WO2019000789A1

WO2019000789A1 - Live video recommending method and device, and server

Info

Publication number: WO2019000789A1
Application number: PCT/CN2017/110935
Authority: WO
Inventors: 刘宇超; 张文明; 陈少杰
Original assignee: 武汉斗鱼网络科技有限公司
Priority date: 2017-06-26
Filing date: 2017-11-14
Publication date: 2019-01-03
Also published as: CN107404656B; CN107404656A

Abstract

The present application relates to the field of Internet recommendation, and provides a live video recommending method and device, and a server. The live video recommending method is applicable to a server in communicative connection with a client. The method comprises: acquiring image data corresponding to multiple broadcast live videos; separately obtaining feature data corresponding to each piece of image data according to the each piece of image data; determining, according to the feature data, whether the similarity between any two of the live videos satisfies a preset standard; searching, according to a received watching request sent by the client, for a live video of which the similarity with a target live video corresponding to the watching request satisfies the preset standard; and recommending the found live video to the client. Timely and effective similarity analysis on broadcast live videos is realized. Moreover, the user recommendation effectiveness can be guaranteed, and the user experience is improved. The gap of real-time live video recommendation in the field of live broadcast is filled.

Description

Live video recommendation method, device and server

Cross-reference to related applications

The present application claims priority to Chinese Patent Application No. 2017104935651, entitled "Real-Video Recommendation Method, Apparatus, and Server", filed on Jun. 26, 2017, the entire contents of which is incorporated herein by reference. .

Technical field

The present application relates to the field of Internet recommendation, and in particular, to a live video recommendation method, apparatus, and server.

Background technique

In the era of information explosion, a variety of information is flooding the Internet. The amount of data on the Internet is so great that users can't screen out the content of interest from all the information on the Internet, and the Internet recommendation comes into being.

The Internet recommends a variety of information, such as articles, videos, and software. At present, the recommended techniques are based on the related text description information and comment information corresponding to the information selected by the user, and then select other related content related to the user's selected information. The recommended technology at this stage is good and mature, but such a recommendation technology faces the emerging industry, but it seems to be powerless. The content of the live broadcast industry is generated in real time, and the text description is free and changeable. Before the live video ends, it is impossible to select other related live video through the current recommendation algorithm.

Summary of the invention

One of the purposes of the present application is to provide a live video recommendation method for improving the above problem.

Another object of the present application is to provide a live video recommendation device for improving the above problem.

It is still another object of the present application to provide a server for improving the above problems.

In order to achieve the above object, the technical solution adopted by the embodiment of the present application is as follows:

The embodiment of the present application provides a live video recommendation method. The method is applied to a server that is in communication with a client, and the method includes: acquiring image data corresponding to a plurality of live video being broadcasted; obtaining feature data corresponding to each of the image data according to each of the image data Determining, according to the feature data, whether the similarity between any two of the live video meets a preset criterion; and searching between the target live video corresponding to the viewing request according to the received viewing request sent by the client The live video that meets the preset criteria is similar; the live video that is found is recommended to the client.

The embodiment of the present application further provides a live video recommendation device. The device is applied to a service that is connected to a client. The device includes: an obtaining module, an extracting module, an evaluating module, a searching module, and a pushing module; wherein the acquiring module is configured to acquire image data corresponding to a plurality of live video being broadcasted; and the extracting module is configured to be respectively configured according to Each of the image data obtains feature data corresponding to each of the image data; the evaluation module is configured to determine, according to the feature data, whether the similarity between any two of the live video meets a preset criterion; And configuring, according to the received viewing request sent by the client, the live video of the similarity between the target live video corresponding to the viewing request that meets the preset criterion; and the pushing module configured to find the The live video is recommended to the client.

The embodiment of the present application further provides a server. The server includes: a memory; a processor; and a live video recommendation device, the live video recommendation device being installed in the memory and including one or more software function modules executed by the processor, the device comprising An obtaining module, an extracting module, an evaluating module, a searching module, and a pushing module; wherein the acquiring module is configured to acquire image data corresponding to a plurality of live video being broadcasted; and the extracting module is configured to obtain, according to each of the image data respectively Each of the image data corresponding to the feature data; the evaluation module is configured to determine, according to the feature data, whether the similarity between any two of the live video meets a preset criterion; the search module is configured to be based on the received client The viewing request sent by the terminal, the search for the live video of the preset standard that matches the target live video corresponding to the viewing request, and the pushing module configured to recommend the found live video to The client.

The embodiment of the present application further provides a computer readable storage medium, on which a computer program is stored, and when the program is executed by the processor, the steps of the foregoing method are implemented.

Compared with the prior art, the present application provides a live video recommendation method. The method determines whether the similarity between any two live video matches the preset criterion according to the feature data corresponding to the image data of the plurality of live video, so that the evaluation result does not depend on the related text description information and the comment information, and the implementation is Similarity analysis between live broadcasts played in real time. Therefore, the user can promptly and accurately push other live video that meets the preset standard with the real-time live video of the viewing, and improve the user experience.

The above described objects, features, and advantages of the present invention will become more apparent from the following description.

DRAWINGS

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings to be used in the embodiments will be briefly described below. It should be understood that the following drawings show only certain embodiments of the present application, and therefore It should be seen as a limitation on the scope, and those skilled in the art can obtain other related drawings according to these drawings without any creative work.

FIG. 1 shows a schematic diagram of an application environment of the present application.

2 is a block schematic diagram of a server according to a preferred embodiment of the present application.

FIG. 3 is a flow chart showing the steps of a live video recommendation method according to a preferred embodiment of the present application.

4 is a flow chart of the sub-steps of step S101 in FIG. 1.

FIG. 5 is a flow chart of the sub-steps of step S103 in FIG. 1.

FIG. 6 is a schematic diagram of a live video recommendation apparatus according to a preferred embodiment of the present application.

FIG. 7 is a schematic diagram of functional sub-modules of the evaluation module of FIG. 6.

Icon: 100-server; 111-memory; 112-processor; 113-communication unit; 200-user terminal; 300-client; 400-live video recommendation device; 401-acquisition module; 402-extraction module; Module; 4031 - calculation sub-module; 4032 - evaluation sub-module; 4033 - determination sub-module; 404 - correspondence establishment module; 405 - search module; 406 - push module.

Detailed ways

The technical solutions in the embodiments of the present application are clearly and completely described in the following with reference to the drawings in the embodiments of the present application. It is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The components of the embodiments of the present application, which are generally described and illustrated in the figures herein, may be arranged and designed in various different configurations. The detailed description of the embodiments of the present application, which is set forth in the claims All other embodiments obtained by a person skilled in the art based on the embodiments of the present application without creative efforts are within the scope of the present application.

It should be noted that similar reference numerals and letters indicate similar items in the following figures. Therefore, once an item is defined in one figure, it is not necessary to further define and explain it in the subsequent figures. Also, in the description of the present application, the terms "first", "second", and the like are used merely to distinguish a description, and are not to be construed as indicating or implying a relative importance.

The following embodiments of the present application can be applied to the environment as shown in FIG. 1 without any particular explanation. As shown in FIG. 1, the server 100 is communicably connected to the user terminal 200 through a wired or wireless network.

The live video recommendation method and apparatus provided by the embodiments of the present application can be applied to the server 100 of the application environment shown in FIG. 1. As shown in FIG. 1, the server 100 and the user terminal 200 are located in a network, and the network includes a wireless network or a wired network. Through the network, the user terminal 200 performs data interaction with the server 100. The server 100 may be a server 100 or a plurality of servers 100 that perform different tasks.

In the embodiment of the present application, the user terminal 200 is preferably a mobile terminal device, and may include, for example, a smart phone, a tablet computer, an e-book reader, a laptop portable computer, a vehicle-mounted computer, a wearable mobile terminal, and the like. A client 300 (for example, a betta client) is installed in the user terminal 200, and corresponds to the server 100. In this embodiment, the client 300 is communicably connected to the server 100 to provide services for the user.

As shown in FIG. 2, it is a block schematic diagram of the server 100. The server 100 includes a live video recommendation device 400, memory 111, processor 112, communication unit 113.

The components of the memory 111, the processor 112, and the communication unit 113 are electrically connected directly or indirectly to each other to implement data transmission or interaction. For example, the components can be electrically connected to one another via one or more communication buses or signal lines. The live video recommendation device 400 includes at least one software function module that can be stored in the memory 111 or solidified in an operating system (OS) of the server 100 in the form of software or firmware. The processor 112 is configured to execute an executable module stored in the memory 111, such as a software function module, a computer program, and the like included in the live video recommendation device 400.

The memory 111 can be, but not limited to, a random access memory (RAM), a read only memory (ROM), and a programmable read-only memory (PROM). Erasable Programmable Read-Only Memory (EPROM), Electric Erasable Programmable Read-Only Memory (EEPROM), and the like. The memory 111 is configured to store programs and voice data, and the processor 112 executes the programs after receiving the execution instructions. The communication unit 113 is configured to establish a communication connection between the server 100 and the user terminal 200 through the network, and is configured to transmit and receive data through the network.

It should be understood that the structure shown in FIG. 2 is only a schematic structural diagram of the server 100, and the server 100 may further include more or less components than those shown in FIG. 1, or have a configuration different from that shown in FIG. . The components shown in Figure 2 can be implemented in hardware, software, or a combination thereof.

First embodiment

Please refer to FIG. 3 , which is a flowchart of a live video recommendation method applied to the server 100 shown in FIG. 2 according to a preferred embodiment of the present application. The live video recommendation method includes the following steps:

Step S101: Acquire image data corresponding to a plurality of live broadcast videos being broadcasted.

In this embodiment, the live video that is being broadcasted may be the live video received by the server 100. Optionally, the server 100 may receive live video uploaded by all the live broadcast clients in real time, and obtain current current image data from the received live video. It should be noted that the client 300 is divided into a live client and a viewing client according to different usage states. The live client is configured to capture the live video upload server 100 to share the live video client 300, and the viewing client is configured as a slave server. 100 acquires the shared live video for the client to view. The viewing client obtains a live video of the live client upload server 100 by sending a request to the server 100. As shown in FIG. 4, step S101 further includes the following sub-steps:

Sub-step S1011: Collect one frame of live broadcast image data from each of the plurality of live broadcast videos being broadcasted according to a preset time period interval.

In this embodiment, the live video that is being broadcasted in real time is uploaded by the live client to the server 100 in real time, and the server 100 collects the latest received video according to the preset time interval. All of the video streams of the live video stream currently have one frame of picture data. For example, the server 100 acquires the previous one-frame picture data of the newly received video stream of the live video at intervals of 5 seconds. This process is ongoing and ensures real-time performance. It should be noted that the plurality of live broadcast videos that are being broadcasted are all live broadcast videos that are uploaded in real time for all live broadcast clients that are connected to the server 100.

Sub-step S1012: generating the image data according to the live broadcast screen data, the live broadcast information data corresponding to the live broadcast screen data, and the anchor information data.

In this embodiment, each live video has corresponding live broadcast information data and anchor information data, and the amount of data between the live broadcast information data and the anchor information data is small, and the information between the live broadcast room and the anchor information can be quickly Find the corresponding live video. The live broadcast information data and the anchor information data corresponding to the live video are used as attribute data of live video data corresponding to the live video to generate corresponding image data.

It can be understood that, in other embodiments of the present application, the image data may also be generated according to the live video data and the attribute data of the live video data corresponding to the live video, where the live video is corresponding to the live video. The attribute data of the live broadcast video data may be data such as live broadcast information data, anchor information data, and the like, and a combination of the foregoing data.

Step S102, obtaining feature data corresponding to each of the image data according to each of the image data.

In the embodiment, the trained deep learning model is preset in the server 100. The server 100 performs incremental training on the deep learning model by collecting live broadcast picture data of a large number of received live video as a sample set. Optionally, the training process adopts a migration learning method, and continuously uses the sample set to train the last of the deep learning model. One layer (ie, the classifier layer), and then continuously improve the parameters of other layers until the accuracy of the classifier layer reaches about 95%. In order to complete the business customization of the deep learning model. For example, the deep learning model may be to download the trained inception v3 model from the GoogleTensorflow official website, set the node of the classifier layer of the inception v3 model to 10000 and then use the sample set to train the classifier layer of the invent v3 model to obtain a live broadcast. Video-customized inception v3 model. The last layer (ie, the classifier layer) of the deep learning model that completes the customization is removed, and the penultimate layer (ie, the feature extraction layer) is used as the output layer. It should be noted that the deep learning includes multiple levels, and the output data of the previous level is the input data of the next level, and the maximum description of the input data is retained between the levels, and the classifier layer is the last one of the deep learning model. Layer, in the prior art, the output result of the classification layer is usually the final output result, and the image data is identified. In the present application, the second last layer (ie, the feature extraction layer) is used as the final output layer, and feature extraction is performed. The output of the layer is a multi-dimensional feature vector extracted from the image data. The feature vector can fully reflect the visual properties of the image, and is convenient for calculation and comparison. Optionally, the feature extraction process is performed on the live picture data of each of the image data obtained by each acquisition by using the deep learning model. The feature vector output from the feature extraction layer of the deep learning model is used as the feature data.

Step S103, determining, according to the feature data, whether the similarity between any two of the live video meets a preset standard.

In this embodiment, in the feature data corresponding to the image data collected in the same time, whether the similarity between the two live video corresponding to the feature data meets a preset criterion is evaluated according to any two feature data. When the similarity of the two live video meets the preset criteria, the two live videos are similar. As shown in FIG. 5, step S103 includes the following sub-steps:

Sub-step S1031, calculating an Euclidean distance between the feature vectors of the image data corresponding to any two live broadcast videos being broadcast.

In this embodiment, the feature data includes a feature vector, and the server 100 respectively calculates each feature vector and other features in the feature vector corresponding to the image data collected according to the feature vector extracted in each acquired image data. Euclidean distance between vectors. It should be noted that the Euclidean distance is a commonly used distance definition, which refers to the true distance between two points in the m-dimensional space, or the natural length of the vector (ie, the distance from the point to the origin). The Euclidean distance in two-dimensional and three-dimensional space is the actual distance between two points. Calculating the Euclidean distance calculation is relatively small and fast.

Sub-step S1032, estimating the similarity of the corresponding two live video according to the Euclidean distance between the two feature vectors.

In the present embodiment, if the two image data are more similar, the Euclidean distances of the feature vectors respectively characterizing the two image data are smaller. The smaller the Euclidean distance, the higher the similarity of the corresponding live video.

Sub-step S1033, determining whether the preset criterion is met according to the similarity of the two live video.

In this embodiment, optionally, the Euclidean distance between the feature vectors of the image data corresponding to the two live video is used to determine whether the similarity between the two live video meets a preset criterion. Optionally, when the Euclidean distance is less than a preset threshold, determining that the similarity between the live video meets the preset criterion.

Step S104: Correspond to the inter-live broadcast information data and the anchor information data in the image data corresponding to the two live video that meet the preset criterion.

In this embodiment, the live broadcast information data and the anchor information corresponding to the two live video that meet the preset standard are associated, so that one live video can be quickly found to find another live video, that is, the similarity is satisfied. The correspondence between the two live video of the preset standard. Optionally, the correspondence may be saved until both live video ends, or may be saved until at least one live video ends.

In step S105, according to the received viewing request sent by the client 300, the live video in which the similarity between the target live video corresponding to the viewing request meets the preset criterion is found.

In this embodiment, when the viewing client requests the server 100 to obtain the target live video corresponding to the request, the server 100 sends the target live video to the viewing client for the customer to view, and also according to the target live video according to the preset time interval. Whether the information data of the live broadcast room and the data of the anchor information are inquired whether the similarity meets the preset standard When the video is broadcast, the process proceeds to step S106. Optionally, the live video that is currently being broadcasted is searched for a live video that meets the preset standard by the similarity of the target live video.

Step S106, recommending the found live video to the client 300.

In this embodiment, the obtained live broadcast information data and the anchor information data corresponding to the live broadcast video are obtained to generate a recommendation list, and the recommendation list is sent to the viewing client to use the user selection of the viewing client. Optionally, before the end of the live broadcast of the target requested by the viewing client or before the viewing client exits the viewing target video, the process returns to step S105 after the end of step S106, and when it is found that is different from the recommended live video, After the live video that meets the preset standard is similar to the target live video, the process proceeds to step S106 again.

Second embodiment

Referring to FIG. 6 , a preferred embodiment of the present application provides a live video recommendation device 400 . The live video recommendation device 400 includes an acquisition module 401 , an extraction module 402 , an evaluation module 403 , a search module 405 , a push module 406 , and a correspondence relationship establishing module 404 .

The obtaining module 401 is configured to acquire image data corresponding to a plurality of live broadcast videos being broadcasted.

In the embodiment of the present application, the step S101, the sub-step S1011, and the sub-step S1012 are all performed by the obtaining module 401.

Further, the obtaining module 401 is further configured to separately collect one frame of live broadcast screen data from the plurality of live broadcast videos that are being broadcasted according to the preset time period interval; and corresponding to the live broadcast screen data according to the live broadcast screen data. The live room information data and the anchor information data generate the image data.

The extraction module 402 is configured to obtain feature data corresponding to each of the image data according to each of the image data.

In the embodiment of the present application, the step S102 may be performed by the extraction module 402. Optionally, the extraction module 402 is configured to perform feature extraction processing on each of the image data by using the deep learning model, and use the feature vector output by the feature extraction layer of the deep learning model as the feature data. It should be noted that the trained deep learning model is preset in the server 100. The server 100 performs incremental training on the deep learning model by collecting live broadcast picture data of a large number of received live video as a sample set. Optionally, the training process adopts a migration learning method, and continuously uses the sample set to train the last of the deep learning model. One layer (ie, the classifier layer), and then continuously improve the parameters of other layers until the accuracy of the classifier layer reaches about 95%. In order to complete the business customization of the deep learning model. For example, the deep learning model may be to download the trained inception v3 model from the GoogleTensorflow official website, set the node of the classifier layer of the inception v3 model to 10000 and then use the sample set to train the classifier layer of the invent v3 model to obtain a live broadcast. Video-customized inception v3 model. The last layer (ie, the classifier layer) of the deep learning model that completes the customization is removed, and the penultimate layer (ie, the feature extraction layer) is used as the output layer. It should be noted that the deep learning includes multiple levels, and the output data of the previous level is the input data of the next level, and is retained between the levels. The maximum description of the input data, the classifier layer is the last layer of the deep learning model. In the prior art, the output result of the classification layer is generally used as the final output result, and the image data has been identified, and in the present application, The penultimate layer (ie, feature extraction layer) is used as the final output layer. The output of the feature extraction layer is a multi-dimensional feature vector extracted from the image data. The feature vector can fully reflect the visual properties of the image, and is convenient for calculation and comparison. . Optionally, the feature extraction process is performed on the live picture data of each of the image data obtained by each acquisition by using the deep learning model. The feature vector output from the feature extraction layer of the deep learning model is used as the feature data.

The evaluation module 403 is configured to determine, according to the feature data, whether the similarity between any two of the live video meets a preset criterion.

In the embodiment of the present application, the step S103 may be performed by the evaluation module 403. As shown in FIG. 7, the evaluation module 403 may further include the following sub-modules:

The calculation sub-module 4031 is configured to calculate an Euclidean distance between the feature vectors of the image data corresponding to any two live broadcast videos being broadcast.

In the embodiment of the present application, the sub-step S1031 may be performed by the calculation sub-module 4031.

The evaluation sub-module 4032 is configured to evaluate the similarity of the corresponding two live video according to the Euclidean distance between the two feature vectors.

In the embodiment of the present application, the sub-step S1032 may be performed by the evaluation sub-module 4032.

The determining sub-module 4033 is configured to determine, according to the similarity between the two live video, whether the preset criterion is met, and when the Euclidean distance is less than a preset threshold, determining that the similarity between the live video meets Presuppose the standard.

In the embodiment of the present application, the sub-step S1033 may be performed by the determining sub-module 4033.

The correspondence relationship establishing module 404 is configured to correspond to the inter-live channel information data and the anchor information data in the image data corresponding to the two live video that meet the preset criterion.

In the embodiment of the present application, the step S104 may be performed by the correspondence establishing module 404.

The searching module 405 is configured to search, according to the received viewing request sent by the client 300, the live video with the similarity between the target live video corresponding to the viewing request that meets the preset criterion.

Further, the searching module 405 is further configured to search for a live video that meets the preset standard by the live video that is currently being broadcasted.

In the embodiment of the present application, the step S105 may be performed by the corresponding search module 405.

The pushing module 406 is configured to recommend the found live video to the client 300. In the embodiment of the present application, the step S106 may be performed by the corresponding pushing module 406.

Further, the push module 406 is configured to obtain the live broadcast information data and the anchor information data corresponding to the found live video to generate a recommendation list, and send the recommendation list to the viewing client.

In summary, the present application provides a live video recommendation method, apparatus, and server. The method is applied to The server is connected to the client, and the method includes: acquiring image data corresponding to the live video that is being broadcasted; obtaining feature data corresponding to each of the image data according to each of the image data; and according to the feature data Determining whether the similarity between any two of the live video meets a preset criterion; according to the received viewing request sent by the client, finding a similarity between the target live video corresponding to the viewing request satisfies the pre- Set the standard live video; recommend the found live video to the client. The similarity analysis between the live broadcast videos being broadcasted is realized, the calculation amount is small, and the analysis speed is fast. And can guarantee the effectiveness of the recommendation to the user and improve the user experience. Filled in the real-time recommendation of live broadcast live broadcast vacancies.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may also be implemented in other manners. The apparatus embodiments described above are merely illustrative, for example, the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and functionality of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present application. operating. In this regard, each block of the flowchart or block diagram can represent a module, a program segment, or a portion of code that includes one or more configurations configured to perform the specified logical functions. Executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the blocks may also occur in a different order than those illustrated in the drawings. For example, two consecutive blocks may be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams and/or flowcharts, and combinations of blocks in the block diagrams and/or flowcharts, can be implemented in a dedicated hardware-based system that performs the specified function or action. Or it can be implemented by a combination of dedicated hardware and computer instructions.

In addition, each functional module in each embodiment of the present application may be integrated to form a separate part, or each module may exist separately, or two or more modules may be integrated to form a separate part.

The functions, if implemented in the form of software functional modules and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application, which is essential or contributes to the prior art, or a part of the technical solution, may be embodied in the form of a software product, which is stored in a storage medium, including The instructions are used to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .

It should be noted that, in this context, relational terms such as first and second are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply these entities or operations. There is any such actual relationship or order between them. Furthermore, the term "comprises" or "comprises" or "comprises" or any other variations thereof is intended to encompass a non-exclusive inclusion, such that a process, method, article, or device that comprises a plurality of elements includes not only those elements but also Other elements, or are included in the process, method, article or equipment Inherent elements. An element that is defined by the phrase "comprising a ..." does not exclude the presence of additional equivalent elements in the process, method, item, or device that comprises the element.

The above description is only the preferred embodiment of the present application, and is not intended to limit the present application, and various changes and modifications may be made to the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application are intended to be included within the scope of the present application. It should be noted that similar reference numerals and letters indicate similar items in the following figures. Therefore, once an item is defined in one figure, it is not necessary to further define and explain it in the subsequent figures.

The foregoing is only a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in the present application. It should be covered by the scope of protection of this application. Therefore, the scope of protection of the present application should be determined by the scope of the claims.

Industrial applicability

The live video recommendation method, device and server provided by the application can realize the similarity analysis between the live video broadcasted in real time without relying on the related text description information and the comment information, so that the user can timely and accurately Push and watch real-time live video similar to other live video that meets preset standards to improve user experience.

Claims

A live video recommendation method, wherein the method is applied to a server that is in communication with a client, the method comprising:

Obtaining image data corresponding to multiple live broadcast videos being broadcasted;

Obtaining feature data corresponding to each of the image data according to each of the image data;

Determining, according to the feature data, whether the similarity between any two of the live video meets a preset criterion;

And searching, according to the received viewing request sent by the client, the live video of the similarity between the target live video corresponding to the viewing request that meets the preset criterion;

The found live video is recommended to the client.
The method of claim 1, wherein the server comprises a preset depth learning model, the feature data comprises a feature vector, and the step of obtaining feature data corresponding to each of the image data comprises:

Feature extraction processing is performed on each of the image data by using the depth learning model, and the feature vector output by the feature extraction layer of the deep learning model is used as the feature data.
The method according to claim 2, wherein the step of determining, according to the feature data, whether the similarity between any two of the live video meets a preset criterion comprises:

Calculating an Euclidean distance between the feature vectors of the image data corresponding to any two live broadcast videos that are being broadcasted;

Evaluating the similarity of the corresponding two live video according to the Euclidean distance between the two feature vectors;

Determining whether the preset criterion is met according to the similarity of the two live video. When the Euclidean distance is less than a preset threshold, determining that the similarity between the live video meets the preset criterion.
The method according to claim 3, wherein the calculating the Euclidean distance between the feature vectors of the image data corresponding to any two live broadcast videos that are being broadcasted comprises:

And calculating, according to the feature vector extracted in the image data collected each time, an Euclidean distance between each feature vector corresponding to the image data corresponding to the acquired image data and other feature vectors.
The method according to claim 1, wherein the step of determining, according to the feature data, whether the similarity between any two of the live video meets a preset criterion comprises:

And determining, according to any two feature data, whether the similarity of the two live video corresponding to the feature data meets the preset criterion according to any two feature data collected by the same image data. ;

When the similarity of the two live video meets the preset criterion, it is determined that the two live videos are similar.
The method of claim 1, wherein the step of acquiring image data corresponding to a plurality of live broadcast videos being broadcasted comprises:

Collecting one frame of live broadcast data from each of the live broadcast videos being broadcasted according to a preset time period interval;

And generating the image data according to the live broadcast screen data, the live broadcast information data corresponding to the live broadcast screen data, and the anchor information data.
The method according to claim 6, wherein after determining whether the similarity between any two of the live video meets a preset criterion, the method further includes:

Corresponding to the inter-live broadcast information data and the anchor information data in the image data corresponding to the two live video that meet the preset standard.
The method according to claim 7, wherein the step of corresponding to the inter-live broadcast information data and the anchor information data in the image data corresponding to the two live video with the similarity satisfying the preset criterion comprises:

Establishing a correspondence between two live video whose similarity meets the preset criterion;

And saving the correspondence until the two live video ends, or saving the correspondence until at least one of the live video ends.
The method of claim 7 wherein:

The step of searching for the live video with the similarity between the target live video corresponding to the viewing request that meets the preset criteria includes:

Searching for a live video that meets the preset standard in a similarity to the target live video from the live video currently being broadcasted;

The step of recommending the found live video to the client includes:

Obtaining the live broadcast information data and the anchor information data corresponding to the found live video to generate a recommendation list.
The method of claim 1, wherein the step of acquiring image data corresponding to a plurality of live broadcast videos being broadcasted comprises:

Collecting one frame of live broadcast data from each of the live broadcast videos being broadcasted according to a preset time period interval;

And generating the image data according to the live broadcast screen data and attribute data of live broadcast screen data corresponding to the live video.
The method of claim 6 or 10, wherein the step of collecting one frame of live broadcast data from each of the plurality of live broadcast videos being broadcasted according to a preset time period interval comprises:

Collecting, according to the preset time period interval, the current frame data of the video stream of all the currently received live video.
The method of claim 1 wherein the method further comprises:

After the similarity between the target live video corresponding to the viewing request is met, the live video step of the preset criterion is met, and when the search is different from the recommended live video, and the target is broadcasted After the video similarity meets the preset standard live video, the step of recommending the found live video to the client is performed.
A live video recommendation device, wherein the device is applied to a server that is in communication with a client, the device comprising:

Obtaining a module, configured to acquire image data corresponding to multiple live broadcast videos being broadcasted;

An extraction module, configured to obtain feature data corresponding to each of the image data according to each of the image data;

An evaluation module, configured to determine, according to the feature data, whether a similarity between any two of the live video meets a preset criterion;

a search module, configured to: according to the received viewing request sent by the client, searching for the live video that meets the preset standard by the similarity between the target live video corresponding to the viewing request;

The pushing module is configured to recommend the found live video to the client.
The apparatus according to claim 13, wherein the server comprises a preset depth learning model, the feature data comprises a feature vector, and the extraction module is further configured to:

Feature extraction processing is performed on each of the image data by using the depth learning model, and the feature vector output by the feature extraction layer of the deep learning model is used as the feature data.
The device of claim 14, wherein the evaluation module comprises:

a calculation submodule configured to calculate an Euclidean distance between the feature vectors of the image data corresponding to any two live broadcast videos that are being broadcasted;

An evaluation submodule configured to evaluate a similarity of the corresponding two live video according to the Euclidean distance between the two feature vectors;

Determining a sub-module, configured to determine, according to the similarity between the two live video, whether the preset criterion is met, and when the Euclidean distance is less than a preset threshold, determining that the similarity between the live video meets the Preset standard.
The apparatus of claim 13, wherein the acquiring module is further configured to separately collect one frame of live video data from the plurality of live broadcast videos being broadcasted according to a preset time period interval; and The image data is generated according to the live broadcast screen data, the live broadcast information data corresponding to the live broadcast screen data, and the anchor information data.
The device of claim 16 wherein said device further comprises:

The correspondence relationship establishing module is configured to correspond to the live broadcast information data and the anchor information data in the image data corresponding to the two live video that meet the preset criterion.
The device of claim 16, wherein the searching module is further configured to: search for a live video that is similar to the target live video and meet the preset standard from a live video currently being broadcasted;

The push module is configured to obtain the live broadcast information data and the anchor information data corresponding to the live broadcast video, to generate a recommendation list, and send the recommendation list to the viewing client.
A server, wherein the server comprises:

Memory

Processor;

The live video recommendation device is installed in the memory and includes one or more software function modules executed by the processor, and the live video recommendation device includes:

Obtaining a module, configured to acquire image data corresponding to multiple live broadcast videos being broadcasted;

An extraction module, configured to obtain feature data corresponding to each of the image data according to each of the image data;

An evaluation module, configured to determine, according to the feature data, whether a similarity between any two of the live video meets a preset criterion;

a search module, configured to: according to the received viewing request sent by the client, searching for the live video that meets the preset standard by the similarity between the target live video corresponding to the viewing request;

The pushing module is configured to recommend the found live video to the client.
A computer readable storage medium having stored thereon a computer program, wherein the program, when executed by a processor, implements the steps of the method of any of claims 1-12.