, updating the live broadcast information of the target live broadcast room to obtain updated live broadcast information; and returning the updated live broadcast information to execute the step of identifying the target live broadcast room to obtain the live broadcast scene of the target live broadcast room according to the live broadcast information, so that when the quality of the second candidate live broadcast cover meets the condition, the second candidate live broadcast cover is determined to be the live broadcast cover, and the quality of the live broadcast cover is ensured.

For example, acquiring live information n1 of a target live broadcasting room, performing scene recognition on the target live broadcasting room according to the live information n1 to obtain live broadcasting scene s1 of the target live broadcasting room, selecting a preset cover generation strategy corresponding to the live broadcasting scene s1 from at least one preset cover generation strategy to obtain the target cover generation strategy m1, generating a second candidate live broadcasting cover c1 of the target live broadcasting room according to the live broadcasting information n1 through the target cover generation strategy m1, determining a quality score corresponding to the second candidate live broadcasting cover c1, wherein the quality score corresponding to the second candidate live broadcasting cover c1 does not meet a preset score condition, updating the live broadcasting information of the target live broadcasting room to n2, performing scene recognition on the target live broadcasting room according to the live broadcasting information n2 to obtain live broadcasting scene s2 of the target live broadcasting room, selecting the preset cover generation strategy corresponding to the live broadcasting scene s2 from the at least one preset cover generation strategy to obtain the target cover generation strategy m2, generating the second candidate cover c2 according to the live broadcasting information n2, determining the quality score corresponding to the second candidate cover c2 of the live broadcasting room according to the live broadcasting information n2, and determining the quality score corresponding to the second candidate cover c2 of the live broadcasting room as the target live broadcasting score.

Optionally, the manner of updating the live broadcast information of the target live broadcast room to obtain updated live broadcast information may be set according to practical situations, for example, when the live broadcast information includes a live broadcast picture of the target live broadcast room, the live broadcast picture in the live broadcast information may be updated so as to update the live broadcast information of the target live broadcast room, and when the live broadcast information includes comment information of the target live broadcast room, the comment information in the live broadcast information may be updated so as to update the live broadcast information.

Optionally, the manner of determining the quality score corresponding to the second candidate live cover may be set according to the actual situation, which is not limited herein in this embodiment.

For example, the process of determining the quality score corresponding to the second candidate live cover may be:

aesthetic scoring processing is carried out on the second candidate live broadcast covers, and the aesthetic scores of the second candidate live broadcast covers are obtained;

and determining the quality score corresponding to the second candidate live cover according to the aesthetic score of the second candidate live cover.

The aesthetic score of the second candidate live cover may be higher, the quality score corresponding to the second candidate live cover is higher, the aesthetic score of the second candidate live cover is lower, and the quality score corresponding to the second candidate live cover is lower.

For another example, the process of determining the quality score corresponding to the second candidate live cover may also be:

performing target detection on the second candidate live broadcast cover to obtain a target detection result;

and determining the quality fraction corresponding to the second candidate live cover according to the target detection result.

The method for performing object detection on the second candidate live cover may be selected according to practical situations, for example, the object detection may be performed on the second candidate live cover through a direction gradient histogram (Histogramof Oriented Gradient, HOG) or an object detection model, which is not limited herein.

Alternatively, the target in the target detection may be set according to the actual situation, for example, the target in the target detection may be a hand or a face, which is not limited herein.

Optionally, according to the target detection result, the process of determining the quality score corresponding to the second candidate live cover may be, for example: and determining whether the target in the second candidate live cover is abnormal or not according to the target detection result, wherein if the target is abnormal, the quality score corresponding to the second candidate live cover is a first score, and if the target is not abnormal, the quality score corresponding to the second candidate live cover is a second score which is larger than the first score.

In some embodiments, when the live broadcast information includes at least one frame of live broadcast picture, different live broadcast covers may be generated according to one frame of live broadcast picture in the at least one frame of live broadcast picture, or different live broadcast covers may be generated according to different live broadcast pictures in the at least one frame of live broadcast picture, that is, at this time, a plurality of live broadcast covers may be generated, and then the plurality of live broadcast covers are sent to the main broadcasting end to be displayed, and in response to a confirmation operation on the live broadcast covers, the live broadcast cover corresponding to the confirmation operation is determined as the live broadcast cover of the target live broadcast room.

Or when the live broadcast information comprises at least one frame of live broadcast picture, generating a live broadcast cover of the target live broadcast room according to the live broadcast information by using a target cover generation strategy, wherein the method comprises the following steps:

dividing at least one frame of live broadcast picture to obtain a picture area corresponding to each frame of live broadcast picture in at least one frame of live broadcast picture;

screening out a target picture area from picture areas corresponding to each frame of live broadcast picture;

splicing the target picture areas to obtain a target live broadcast picture;

and generating a live broadcast cover of the target live broadcast room according to the target live broadcast picture through a target cover generation strategy.

In this case, if the live cover is directly generated according to the live picture, the quality of the live cover is reduced because the picture area corresponding to each frame of live picture may have a poor (blurred) picture area.

Therefore, in the embodiment of the application, when the live broadcast information includes at least one frame of live broadcast picture, the at least one frame of live broadcast picture is subjected to segmentation processing, so as to obtain a picture area corresponding to each frame of live broadcast picture in the at least one frame of live broadcast picture; screening out target picture areas meeting quality conditions from picture areas corresponding to each frame of live broadcast picture; splicing the target picture areas to obtain a target live broadcast picture; and generating a live cover of the target live broadcasting room according to the target live broadcasting picture through a target cover generation strategy, so that the quality of the obtained live cover is higher.

For example, at least one frame of live broadcast picture comprises a live broadcast picture h1 and a live broadcast picture h2, the live broadcast picture h1 is segmented to obtain a picture area h11 and a picture area h12 corresponding to the live broadcast picture h1, the live broadcast picture h2 is segmented to obtain a picture area h21 and a picture area h22 corresponding to the live broadcast picture h2, the similarity between the picture area h11 and the picture area h21 meets a preset similarity condition, the similarity between the picture area h12 and the picture area h22 meets the preset similarity condition, the quality score of the picture area h11 is larger than the quality score of the picture area h21, the quality score of the picture area h12 is smaller than the quality score of the picture area h22, and then the picture area h11 and the picture area h22 are spliced into the target live broadcast picture.

In some embodiments, generating, by the target cover generation policy, a live cover of the target live room according to the live information includes:

generating live covers of different styles of a target live room according to live information through a target cover generation strategy;

after generating the live cover of the target live room according to the live information through the target cover generation strategy, the method further comprises the following steps:

obtaining style characteristics of target account numbers of all audience terminals;

and displaying the live broadcast cover of the style matched with the style characteristics as the live broadcast cover of the target live broadcast room in the audience terminal.

According to the method, through a target cover generation strategy, different styles of live covers of a target live room are generated according to live information, and style characteristics of target account numbers of all audience terminals are obtained; the live cover of the style matched with the style characteristics is used as the live cover of the target live broadcasting room in the audience terminal to be displayed, so that the live covers of the target live broadcasting rooms in different audience terminals are different, the live cover of the target live broadcasting room meets the requirements of the audience more, the probability of the audience entering the target live broadcasting room is improved, and the exposure conversion rate of the target live broadcasting room is further improved.

For example, the live covers of different styles include a live cover c1, a live cover c2 and a live cover c3, the style of the live cover c1 is f1, the style of the live cover c2 is f2, the style of the live cover c3 is f3, the style of the target account a1 of the viewer end d1 is f1, the style of the target account a2 of the viewer end d2 is f3, the style of the target account a3 of the viewer end d3 is f2, the live cover c1 is used as the live cover of the target living room in the viewer end d1, the live cover c3 is used as the live cover of the target living room in the viewer end d2, and the live cover c2 is used as the live cover of the target living room in the viewer end d 3.

In some embodiments, acquiring live information of a target live room includes:

acquiring a cover generation period of a target live broadcasting room;

acquiring live broadcast information of a target live broadcast room according to the cover generation period;

after generating the live cover of the target live room according to the live information through the target cover generation strategy, the method further comprises the following steps:

displaying a live cover at a viewer end;

acquiring the number of audiences entering a target living room within a preset time length after the live cover is displayed;

according to the number of audiences, adjusting the cover generation period to obtain an adjusted cover generation period;

And taking the adjusted cover generation period as a cover generation period, and returning to the step of executing the cover generation period of the acquisition target live broadcasting room.

For example, according to a cover generation period t1, acquiring live information n1 of a target live broadcasting room, generating a live broadcasting cover of the target live broadcasting room as a live broadcasting cover c1 according to a live broadcasting information n1 through a target cover generation strategy, displaying the live broadcasting cover c1 at a viewer end, acquiring the number of viewers entering the target live broadcasting room within a preset time length after displaying the live broadcasting cover c1 as b1, and the cover generation period corresponding to b1 as t2, adjusting the cover generation period t1 as a cover generation period t2, then according to the cover generation period t2, acquiring live broadcasting information n2 of the target live broadcasting room, generating the live broadcasting cover of the target live broadcasting room as a live broadcasting cover c2 according to the live broadcasting information n2 through the target cover generation strategy, displaying the live broadcasting cover c2 at the viewer end, acquiring the number of viewers entering the target live broadcasting room within the preset time length after displaying the live broadcasting cover c2 as b2, and directly returning to the step of acquiring the live broadcasting information of the target live broadcasting room according to the cover generation period t2 without adjusting the cover generation period t 2.

In the embodiment of the application, a cover generation period of a target living room is acquired; acquiring live broadcast information of a target live broadcast room according to the cover generation period, and displaying the live broadcast cover at the audience terminal; acquiring the number of audiences entering a target living room within a preset time length after the live cover is displayed; according to the number of audiences, adjusting the cover generation period to obtain an adjusted cover generation period; and taking the adjusted cover generation period as a cover generation period, and returning to the step of executing the cover generation period of the target living broadcast room, so that the period of updating the living broadcast cover can be adjusted in real time according to the number of audiences in the target living broadcast room, and the number of the audiences entering the target living broadcast room is further improved.

Alternatively, since the fewer the number of viewers entering the target living room, the lower the attraction of the live seal of the target living room to the viewers, the cover generation period is adjusted according to the number of viewers, and the smaller the number of viewers, the shorter the adjusted cover generation period, the greater the number of viewers, and the longer the adjusted cover generation period.

From the above, in the embodiment of the present application, live broadcast information of a target live broadcast room is obtained; according to the live broadcast information, carrying out scene identification on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room; selecting a preset cover generation strategy corresponding to the live scene from at least one preset cover generation strategy to obtain a target cover generation strategy; and generating the live cover of the target live broadcasting room according to the live broadcasting information through the target cover generation strategy, so that when live broadcasting scenes are different, the cover generation strategy is different, the quality and the stability of the live cover are improved, and the exposure conversion rate of the live broadcasting room is improved.

The method for processing the live cover provided in the present application is further described with reference to fig. 3, and in this embodiment, the main cast end is used as an execution subject.

S301, displaying a target live broadcast room, wherein the target live broadcast room comprises a comment display area.

The position of the comment display area in the target live broadcast room may be set according to the actual situation, for example, the comment display area may be below the target live broadcast room as shown in fig. 4, for example, the comment display area may be above the target live broadcast room as shown in fig. 5, which is not limited herein.

S302, displaying a live broadcast cover corresponding to the target live broadcast room on the comment display area, wherein the live broadcast cover is determined based on a scene identification result of the target live broadcast room.

The method for determining the live cover based on the scene recognition result of the target live room may be set according to actual situations, for example, the method for determining the live cover based on the scene recognition result of the target live room may be the method described in the live cover processing method embodiment, or may not be the method described in the live cover processing method embodiment, which is not limited herein.

The corresponding live cover of the target live room may include at least one. For example, displaying the corresponding live cover of the target live room on the comment display area may be as shown in fig. 6.

And S303, responding to the confirmation operation of the live cover, and sending the live cover to the audience terminal for display.

After the live cover is displayed, the host side can respond to the confirmation operation of the live cover and send the live cover to the audience side for display.

Optionally, when the live broadcast covers corresponding to the target live broadcast room include a plurality of live broadcast covers, the anchor terminal may respond to the confirmation operation on the live broadcast covers, and send the live broadcast covers corresponding to the confirmation operation to the audience terminal for display.

For example, as shown in fig. 7, the generated live cover includes a live cover c1, a live cover c2, a live cover c3, and a live cover c4, and the anchor terminal may send the live cover c4 to the audience terminal for display in response to the confirmation operation on the live cover c 4.

Optionally, the type of the confirmation operation on the live cover may be set according to the actual situation, for example, the confirmation operation on the live cover may be a click operation on the live cover, or the comment display area may further include a digital control, and the confirmation operation on the live cover may be a click operation on the digital control.

When the confirmation operation of the live cover is a click operation of the digital control, the live cover is sent to the audience end for display in response to the confirmation operation of the live cover, and the method comprises the following steps:

responding to clicking operation of the digital control, and determining the numbers displayed in the digital control;

and sending the live cover corresponding to the number to the audience terminal for display.

The shape of the digital control may be set according to the actual situation, for example, when the digital control is a cube, the digital control may also be referred to as a virtual sub (for example, the digital control may be as shown in fig. 7), where the number displayed in the digital control may be an upward number in the virtual sub or a number facing the user, or the digital control may also be a circular control, where the number displayed in the digital control may be a number displayed on the circular control.

In the embodiment of the application, responding to clicking operation of the digital control, and determining the number displayed in the digital control; and sending the live cover corresponding to the number to the audience end for display, so that the interestingness in the process of determining the live cover is improved, and the probability of adopting the live cover by a host is improved.

For example, the generated live cover includes a live cover c1, a live cover c2, a live cover c3 and a live cover c4, the digital control is a virtual sub-body, after the host clicks the virtual sub-body, the virtual sub-body is 3 in upward digital, the live cover c3 is used as the live cover of the target live room, and the live cover c3 is sent to the audience for displaying.

In some embodiments, the generated live cover includes a plurality of live covers, the comment display area includes a digital control, and after the live cover corresponding to the target live room is displayed on the comment display area, the method further includes:

responding to the triggering operation of the digital control, and determining the number displayed in the digital control;

and displaying the live broadcast covers corresponding to the numbers on the comment display area.

For example, the generated live cover includes a live cover c1, a live cover c2, a live cover c3 and a live cover c4, the host side firstly displays the live cover c1 on the comment display area, after clicking the digital control by the host side, the host side shakes the digital control in response to the triggering operation of the digital control to obtain the number shown in the digital control to be 3, the host side replaces the live cover c1 displayed on the comment display area with the live cover c3, and finally, the host side can send the live cover c3 to the audience side for display in response to the confirmation operation of the live cover c 3.

In some embodiments, the generated live cover includes a plurality of live covers, the comment display area includes a digital control, and after the live cover corresponding to the target live room is displayed on the comment display area, the method further includes:

responding to the triggering operation of the digital control, and determining the number displayed in the digital control;

and displaying live covers corresponding to the target number on the comment display area, wherein the target number is a number corresponding to the number.

For example, the generated live cover includes a live cover c1, a live cover c2, a live cover c3 and a live cover c4, the main broadcasting end displays the live cover c1 on the comment display area according to the default number, after the main broadcasting clicks the digital control, the main broadcasting end shakes the digital control in response to the triggering operation of the digital control to obtain the number displayed in the digital control as 4, that is, the live cover is displayed on the comment display area, and then the main broadcasting end displays the live cover c1, the live cover c2, the live cover c3 and the live cover c4 on the comment display area.

In some embodiments, the comment display area further includes an association control, and after displaying the live cover corresponding to the target live room on the comment display area, the method further includes:

Responding to triggering operation of the association control, and displaying a note list corresponding to the target anchor account in the target live broadcasting room, wherein the note list comprises at least one candidate note;

in response to a selection operation of a target note in the at least one candidate note, associating the target note with the live cover;

in response to a confirmation operation on the live cover, transmitting the live cover to a viewer end for display, including:

and responding to the confirmation operation of the live cover, sending the live cover to a spectator terminal for displaying, and displaying target notes associated with the live cover in a target live room in the spectator terminal.

The position of the association control on the comment display area may be set according to the actual situation, for example, the association control may be located at a position with a preset distance from the live cover, as shown in fig. 6, or the association control may be located at another position on the comment display area, for example, as shown in fig. 8.

The type of the triggering operation of the associated control may be set according to the actual situation, for example, the triggering operation of the associated control may be a clicking operation or a long-press operation, which is not limited herein.

Alternatively, the manner of displaying the note list corresponding to the target anchor account in the target live broadcasting room may be set according to the actual situation, for example, the note list may be displayed on the comment display area as shown in 901 in fig. 9, or the note list may be displayed in a pop-up window form as shown in 902 in fig. 9, which is not limited herein.

Alternatively, the form of displaying the target note in the target living room on the viewer side may be set according to the actual situation, for example, the target note may be displayed in the comment display area as shown in 1001 in fig. 10, or the target note may be displayed in the form of a card in the target living room on the viewer side as shown in 1002 in fig. 10, which is not limited herein.

In the embodiment of the application, in response to triggering operation of the association control, a note list corresponding to a target anchor account in a target live broadcasting room is displayed, wherein the note list comprises at least one candidate note; in response to a selection operation of a target note in at least one candidate note, associating the target note with the live cover, in response to a confirmation operation of the live cover, transmitting the live cover to the audience terminal for display, and displaying the target note associated with the live cover in a target live room in the audience terminal, so that the audience terminal can display the target note in the target live room after entering the target live room.

Optionally, the displayed candidate notes may be all candidate notes in the note list, that is, the displayed candidate notes are at least one candidate note, or the displayed candidate notes may also be part of the candidate notes in the note list, where, in response to a triggering operation on the association control, a note list corresponding to the target anchor account in the target live room is displayed, including:

responding to the triggering operation of the associated control, and acquiring a scene identification result of the target live broadcasting room;

screening candidate notes matched with the scene recognition result from a note list corresponding to a target anchor account in a target live broadcasting room;

and displaying the candidate notes matched with the scene recognition result.

The scene recognition result may be a scene recognition result obtained by the live cover processing method embodiment, or may not be a scene recognition result obtained by the live cover processing method embodiment, which is not limited herein.

The candidate note matching with the scene recognition result may also be understood as that the scene recognition result of the candidate note matches with the scene recognition result of the target live broadcast room, for example, the scene recognition result of the target live broadcast room is a live broadcast scene with goods, and the scene recognition result of the candidate note matches with the scene recognition result of the target live broadcast room.

In the embodiment of the application, a scene identification result of a target live broadcast room is obtained in response to triggering operation of an associated control; screening candidate notes matched with the scene recognition result from a note list corresponding to a target anchor account in a target live broadcasting room; and displaying the candidate notes matched with the scene recognition result, and displaying the candidate notes matched with the scene recognition result of the target living broadcast room at the audience side in the target living broadcast room, so that the audience can better know the related information of the target living broadcast room through the candidate notes.

In some embodiments, the target live room further includes a switch control, and after displaying the live cover corresponding to the target live room on the comment display area, the method further includes:

responding to the triggering operation of the switching control, and starting a cover switching function;

after the live cover is sent to the audience end for display in response to the confirmation operation of the live cover, the method further comprises the following steps:

detecting whether a scene recognition result changes according to the cover switching function;

if the scene identification result changes, updating the live cover, and sending the updated live cover to the audience terminal for display.

The position of the switching control on the target live room may be set according to the actual situation, for example, the switching control may be located at a position with a preset distance from the target of the live cover, as shown in fig. 6, or the switching control may be located at another position of the target live room, as shown in fig. 11, which is not limited herein.

The process of updating the live cover can refer to the process of generating the live cover of the target live room according to the live information by the target cover generation strategy, and the embodiment of the application is not repeated here.

In the embodiment of the application, whether a scene recognition result changes or not is detected according to the cover switching function; if the scene identification result changes, the live cover is updated, and the updated live cover is sent to the audience end for display, so that the live cover changes along with the scene change of the target live room according to the field of the target live room.

From the above, in the embodiment of the application, the target live broadcasting room is displayed, the target live broadcasting room comprises a comment display area, and the live broadcasting cover corresponding to the target live broadcasting room is displayed on the comment display area, wherein the live broadcasting cover is a cover determined based on the scene identification result of the target live broadcasting room, and the live broadcasting cover is sent to the audience terminal for display in response to the confirmation operation of the live broadcasting cover, so that the live broadcasting cover is displayed in the comment display area, whether the live broadcasting cover is adopted or not is conveniently confirmed by a host, and the utilization rate of the live broadcasting cover is improved.

In the embodiment, taking at least one preset cover generation model as at least one preset cover generation policy as an example, taking live scenes determined by a trained scene recognition model as an example, and taking personal information of a live scene and a host of a target live room as live information as an example, the live cover processing method provided by the application is further described with reference to fig. 12 and 13.

The training process of at least one preset cover creation model and the training process of the trained scene recognition model will be described with reference to fig. 12.

S1201, the server acquires a sample live broadcast picture of the sample live broadcast room and personal information of a host broadcast of the sample live broadcast room.

S1202, the server performs feature extraction on the sample live broadcast picture through the scene recognition model to be trained to obtain sample visual mode features corresponding to the sample live broadcast room, and performs feature extraction on personal information of a host broadcast in the sample live broadcast room to obtain sample text mode features corresponding to the sample live broadcast room.

S1203, the server performs fusion processing on the sample visual mode characteristics and the sample text mode characteristics through a scene recognition model to be trained to obtain sample multi-mode characteristics corresponding to the sample live broadcast room, and performs scene recognition on the sample live broadcast room according to the sample multi-mode characteristics to obtain a sample live broadcast scene of the sample live broadcast room.

S1204, the server determines a loss function value according to the sample live broadcast scene and the scene label corresponding to the sample live broadcast room, and determines whether the scene recognition model to be trained is converged according to the loss function value.

And S1205, if the scene recognition model to be trained is not converged, updating the model parameters of the scene recognition model to be trained according to the loss function value by the server, and returning to the step S1201.

S1206, if the scene recognition model to be trained converges, determining the scene recognition model to be trained as a trained scene recognition model.

S1207, the server performs feature extraction on the sample live broadcast picture through the trained scene recognition model to obtain a first sample visual mode feature corresponding to the sample live broadcast room, and performs feature extraction on personal information of a host broadcast in the sample live broadcast room to obtain a first sample text mode feature corresponding to the sample live broadcast room.

S1208, the server performs fusion processing on the first sample visual mode characteristics and the first sample text mode characteristics through the scene recognition model to be trained to obtain first sample multi-mode characteristics corresponding to the sample live broadcast room, and performs scene recognition on the sample live broadcast room according to the first sample multi-mode characteristics to obtain a first sample live broadcast scene of the sample live broadcast room.

S1209, the server selects a to-be-trained generating model corresponding to the first sample live scene from at least one to-be-trained generating model to obtain a target to-be-trained generating model.

S12010, the server generates a model to be trained through the target, and generates a sample live broadcast cover of the sample live broadcast room according to the sample live broadcast picture.

S12011, the server determines a target loss function value according to the sample live cover, and determines whether the target to-be-trained generation model is converged according to the target loss function value.

In the process of determining the objective loss function value according to the sample live cover by the server, whether the sample live cover comprises sample objective information (the sample objective information can be set according to actual conditions, for example, the sample objective information can be at least one of sample hand information and sample face information), if the sample live cover comprises the sample objective information, the objective loss function value is determined according to the sample objective information and the target information of the anchor in the sample live room (when the sample objective information is the sample face information, the objective loss function value is determined according to the sample objective information, the target information of the anchor in the sample live room and the target mark of the anchor in the sample live room), so that the generation stability of the live cover can be improved according to the sample objective information as auxiliary information for training at least one model to be trained, and the problems of hand deformity, low face similarity and the like in the live cover are solved.

S12011, the server determines a target loss function value according to the sample live cover, and determines whether the target to-be-trained generation model is converged according to the target loss function value.

S12012, if the target to-be-trained generation model is not converged,

S12012, if the target to-be-trained generation model is not converged, the server updates model parameters of the target to-be-trained generation model according to the target loss function value, and returns to execute the step S1207.

S12013, if the target to-be-trained generation model converges, the server determines the target to-be-trained generation model as a preset cover generation model.

At this time, the server may stop the training process when training all the to-be-trained generative models in the at least one to-be-trained generative model into the preset cover generative model.

In this embodiment, the live cover may be generated by the trained scene recognition model and at least one preset cover generation model, and at this time, the model structure in this embodiment may be as shown in fig. 14.

The following describes an application process of at least one preset cover creation model and an application process of a trained scene recognition model according to fig. 13.

S1301, the anchor terminal responds to the triggering operation of the recommendation control, acquires the live broadcast picture of the target live broadcast room and the personal information of the anchor according to the cover generation period of the target live broadcast room, and generates a cover generation task according to the live broadcast picture of the target live broadcast room and the personal information of the anchor.

S1302, the anchor side sends the cover generation task to a task queue for sorting.

S1303, the server acquires a target cover generation task from the task queue, performs feature extraction on live broadcast pictures in the target cover generation task through the trained scene recognition model to obtain visual mode features corresponding to a target live broadcast room, and performs feature extraction on personal information of a host broadcast in the target cover generation task to obtain text mode features corresponding to the target live broadcast room.

And 1304, the server performs fusion processing on the visual mode characteristics and the text mode characteristics through the trained scene recognition model to obtain multi-mode characteristics corresponding to the target live broadcasting room, and performs scene recognition on the target live broadcasting room according to the multi-mode characteristics to obtain a live broadcasting scene of the target live broadcasting room.

S1305, the server selects a preset cover generation model corresponding to the live scene from at least one preset cover generation model to obtain a target cover generation model.

S1306, the server generates a second candidate live cover with different styles according to the live picture in the target cover generation task through the target cover generation model.

The process of generating the second candidate live covers with different styles according to the live pictures in the target cover generation task through the target cover generation model may be:

Dividing at least one frame of live broadcast picture in the target cover generation task to obtain a picture area corresponding to each frame of live broadcast picture in the at least one frame of live broadcast picture;

screening out a target picture area from picture areas corresponding to each frame of live broadcast picture;

splicing the target picture areas to obtain a target live broadcast picture;

and generating a second candidate live broadcast cover with different styles according to the target live broadcast picture through the target cover generation model.

And S1307, the server determines the quality score of the second candidate live cover.

S1308, if the quality score meets the preset score condition, the server determines the second candidate live broadcast cover as the live broadcast cover of the target live broadcast room.

S1309, if the quality score does not meet the preset score condition, determining whether the face information exists in the target live broadcasting room.

And S13010, if no face information exists, updating the live broadcast picture in the live broadcast information by the server to obtain updated live broadcast information, and returning the updated live broadcast information to the execution step S903 as live broadcast information.

And S13011, if the face information exists, the server determines the face similarity according to the face information and the face information in the face area of the second candidate live broadcast cover.

S13012, if the face similarity does not meet the preset similarity, the server adjusts the face area in the second candidate live broadcast cover according to the face information to obtain the live broadcast cover of the target live broadcast room.

And S13013, if the face similarity meets the preset similarity, updating the live broadcast picture in the live broadcast information by the server to obtain updated live broadcast information, taking the updated live broadcast information as the live broadcast information, and returning to the step S1303.

The face information may be face key point information. When at least one generation model to be trained is trained according to sample face information, the server may perform steps S1307-S13013 through the target cover generation model.

It should be noted that, when there is a second candidate live cover corresponding to the quality score that does not meet the preset score condition, in the process of returning to the executing step S1302, only the cover of the style to which the second candidate live cover corresponding to the quality score that does not meet the preset score condition belongs may be regenerated, or a live cover of a different style may also be regenerated.

In this embodiment, the server acquires a target cover generation task from the task queue, performs image content understanding on live broadcast pictures in the target cover generation task and performs semantic understanding on personal information through the trained scene recognition model so as to determine a live broadcast scene of the target live broadcast room, then screens out the target cover generation model from at least one preset cover generation model according to the live broadcast scene, generates a second candidate live broadcast cover of the target live broadcast room according to the live broadcast pictures through the target cover generation model, finally performs quality evaluation on the second candidate live broadcast cover, and determines the second candidate live broadcast cover as the live broadcast cover if the quality evaluation passes, for example, as shown in fig. 15.

S13014, the server sends the live cover of the target live room to the anchor terminal.

S13015, the host side displays the live cover of the target live broadcasting room on the comment display area, and sends a confirmation instruction to the server in response to confirmation operation on the live cover.

S13016, the server acquires style characteristics of target account numbers of all audience terminals, and sends the live broadcast covers of the styles matched with the style characteristics to the audience terminals to serve as live broadcast covers of target live rooms for display.

For example, as shown in fig. 16, the anchor side obtains the live broadcast picture of the target live broadcast room and the personal information of the anchor, generates a cover generation task according to the live broadcast picture and the personal information, sends the cover generation task to the task queue for sorting, the server obtains the target cover generation task from the task queue, generates the live broadcast cover according to the target cover generation task, returns the live broadcast cover to the result queue, the anchor side obtains the live broadcast cover corresponding to the cover generation task from the result queue, displays the live broadcast cover on the comment display area, and sends the live broadcast cover to the audience side for display in response to the confirmation operation of the anchor on the live broadcast cover.

The interaction between the anchor side, the server and the viewer side in this embodiment may be as shown in fig. 17.

Optionally, after the audience end presents the live cover, the method can further comprise;

the method comprises the steps that a server obtains the number of audiences entering a target living room within a preset time length after a living cover is displayed;

the server adjusts the cover generation period according to the number of audiences to obtain an adjusted cover generation period;

the server sends the adjusted cover generation period to the anchor terminal;

the anchor terminal takes the adjusted cover generation period as a cover generation period and returns to the step of executing the cover generation period of the target living broadcast room.

The specific implementation manner and the corresponding beneficial effects in this embodiment may refer to the above live cover processing method embodiment specifically, and this embodiment is not described herein again.

In order to facilitate better implementation of the live cover processing method provided by the embodiment of the application, the embodiment of the application also provides a device based on the live cover processing method. The meaning of the nouns is the same as that in the live cover processing method, and specific implementation details can be referred to the description in the embodiment of the method.

For example, as shown in fig. 18, the live cover processing apparatus may include:

And the acquiring module 1801 is used for acquiring the live broadcast information of the target live broadcast room.

And the identification module 1802 is configured to identify a scene of the target live broadcast room according to the live broadcast information, so as to obtain a live broadcast scene of the target live broadcast room.

The selecting module 1803 is configured to select a preset cover generation policy corresponding to the live scene from at least one preset cover generation policy, so as to obtain a target cover generation policy.

The generating module 1804 is configured to generate, according to the live information, a live cover of the target live room by using the target cover generating policy.

Optionally, the live information includes a live screen of the target live room and text information associated with the target live room, and the identification module 1802 is specifically configured to perform:

extracting features of the live broadcast pictures to obtain visual mode features corresponding to the target live broadcast room;

extracting the characteristics of the text information to obtain the text modal characteristics corresponding to the target live broadcasting room;

fusion processing is carried out on the video modality features and the text modality features to obtain multi-modality features corresponding to the target live broadcasting room;

and carrying out scene recognition on the target live broadcasting room according to the multi-mode characteristics to obtain a live broadcasting scene of the target live broadcasting room.

Optionally, the generating module 1804 is specifically configured to perform:

Generating a first candidate live broadcast cover of the target live broadcast room according to live broadcast information through a target cover generation strategy;

acquiring face information of a target live broadcasting room;

and according to the face information, adjusting the face area in the first candidate live broadcast cover to obtain the live broadcast cover of the target live broadcast room.

Optionally, the generating module 1804 is specifically configured to perform:

generating a second candidate live broadcast cover of the target live broadcast room according to the live broadcast information through a target cover generation strategy;

determining the quality fraction corresponding to the second candidate live cover;

and if the quality score meets the preset score condition, determining the second candidate live broadcast cover as the live broadcast cover of the target live broadcast room.

Optionally, the generating module 1804 is further configured to perform:

if the quality score does not meet the preset score condition, updating the live broadcast information of the target live broadcast room to obtain updated live broadcast information;

and returning the updated live broadcast information as live broadcast information to execute the step of identifying the scene of the target live broadcast room according to the live broadcast information to obtain the live broadcast scene of the target live broadcast room.

Optionally, the generating module 1804 is specifically configured to perform:

performing target detection on the second candidate live broadcast cover to obtain a target detection result;

And determining the quality fraction corresponding to the second candidate live cover according to the target detection result.

Optionally, the live information includes at least one frame of live screen of the target live room, and the generating module 1804 is specifically configured to perform:

dividing at least one frame of live broadcast picture to obtain a picture area corresponding to each frame of live broadcast picture in at least one frame of live broadcast picture;

screening out a target picture area from picture areas corresponding to each frame of live broadcast picture;

splicing the target picture areas to obtain a target live broadcast picture;

and generating a live broadcast cover of the target live broadcast room according to the target live broadcast picture through a target cover generation strategy.

Optionally, the generating module 1804 is specifically configured to perform:

and generating the live broadcast covers of different styles of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

The live cover processing device further comprises:

a transmitting module, configured to perform:

obtaining style characteristics of target account numbers of all audience terminals;

and displaying the live broadcast cover of the style matched with the style characteristics as the live broadcast cover of the target live broadcast room in the audience terminal.

Optionally, the acquiring module 1801 is specifically configured to perform:

acquiring a cover generation period of a target live broadcasting room;

Acquiring live broadcast information of a target live broadcast room according to the cover generation period;

displaying a live cover at a viewer end;

acquiring the number of audiences entering a target living room within a preset time length after the live cover is displayed;

according to the number of audiences, adjusting the cover generation period to obtain an adjusted cover generation period;

and taking the adjusted cover generation period as a cover generation period, and returning to the step of executing the cover generation period of the acquisition target live broadcasting room.

Optionally, the target live room includes at least one, and the identification module 1802 is specifically configured to perform:

generating a cover generation task according to the live broadcast information;

sending the cover generating task to a task queue for sorting;

acquiring a target cover generation task from a task queue;

and carrying out scene recognition on the target live broadcasting room according to the target cover generation task to obtain a live broadcasting scene of the target live broadcasting room.

Optionally, the identification module 1802 is specifically configured to perform:

transmitting the cover generating task to a task queue;

acquiring characteristic information of a target living broadcast room corresponding to each cover generation task in a task queue;

and sequencing all cover generating tasks in the task queue according to the characteristic information.

In the specific implementation, each module may be implemented as an independent entity, or may be combined arbitrarily, and implemented as the same entity or a plurality of entities, and the specific implementation and the corresponding beneficial effects of each module may be referred to the foregoing method embodiments, which are not described herein again.

In order to facilitate better implementation of the live cover processing method provided by the embodiment of the application, the embodiment of the application also provides a device based on the live cover processing method. The meaning of the nouns is the same as that in the live cover processing method, and specific implementation details can refer to the description in the embodiment of the method.

For example, as shown in fig. 19, the live cover processing apparatus may include:

a first display module 1901 is configured to display a target live room, where the target live room includes a comment display area.

And the second display module 1902 is configured to display, on the comment display area, a live broadcast cover corresponding to the target live broadcast room, where the live broadcast cover is a cover determined based on a scene recognition result of the target live broadcast room.

And the operation response module 1903 is used for responding to the confirmation operation of the live cover and sending the live cover to the audience end for display.

Optionally, the comment display area further includes an association control, and the live cover processing apparatus further includes:

An association module for performing:

responding to triggering operation of the association control, and displaying a note list corresponding to the target anchor account in the target live broadcasting room, wherein the note list comprises at least one candidate note;

in response to a selection operation of a target note in the at least one candidate note, associating the target note with the live cover;

the operation response module 1903 is specifically configured to perform:

and responding to the confirmation operation of the live cover, sending the live cover to a spectator terminal for displaying, and displaying target notes associated with the live cover in a target live room in the spectator terminal.

Optionally, the association module is further configured to perform:

responding to the triggering operation of the associated control, and acquiring a scene identification result of the target live broadcasting room;

screening candidate notes matched with the scene recognition result from a note list corresponding to a target anchor account in a target live broadcasting room;

and displaying the candidate notes matched with the scene recognition result.

Optionally, the target live room further includes a switch control, and the live cover processing device further includes:

a starting module, configured to perform:

responding to the triggering operation of the switching control, and starting a cover switching function;

detecting whether a scene recognition result changes according to the cover switching function;

If the scene identification result changes, updating the live cover, and sending the updated live cover to the audience terminal for display.

In the specific implementation, each module may be implemented as an independent entity, or may be combined arbitrarily, and implemented as the same entity or a plurality of entities, and the specific implementation and the corresponding beneficial effects of each module may be referred to the foregoing method embodiments, which are not described herein again.

The embodiment of the application also provides an electronic device, which may be a server or a terminal, as shown in fig. 20, and shows a schematic structural diagram of the electronic device according to the embodiment of the application, specifically:

the electronic device may include one or more processing cores 'processors 2001, one or more computer-readable storage media's memory 2002, power supply 2003, and input unit 2004, among other components. It will be appreciated by those skilled in the art that the electronic device structure shown in fig. 20 is not limiting of the electronic device and may include more or fewer components than shown, or may combine certain components, or a different arrangement of components. Wherein:

The processor 2001 is a control center of the electronic device, connects various parts of the entire electronic device using various interfaces and lines, performs various functions of the electronic device and processes data by running or executing computer programs and/or modules stored in the memory 2002, and invoking data stored in the memory 2002. Optionally, processor 2001 may include one or more processing cores; preferably, the processor 2001 may integrate an application processor and a modem processor, wherein the application processor primarily handles operating systems, user interfaces, application programs, etc., and the modem processor primarily handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 2001.

The memory 2002 may be used to store computer programs and modules, and the processor 2001 may perform various functional applications and data processing by executing the computer programs and modules stored in the memory 2002. The memory 2002 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, a computer program (such as a sound playing function, an image playing function, etc.) required for at least one function, and the like; the storage data area may store data created according to the use of the electronic device, etc. In addition, memory 2002 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, memory 2002 may also include a memory controller to provide processor 2001 with access to memory 2002.

The electronic device further comprises a power supply 2003 for powering the various components, preferably the power supply 2003 is logically connected to the processor 2001 by a power management system whereby charging, discharging, power consumption management and the like are managed. The power supply 2003 may also include one or more of any components, such as a direct current or alternating current power supply, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, and the like.

The electronic device may also include an input unit 2004, which input unit 2004 may be used to receive input numeric or character information and to generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

Although not shown, the electronic device may further include a display unit or the like, which is not described herein. In particular, in this embodiment, the processor 2001 in the electronic device loads executable files corresponding to the processes of one or more computer programs into the memory 2002 according to the following instructions, and the processor 2001 executes the computer programs stored in the memory 2002, so as to implement various functions, for example:

Acquiring live broadcast information of a target live broadcast room;

according to the live broadcast information, carrying out scene identification on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room;

selecting a preset cover generation strategy corresponding to the live scene from at least one preset cover generation strategy to obtain a target cover generation strategy;

and generating a live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

The specific embodiments and the corresponding beneficial effects of the above operations can be referred to the above detailed description of the live cover processing method, and are not described herein.

It will be appreciated by those of ordinary skill in the art that all or part of the steps of the various methods of the above embodiments may be performed by a computer program, or by computer program control related hardware, which may be stored in a computer readable storage medium and loaded and executed by a processor.

To this end, embodiments of the present application provide a computer readable storage medium having stored therein a computer program that is capable of being loaded by a processor to perform steps in any of the live cover processing methods provided by embodiments of the present application. For example, the computer program may perform the steps of:

Acquiring live broadcast information of a target live broadcast room;

according to the live broadcast information, carrying out scene identification on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room;

selecting a preset cover generation strategy corresponding to the live scene from at least one preset cover generation strategy to obtain a target cover generation strategy;

and generating a live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

The specific embodiments and the corresponding beneficial effects of each of the above operations can be found in the foregoing embodiments, and are not described herein again.

Wherein the computer-readable storage medium may comprise: read Only Memory (ROM), random access Memory (RAM, random Access Memory), magnetic or optical disk, and the like.

Since the computer program stored in the computer readable storage medium may execute the steps in any live cover processing method provided in the embodiments of the present application, the beneficial effects that any live cover processing method provided in the embodiments of the present application may be achieved are detailed in the previous embodiments and will not be described herein.

Among other things, according to one aspect of the present application, a computer program product or computer program is provided that includes computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions so that the computer device performs the live cover processing method described above.

The foregoing has described in detail the method, apparatus, electronic device and computer storage medium for processing live covers provided in the embodiments of the present application, and specific examples have been applied to illustrate the principles and embodiments of the present application, where the foregoing description of the embodiments is only for helping to understand the method and core ideas of the present application; meanwhile, those skilled in the art will have variations in the specific embodiments and application scope in light of the ideas of the present application, and the present description should not be construed as limiting the present application in view of the above.

Claims (18)

1. A live cover processing method, comprising:

acquiring live broadcast information of a target live broadcast room;

according to the live broadcast information, carrying out scene recognition on the target live broadcast room to obtain a live broadcast scene of the target live broadcast room;

selecting a preset cover generation strategy corresponding to the live broadcast scene from at least one preset cover generation strategy to obtain a target cover generation strategy;

and generating a live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

2. The live cover processing method according to claim 1, wherein the live information includes live pictures of the target live room and text information associated with the target live room, and the scene recognition is performed on the target live room according to the live information to obtain live scenes of the target live room, including:

Extracting features of the live broadcast pictures to obtain visual mode features corresponding to the target live broadcast room;

extracting the characteristics of the text information to obtain text modal characteristics corresponding to the target live broadcasting room;

carrying out fusion processing on the visual mode characteristics and the text mode characteristics to obtain multi-mode characteristics corresponding to the target live broadcasting room;

and carrying out scene recognition on the target live broadcasting room according to the multi-mode characteristics to obtain a live broadcasting scene of the target live broadcasting room.

3. The live cover processing method according to claim 1, wherein the generating, by the target cover generating policy, the live cover of the target live room according to the live information includes:

generating a first candidate live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy;

acquiring face information of the target live broadcasting room;

and according to the face information, adjusting the face area in the first candidate live broadcast cover to obtain the live broadcast cover of the target live broadcast room.

4. The live cover processing method according to claim 1, wherein the generating, by the target cover generating policy, the live cover of the target live room according to the live information includes:

Generating a second candidate live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy;

determining the quality fraction corresponding to the second candidate live cover;

and if the quality score meets a preset score condition, determining the second candidate live broadcast cover as the live broadcast cover of the target live broadcast room.

5. The live cover processing method according to claim 4, further comprising, after the determining the quality score corresponding to the second candidate live cover:

if the quality score does not meet the preset score condition, updating the live broadcast information of the target live broadcast room to obtain updated live broadcast information;

and returning the updated live information as the live information to execute the step of identifying the scene of the target live broadcasting room according to the live information to obtain the live broadcasting scene of the target live broadcasting room.

6. The live cover processing method according to claim 4, wherein the determining the quality score corresponding to the second candidate live cover includes:

performing target detection on the second candidate live broadcast cover to obtain a target detection result;

And determining the quality fraction corresponding to the second candidate live cover according to the target detection result.

7. The live cover processing method according to claim 1, wherein the live information includes at least one frame of live pictures of the target live room, the generating, by the target cover generating policy, the live cover of the target live room according to the live information includes:

dividing the at least one frame of live broadcast picture to obtain a picture area corresponding to each frame of live broadcast picture in the at least one frame of live broadcast picture;

screening out a target picture area from picture areas corresponding to each frame of live broadcast picture;

splicing the target picture areas to obtain a target live broadcast picture;

and generating a live broadcast cover of the target live broadcast room according to the target live broadcast picture through the target cover generation strategy.

8. The live cover processing method according to claim 1, wherein the generating, by the target cover generating policy, the live cover of the target live room according to the live information includes:

generating different styles of live covers of the target live broadcasting room according to the live broadcasting information through the target cover generation strategy;

After the target cover generation strategy is adopted to generate the live cover of the target live room according to the live information, the method further comprises the following steps:

obtaining style characteristics of target account numbers of all audience terminals;

and displaying the live broadcast cover of the style matched with the style characteristics as the live broadcast cover of the target live broadcast room in the audience terminal.

9. The live cover processing method according to claim 1, wherein the acquiring live information of the target live room includes:

acquiring a cover generation period of the target live broadcasting room;

acquiring live broadcast information of a target live broadcast room according to the cover generation period;

after the target cover generation strategy is adopted to generate the live cover of the target live room according to the live information, the method further comprises the following steps:

displaying the live cover at the audience end;

acquiring the number of audiences entering the target living room within a preset time length after the living cover is displayed;

according to the number of the audiences, adjusting the cover generation period to obtain an adjusted cover generation period;

and taking the adjusted cover generation period as the cover generation period, and returning to the step of acquiring the cover generation period of the target living broadcast room.

10. The live cover processing method according to claim 1, wherein the target live room includes at least one, and the performing scene recognition on the target live room according to the live information to obtain a live scene of the target live room includes:

generating a cover generation task according to the live broadcast information;

sending the cover generating task to a task queue for sorting;

acquiring a target cover generation task from the task queue;

and generating a task according to the target cover, and identifying the scene of the target live broadcasting room to obtain the live broadcasting scene of the target live broadcasting room.

11. The live cover processing method of claim 10, wherein the sending the cover creation task to a task queue for ordering comprises:

transmitting the cover generating task to a task queue;

acquiring characteristic information of a target living broadcast room corresponding to each cover generation task in the task queue;

and sequencing all cover generating tasks in the task queue according to the characteristic information.

12. The live cover processing method is characterized by being applied to a host side and comprising the following steps of:

Displaying a target live broadcast room, wherein the target live broadcast room comprises a comment display area;

displaying a live broadcast cover corresponding to the target live broadcast room on the comment display area, wherein the live broadcast cover is determined based on a scene identification result of the target live broadcast room;

and responding to the confirmation operation of the live cover, and sending the live cover to a spectator terminal for display.

13. The live cover processing method of claim 12, wherein the comment display area further includes an association control, and after displaying the live cover corresponding to the target live room on the comment display area, further includes:

responding to the triggering operation of the association control, and displaying a note list corresponding to a target anchor account in the target live broadcasting room, wherein the note list comprises at least one candidate note;

in response to a selection operation of a target note in the at least one candidate note, associating the target note with the live cover;

transmitting the live cover to a viewer end for display in response to a confirmation operation of the live cover, wherein the live cover comprises the following steps:

and responding to the confirmation operation of the live cover, sending the live cover to a spectator terminal for displaying, and displaying the target notes associated with the live cover in the spectator terminal by the target live room.

14. The live cover processing method according to claim 13, wherein the displaying, in response to the triggering operation of the association control, a note list corresponding to a target anchor account in the target live room includes:

responding to the triggering operation of the association control, and acquiring a scene identification result of the target live broadcasting room;

screening candidate notes matched with the scene recognition result from a note list corresponding to a target anchor account in the target live broadcasting room;

and displaying candidate notes matched with the scene recognition result.

15. The live cover processing method according to claim 13, wherein the target live room further includes a switch control, and after the live cover corresponding to the target live room is displayed on the comment display area, the method further includes:

responding to the triggering operation of the switching control, and starting a cover switching function;

after the live cover is sent to the audience end for display in response to the confirmation operation of the live cover, the method further comprises the following steps:

detecting whether the scene recognition result changes according to the cover switching function;

if the scene identification result changes, updating the live cover, and sending the updated live cover to the audience terminal for display.

16. A live cover processing apparatus, comprising:

the acquisition module is used for acquiring live broadcast information of the target live broadcast room;

the identification module is used for carrying out scene identification on the target live broadcasting room according to the live broadcasting information to obtain a live broadcasting scene of the target live broadcasting room;

the selecting module is used for selecting a preset cover generation strategy corresponding to the live broadcast scene from at least one preset cover generation strategy to obtain a target cover generation strategy;

and the generation module is used for generating a live broadcast cover of the target live broadcast room according to the live broadcast information through the target cover generation strategy.

17. An electronic device comprising a processor and a memory, the memory storing a computer program, the processor configured to execute the computer program in the memory to perform the live cover processing method of any of claims 1 to 15.

18. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program adapted to be loaded by a processor to perform the live cover processing method of any of claims 1 to 15.