CN113824950A - Service processing method, device, equipment and medium - Google Patents

Service processing method, device, equipment and medium Download PDF

Info

Publication number
CN113824950A
CN113824950A CN202110780207.5A CN202110780207A CN113824950A CN 113824950 A CN113824950 A CN 113824950A CN 202110780207 A CN202110780207 A CN 202110780207A CN 113824950 A CN113824950 A CN 113824950A
Authority
CN
China
Prior art keywords
data stream
live broadcast
decision
result
service
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110780207.5A
Other languages
Chinese (zh)
Inventor
蒋政胜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tencent Technology Shenzhen Co Ltd
Original Assignee
Tencent Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tencent Technology Shenzhen Co Ltd filed Critical Tencent Technology Shenzhen Co Ltd
Priority to CN202110780207.5A priority Critical patent/CN113824950A/en
Publication of CN113824950A publication Critical patent/CN113824950A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N17/00Diagnosis, testing or measuring for television systems or their details
    • H04N17/004Diagnosis, testing or measuring for television systems or their details for digital television systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)

Abstract

The embodiment of the application provides a service processing method, a device, equipment and a medium, wherein the method comprises the following steps: acquiring a target live broadcast data stream in a live broadcast service; calling a service model to identify the target live broadcast data stream to obtain an identification result of the target live broadcast data stream; performing decision processing on the label display of the target live broadcast data stream according to the recognition result to obtain a decision result; and testing the business model based on the decision result. By adopting the embodiment of the application, the test efficiency of the service model can be improved.

Description

Service processing method, device, equipment and medium
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method, an apparatus, a device, and a medium for processing a service.
Background
In many service scenarios (e.g., live service scenarios, music playing scenarios, etc.), there is a need to identify multimedia data (e.g., video, image, audio, etc.) and label the multimedia data according to the identification result. For example: in a live broadcast service scene, a live broadcast video stream can be identified, and a dance tag, a singing tag, a motion tag and the like can be marked on live broadcast data according to an identification result.
The identification of the multimedia data is realized by depending on a network model (or called a service model), and the identification performance of the network model has a large influence on the identification result. In the prior art, the identification performance of a test network model is measured by manually broadcasting a test material and observing whether a label is accurately displayed on a terminal; this makes the test link to the network model longer, and the test efficiency is lower. How to improve the testing efficiency of the network model becomes a hot topic of research.
Disclosure of Invention
The embodiment of the application provides a service processing method, a service processing device, service processing equipment and a service processing medium, and the service processing method, the service processing device, the service processing equipment and the service processing medium can improve the efficiency of testing a service model.
In one aspect, an embodiment of the present application provides a service processing method, where the method includes:
acquiring a target live broadcast data stream in a live broadcast service;
calling a service model to identify a target live broadcast data stream to obtain an identification result of the target live broadcast data stream, wherein the identification result comprises a decision value of the target live broadcast data stream belonging to a target category;
performing decision processing on the label display of the target live broadcast data stream according to the recognition result to obtain a decision result; and the number of the first and second groups,
and testing the service model based on the decision result.
On the other hand, an embodiment of the present application provides a service processing apparatus, where the apparatus includes:
the acquisition unit is used for acquiring a target live broadcast data stream in a live broadcast service;
the processing unit is used for calling the service model to identify the target live broadcast data stream to obtain an identification result of the target live broadcast data stream, and the identification result comprises a decision value of the target live broadcast data stream belonging to a target category;
the processing unit is also used for carrying out decision processing on the label display of the target live broadcast data stream according to the identification result to obtain a decision result; and the number of the first and second groups,
and the processing unit is also used for testing the service model based on the decision result.
In one implementation, a live broadcast service is accessed to a micro service interface; the processing unit is configured to, when acquiring a target live broadcast data stream in a live broadcast service, specifically:
and calling a micro-service interface to receive a target live broadcast data stream of the live broadcast service.
In an implementation manner, the processing unit is configured to, when calling the service model to perform recognition processing on the target live broadcast data stream to obtain a recognition result of the target live broadcast data stream, specifically:
performing framing processing on a target live broadcast data stream to obtain one or more image frames of the target live broadcast data stream;
calling a service model to identify each image frame to obtain an identification sub-result of each image frame;
and calculating to obtain the identification result of the target live broadcast data stream based on the identification sub-result of each image frame.
In one implementation, the decision result comprises a first decision result; the processing unit is configured to perform decision processing on the tag display of the target live broadcast data stream according to the recognition result, and when a decision result is obtained, the processing unit is specifically configured to:
comparing the decision value with a decision threshold for the target class;
and if the decision value is greater than the decision threshold value, obtaining a first decision result, wherein the first decision result is used for indicating a category label of a target category displayed in a playing interface of the target live broadcast data stream.
In one implementation, the decision result comprises a second decision result; a processing unit further to:
and if the decision value is not greater than the decision threshold value, obtaining a second decision result, wherein the second decision result is used for indicating that the category label of the target category is not displayed in the playing interface of the target live broadcast data stream.
In one implementation, the live broadcast service includes N stream segments, where N is an integer greater than 1, and the target live broadcast data stream is any one of the N stream segments; a processing unit further to:
calling a service model to identify N-1 stream segments except the target live broadcast data stream in the N stream segments to obtain identification results of the N-1 stream segments;
and generating a recognition result curve graph based on the recognition results of the N-1 stream segments and the recognition result of the target live broadcast data stream.
In an implementation, the processing unit is configured to, when testing the service model based on the decision result, specifically:
acquiring a reference label of a target live broadcast data stream, wherein the reference label indicates that a category label of a target category should be displayed or not displayed in a playing interface of the target live broadcast data stream;
comparing the decision result with the reference label;
if the decision result is matched with the reference label, determining that the identification performance of the service model meets the test requirement;
and if the decision result is not matched with the reference label, determining that the identification performance of the service model cannot meet the test requirement.
In one implementation, the target live broadcast data stream is a stream segment in a test set, the test set includes M stream segments and a reference label corresponding to each stream segment, and M is an integer greater than 1; the test set is used for testing the service model; the processing unit is configured to, when testing the service model based on the decision result, specifically:
calling a service model to identify M-1 stream segments except the target live broadcast data stream in the M stream segments to obtain identification results of the M-1 stream segments;
performing decision processing on each stream segment in the M-1 stream segments according to the identification results of the M-1 stream segments to obtain decision results of the M-1 stream segments;
comparing the decision results of the M-1 stream segments and the decision result of the target live data stream with corresponding reference labels;
and counting the target number of the flow segments of which the decision results are matched with the reference labels, and generating a performance test result for the service model test based on the target number and the M flow segments.
In one implementation, the processing unit is further configured to:
if the performance test result is greater than the test result threshold value, determining that the identification performance of the service model meets the test requirement;
and if the performance test result is not greater than the test result threshold value, determining that the identification performance of the service model does not meet the test requirement.
In one implementation, if the recognition performance of the service model does not meet the test requirement, the processing unit is further configured to:
adding the stream fragments of which the decision results are not matched with the reference labels into a training set, wherein the training set is used for training the service model;
and training the business model by adopting a training set to obtain the optimized business model.
In another aspect, the present application provides a service processing device, including:
a processor for loading and executing a computer program;
a computer-readable storage medium, in which a computer program is stored, which, when executed by a processor, implements the above-described service processing method.
In another aspect, the present application provides a computer-readable storage medium storing a computer program adapted to be loaded by a processor and to execute the above-mentioned business process method.
In another aspect, the present application provides a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The processor of the service processing device reads the computer instruction from the computer readable storage medium, and the processor executes the computer instruction, so that the service processing device executes the service processing method.
In the embodiment of the application, the target live broadcast data stream in the live broadcast service can be acquired, and the service model is called to perform recognition processing and decision processing on the target live broadcast data stream, so that the service model can be tested according to the decision result of the decision processing. In the scheme, the target live broadcast data stream can be directly pulled from the live broadcast service for identification processing and decision processing, and the test personnel do not need to manually start broadcasting; and the service model is tested by adopting a decision result of the target live broadcast data stream, and the mode of directly pulling the target live broadcast data stream from the live broadcast service to test the service model shortens the test link length of the service model and improves the test efficiency of the service model.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1a is a diagram illustrating material information for training and testing business models provided by an exemplary embodiment of the present application;
FIG. 1b is a schematic diagram illustrating an architecture of a business processing system according to an exemplary embodiment of the present application;
FIG. 1c is a flow diagram illustrating a business process scenario provided by an exemplary embodiment of the present application;
FIG. 1d is a schematic diagram illustrating a scenario of a business processing system according to an exemplary embodiment of the present application;
FIG. 1e is a block diagram illustrating a mainstream architecture for testing a business model according to an exemplary embodiment of the present application;
fig. 2 is a flowchart illustrating a business processing method according to an exemplary embodiment of the present application;
FIG. 3 is a diagram illustrating a dance tag in a playback interface of a target live data stream according to an exemplary embodiment of the present application;
FIG. 4 is a diagram illustrating a determination of category labels presented in a play interface according to an exemplary embodiment of the present application;
fig. 5 is a flowchart illustrating a business processing method according to an exemplary embodiment of the present application;
FIG. 6 illustrates a schematic diagram of a graph for generating recognition results provided by an exemplary embodiment of the present application;
FIG. 7 is a diagram illustrating a graph of identification results of decision value composition for a class flow segment according to an exemplary embodiment of the present application;
fig. 8 is a schematic structural diagram of a service processing apparatus according to an exemplary embodiment of the present application;
fig. 9 shows a schematic structural diagram of a service processing device according to an exemplary embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The technical terms and concepts related to the embodiments of the present application will be briefly described below, wherein:
1) artificial Intelligence (AI).
Artificial intelligence is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making. The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.
The embodiment of the application relates to machine learning/deep learning directions and the like included in the artificial intelligence technology. Machine Learning (ML) is a multi-domain cross subject, and relates to multiple subjects such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like, and is used for specially researching how a computer simulates or realizes human Learning behaviors to acquire new knowledge or skills and reorganizing an existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning. Machine learning can be viewed as a task whose goal is to let machines (computers in a broad sense) learn to obtain human-like intelligence. For example, a human will play go, and a computer program (AlphaGo or AlphaGo Zero) is designed to master the knowledge of go, a program to play go. Various methods can be used for realizing the task of machine learning, such as neural networks, linear regression, decision trees, support vector machines, Bayesian classifiers, reinforcement learning, probability map models, clustering and other various methods.
2) And (4) a label.
A tag may be a type of mark (or logo) used to mark a product or multimedia data (e.g., video, picture, audio, etc.). Specifically, the tag often contains a keyword which is highly related to the multimedia data (or the product), so that by tagging the multimedia data, the feature of the multimedia data can be indicated through the tag, thereby realizing the classification of the multimedia data, and facilitating the user to easily realize the viewing, retrieving, sharing and the like of the multimedia data. For example: in the process of live broadcasting by a main broadcast, a tag can be displayed on a playing interface for playing a live data stream (or called a video stream) related to the main broadcast, and the tag can be used for marking the category of an action executed by the main broadcast so as to realize the classification of the live data stream. For example, if the host broadcasts a live song, the label displayed on the display interface may be a song label (e.g., the song label contains the character string "song"); if the anchor dances on the live broadcast, the label displayed on the playing interface can be a dance label (if the dance label contains a character string 'dance'), and the like; this facilitates a user (e.g., a broadcaster or a viewer watching a live broadcast) to quickly know the characteristics or categories of the live broadcast data stream through the tags of the live broadcast data stream, thereby helping the user select a preferred live broadcast data stream for viewing.
For convenience of description, multimedia data is taken as a live data stream in a live service scene, and a category tag of the live data stream is taken as an example for introduction, which is described herein specifically.
Currently, a business model can be adopted to identify a live broadcast data stream, so as to determine the category to which the live broadcast data stream belongs and a corresponding category label according to an identification result; the business model can be obtained by training a machine learning model (such as a neural network model) by adopting a sample data stream (or called sample data), and the trained business model has the functions of image recognition, audio recognition and the like. Therefore, the identification performance of the service model directly affects the accuracy of identification of the category and the category label to which the live data stream belongs, and it is important to test the performance of the service model to know the performance of the service model. An exemplary material information for training and testing the business model may be referred to fig. 1a, which is a first diagram shown in fig. 1a, and a training set for training the business model to identify a singing category may include 5000 sample data streams; the category to which 1000 sample data streams belong in 5000 sample data streams is a singing category, and the category to which 4000 sample data streams belong is a non-singing category; the test set for testing the recognition performance of the service model for recognizing the singing category may include 1000 test data streams, wherein 200 test data streams among the 1000 test data streams belong to the category of the singing category, and 800 test data streams belong to the category of the non-singing category. Similarly, for the material information used for training or testing the business model to identify the dance category, reference may be made to the second diagram shown in fig. 1a, and for a specific implementation manner, reference may be made to the related description of the first diagram shown in fig. 1a, which is not described herein again. It should be noted that the number and types of the sample data streams and the test data streams are all given as examples, and the embodiments of the present application do not limit the number and types of the sample data streams and the test data streams.
The embodiment of the present application provides a service processing scheme for testing a service model, and an architecture diagram of the scheme can be referred to fig. 1 b; as shown in fig. 1b, in the embodiment of the present application, an intermediate service is added between the identification module and the background server, where the intermediate service may be a flash service, and the flash service is a lightweight framework and may be used for developing a website or a web service. That is to say, the identification service provided by the embodiment of the application is implemented based on the flash service, that is, each function (such as a function of splitting a live data stream into image frames, a function of identifying the image frames, and the like) provided by the identification service is deployed on the flash service, which reduces the workload of developing the identification service to a certain extent, and is easier to perform an interface automation test based on the flash service; the interface automatic test can be that a live data stream (or a test data stream) is directly acquired through an interface to carry out automatic identification so as to realize the test of a service model, and the live data stream (or the test data stream) does not need to be played manually (for example, a video is triggered to be played) so as to realize the test.
Specifically, after an intermediate service is newly added to the identification service, the background server can actively access the test data stream (or the live data stream in the live service) in the test set through an interface provided by the intermediate service, and the identification service does not need to actively request the test data stream from the background server every time. In addition, the identification of invalid live data streams can be prevented in time in an interface automatic test mode, and the identification speed and the identification efficiency are improved to a certain extent; the invalid live stream may be a live stream whose recognition processing is canceled, or the like. For example, when the background server detects that the test data stream currently identified by the identification module is an invalid data stream, the background server may actively issue a next test data stream, so that the identification module stops the identification task for the currently identified test data stream, and receives and starts to execute the identification task for the next test data stream. And only one identification thread can be created when one live data stream is identified, the previous identification result is multiplexed, and the identification performance is good.
Based on the architecture diagram shown in fig. 1b, a flow diagram of a service processing scheme provided in the embodiment of the present application may be referred to in fig. 1 c; the following briefly introduces a flow of a service processing scheme proposed in the embodiment of the present application with reference to schematic diagrams shown in fig. 1b and fig. 1 c: firstly, acquiring a target live broadcast data stream (such as any live broadcast data stream) from a live broadcast service; secondly, calling a service model to identify the target live broadcast data stream, and performing decision processing on the label display of the target live broadcast data stream according to the identification result of the service model to obtain a decision result, wherein the decision result indicates that the label is displayed or not displayed in a playing interface of the target live broadcast data stream; and finally, testing the service model according to the decision result of the decision processing. The process can directly pull the target live broadcast data stream from the live broadcast service without manual broadcasting by testers; and moreover, the service model is tested by adopting a decision result of the target live broadcast data stream, so that the link length of the service model is shortened, and the testing efficiency of the service model is improved.
The following describes a service processing scheme according to an embodiment of the present application with reference to an actual service processing scenario. Referring to fig. 1d, fig. 1d is a schematic diagram illustrating an architecture of a service processing system according to an exemplary embodiment of the present application; as shown in fig. 1d, the service processing system includes a terminal 101, a terminal 102 and a server 102; in the embodiment of the present application, the names and numbers of the terminals (e.g., the terminal 101 and the terminal 102) and the names and numbers of the servers are not limited. The terminal 101 is a terminal device that can be used to capture images (or videos) of the environment where the anchor is located, and the terminal can be any device with image capturing function. The terminal 102 is a terminal used by a user (or simply, viewer) watching a live broadcast, and the terminal 102 is operable to receive and present a live data stream. Terminal 101 or terminal 102 may include, but is not limited to: the smart device comprises smart devices capable of performing touch screen, such as a smart phone, a tablet computer, a portable personal computer, a mobile internet device, a smart television, a vehicle-mounted device, and a head-mounted device. Both the terminal 101 and the terminal 102 can run an application program (or simply referred to as an application) to realize the function of live broadcasting or live broadcasting watching.
The server 103 may be the aforementioned background server, or the server 103 may be a Graphics Processing Unit (GPU) with a recognition service deployed; for convenience of illustration, the following description will take a server as an example of a device with an identification service deployed therein. Specifically, the server 103 may be a background server of the terminal 101 or the terminal 102, or the server 103 may be a background server of an application running in the terminal 101 or an application running in the terminal 102, and is configured to interact with a terminal (such as the terminal 101 or the terminal 102) running an application, so as to provide computing and application service support for any application. The server 103 may be an independent physical server, or may be a server cluster or a distributed system including a plurality of physical servers. The terminal (including the payment terminal and the user terminal) and the server may be directly or indirectly communicatively connected in a wired or wireless manner, and the connection manner between the terminal and the server is not limited in the embodiments of the present application.
With reference to fig. 1d, the overall process of the service processing scenario according to the embodiment of the present application may include: firstly, when the anchor is opened and the terminal 101 is used for live broadcasting, the terminal 101 can acquire images of the environment where the anchor is located, and obtain a live data stream composed of multiple frames of images. Secondly, the server 103 may pull the live data stream from the terminal 101, and perform recognition processing and decision processing on the live data stream to obtain a decision result. Finally, the server 103 tests the service model according to the decision result to test the recognition performance of the service model. Of course, the server 103 also sends the decision result to the terminal 102 used by the viewer, so that the terminal 102 determines whether to display the tag in the playing interface of the live data stream and which type of tag is displayed according to the decision result. With reference to fig. 1d, after the server 103 processes the live data stream to obtain a decision result, if the decision result indicates that the category tag of the target category to which the live data stream belongs is displayed in the playing interface of the live data stream, the category tag 104 of the target category is displayed on the playing interface of the live data stream displayed by the terminal 102 used by the viewer; this helps the viewer to quickly understand the actions performed by the anchor and enrich the display content of the playing interface.
It should be noted that, after obtaining the decision result about the live data stream, the server 103 may also return the decision result to the terminal 101 used by the anchor, so that the terminal 101 determines whether to display the tag of the live data stream and which type of tag to display on the terminal screen according to the decision result; therefore, the anchor can know whether the action executed by the anchor is correctly identified by the identification service in real time, and can feed back the problems generated in live broadcasting more quickly.
Practice shows that the scheme of the embodiment of the application has remarkable advantages in the process of testing the service model. The following description will be made by taking a comparison between the present embodiment and the conventional main embodiment as an example. The schematic architecture of the existing mainstream scheme for testing the service model can be seen in fig. 1 e. As shown in fig. 1e, each time the existing identification service performs an identification operation, a tester needs to actively and manually start a test data stream, and then test a service model by observing whether a tag is displayed on a terminal screen; the testing link is long, a large amount of testing manpower is consumed, and the testing efficiency is not high. The scheme of the application has the following advantages: the service processing scheme provided by the embodiment of the application adds the flash service in the identification service, so that the automatic test of the interface is easier to realize, and compared with the mainstream manual broadcasting of the test data stream by a tester, the service processing scheme can realize the identification of batch test data streams by leading the test data streams into the identification service through the interface, thereby shortening the test link length of the service model and improving the test efficiency of the service model. And the flash service is also made into a scoring platform, when the live data stream with wrong identification is fed back by an anchor or an operator, the live data stream (or playback data stream and the like) can be directly input into the scoring platform, and whether the business model identifies the live data stream accurately is judged through a decision value on the scoring platform, so that the problem is quickly positioned, the time for troubleshooting is shortened, and the cost for solving the problem is reduced.
Based on the service processing scheme described above, a more detailed service processing method is provided in the embodiments of the present application, and the service processing method provided in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.
Fig. 2 is a flowchart illustrating a business processing method according to an exemplary embodiment of the present application; the service processing method may be executed by a server, and may include, but is not limited to, steps S201 to S204:
s201: and acquiring a target live broadcast data stream in the live broadcast service.
In specific implementation, the embodiment of the application supports the access of a live broadcast service to a micro service interface, so as to call the micro service interface to receive a target live broadcast data stream of the live broadcast service. The target live data stream may refer to any live data stream included in the live service, such as a video stream of a target time period (e.g., 10 seconds) generated in the live process. The micro-service may include a flash service, and the micro-service interface may refer to an interface developed based on the flash service for receiving a target live data stream. The live broadcast service can be provided by any application program with a live broadcast function. An application refers to a computer program designed to perform a specific task or tasks. According to the operation mode of the application program, the application program can include but is not limited to: the method includes installing and running an application program in a terminal. ② the application program without installation, namely the application program which can be used without downloading and installation, this kind of application program is also commonly called small program, and it is usually operated in the client as subprogram. A web application opened by the browser; and so on. The embodiment of the present application does not limit the types of application programs for providing live broadcast services, and is specifically described herein.
S202: and calling a service model to identify the target live broadcast data stream to obtain an identification result of the target live broadcast data stream.
The identification result of the target live data stream may include a decision value that the target live data stream belongs to the target category. Wherein the object class corresponds to the business model, i.e. the business model can be used to identify the object class (e.g. dance class, sports class, gymnastics class, etc.). The decision value of the target live data stream belonging to the target category can be used to indicate that: probability that a target live data stream belongs to a target category; for example, if the decision value of the first target live data stream belonging to the target category is 50%, and the decision value of the second target live data stream belonging to the target category is 70%, it may be determined that: the probability that the second target live data stream belongs to the target category is greater than the probability that the first target live data stream belongs to the target category.
The implementation manner of calling the service model to identify and process the target live broadcast data stream may include: firstly, performing framing processing on a target live broadcast data stream to obtain one or more image frames of the target live broadcast data stream; as described above, the identification service provided by the embodiment of the present application is implemented based on the flash service, so that the frame division processing of the target live broadcast data stream can be automatically implemented based on the flash service, the frame division speed and efficiency are improved, and the identification efficiency is further improved. Secondly, calling a service model to identify each image frame to obtain an identification sub-result of each image frame; the recognition sub-result of any image frame may be used to indicate the probability that the any image frame belongs to the object class. Finally, obtaining an identification result of the target live broadcast data stream based on the identification sub-result of each image frame; specifically, the weighted calculation may be performed on the recognition sub-results of one or more image frames included in the target live broadcast data stream to obtain the recognition result of the target live broadcast data stream.
For example, the target live data stream includes 5 image frames, wherein the identification sub-result of the image frame 1 includes a decision value of 20%, the identification sub-result of the image frame 2 includes a decision value of 60%, the identification sub-result of the image frame 3 includes a decision value of 50%, the identification sub-result of the image frame 4 includes a decision value of 60%, and the identification sub-result of the image frame 5 includes a decision value of 50%, and then the average calculation of the decision values included in the identification sub-results of the 5 image frames results in an identification result of (20% + 60% + 50% + 60% + 50%)/5 ═ 46%. Of course, in addition to using percentage to represent the decision value, the decision value of the target live data stream may also be represented in the form of decimal, fractional, or the like, which is not limited in this embodiment of the application.
In addition, the business model mentioned in the embodiment of the present application may support the identification of one category, or support the identification of multiple categories at the same time. In one implementation, a business model may support identifying a category. Specifically, assuming that the service model can be used for identifying the target category, the service model is used for identifying the target live broadcast data stream, and the obtained identification result may include a decision value that the target live broadcast data stream belongs to the target category. In this implementation manner, if there is a need to identify multiple categories of the target live broadcast data stream, multiple service models supporting identification of different categories may be simultaneously used to identify the target live broadcast data stream, so as to obtain a decision value of each category to which the target live broadcast data stream belongs. For example, assuming that a first business model supports identification of a first target category and a second business model supports identification of a second target category, the first business model is adopted to identify a target live broadcast data stream, and the available target live broadcast data stream belongs to a decision value of the first target category; similarly, the target live broadcast data stream is identified by adopting a second service model, and the obtained target live broadcast data stream belongs to a decision value of a second target category.
In other implementations, the business model may support identifying multiple categories. Specifically, assuming that the service model can support identification of a first target category and a second target category, after the service model is used to identify the target live broadcast data stream, the obtained identification result may include a decision value that the target live broadcast data stream belongs to the first target category and a decision value that the target live broadcast data stream belongs to the second target category. For example, the service model may be used to identify a dance category and a singing category, and then after the target live broadcast data stream is identified by using the service model, the obtained identification result may include decision values of the target live broadcast data stream belonging to the dance category and the singing category. The embodiments of the present application do not limit the types of the categories that the business model supports to identify, and are described herein.
S203: and performing decision processing on the label display of the target live broadcast data stream according to the recognition result to obtain a decision result.
The decision result can be used to indicate whether to show a category label of the target category in a play interface of the target live data stream. The decision result may include a first decision result, where the first decision result is used to indicate that a category label of a target category is displayed in a playing interface of the target live broadcast data stream, that is, the target live broadcast data stream belongs to the target category; the decision result may further include a second decision result, where the second decision result is used to indicate that the category label of the target category is not displayed in the playing interface of the target live broadcast data stream, that is, the target live broadcast data stream does not belong to the target category. In specific implementation, a decision value contained in the recognition result can be compared with a decision threshold of a target category; if the decision value is larger than the decision threshold, obtaining a first decision threshold, and indicating that the target live broadcast data stream belongs to a target category; otherwise, if the decision value is not greater than (i.e., less than or equal to) the decision threshold, a second decision result is obtained, which indicates that the target live broadcast data stream does not belong to the target category. The decision threshold corresponding to each category may not be the same, and the decision threshold may be preset by a service person according to a service requirement.
For example, assume that a service model is called to identify a target live broadcast data stream, and the obtained identification result indicates that the target live broadcast data stream belongs to a dancing category with a decision value of 80 and a decision threshold of 70; and performing decision processing on the label display of the target live broadcast data stream according to the identification result, wherein the obtained decision result is a first decision result, namely a dance label needs to be displayed in a playing interface of the target live broadcast data stream. An exemplary schematic diagram of showing a dance tag in a playing interface of a target live data stream can be seen in fig. 3, as shown in fig. 3, a dance tag 3011 is shown in a playing interface 301 of a target live data stream, which can help a viewer to quickly know the category of the target live data stream without the viewer playing the target live data stream.
As described in step S202, if different service models are used to identify the target live broadcast data stream, or one service model with multiple categories is used to identify the target live broadcast data stream, multiple identification results of the target live broadcast data stream are obtained, where each identification result corresponds to one category. In this implementation manner, the decision-making process is performed on the tag display of the target live broadcast data stream according to the multiple recognition results, and the implementation manner of obtaining the decision-making result may include: assuming that the identification result corresponding to the target live broadcast data stream comprises an identification result 1, an identification result 2 and an identification result 3, wherein the identification result 1 corresponds to a first category, the identification result 2 corresponds to a second category and the identification result 3 corresponds to a third category; then, the decision value contained in each recognition result is compared with the decision threshold of the corresponding category, for example, the decision value contained in the recognition result 1 is compared with the decision threshold of the first category, so as to obtain 3 decision results; therefore, whether the category label is displayed in the playing interface of the target live broadcast data stream and which category label is displayed can be judged according to the number of the first decision results and the number of the second decision results contained in the 3 decision results.
In the following, several optional implementation manners for determining which category of category label is displayed in the play interface of the target live broadcast data stream according to the 3 decision results are described, taking the number of the decision results of the target live broadcast data stream as 3 as an example. Wherein: (1) and if each decision result in the 3 decision results is a second decision result, namely the target live broadcast data stream does not belong to any one of the first category, the second category and the third category, determining that no category label is displayed in a playing interface of the target live broadcast data stream. (2) And if any decision result in the 3 decision results is a first decision result, that is, the target live broadcast data stream belongs to any one of a first category, a second category and a third category, and if the target live broadcast data stream belongs to the first category, determining that a category label of the first category is displayed in a playing interface of the target live broadcast data stream. (3) If at least 2 decision results in the 3 decision results are first decision results, namely the target live broadcast data stream belongs to at least two categories of a first category, a second category and a third category, determining that category labels of the at least two categories are displayed in a playing interface of the target live broadcast data stream; or randomly selecting one or more categories of category labels from the at least 2 categories to display; or, a target decision value (e.g., the decision value with the largest value) is selected from the decision values corresponding to the at least 2 decision results according to the weight, and a category label of a category corresponding to the target decision value is displayed in a play interface of the target live broadcast data stream. The embodiment of the present application is not limited to which implementation described above is specifically adopted to display the category label, and is described here.
With reference to fig. 4, for example, the category corresponding to the decision value with the largest weight is selected from the multiple decision values to perform tag display, and the display of the tag in the play interface of the target live broadcast data stream under the multiple recognition results is briefly described. Referring to fig. 4, it is assumed that there are a service model 1 having an audio recognition function and a recognition model 2 having a video recognition function; wherein the service model 1 can identify a singing category, a chatting category, a music category, etc., and the service model 2 can identify a dancing category, a sports category, a show category, etc. Now, after the business model 1 is adopted to identify the target live broadcast data stream 1, the decision value that the target live broadcast data stream 1 belongs to the singing category is 70, and after the business model 2 is adopted to identify the target live broadcast data stream 1, the decision that the target live broadcast data stream 1 belongs to the show field is 50; assuming that the decision threshold for the singing category is 60 and the decision threshold for the show category is 40, it may be determined that the target live data stream belongs to both the singing category and the show category, but since the decision value 70 of the target live data stream belonging to the singing category is greater than the decision value 50 of the target live data stream belonging to the show category, it is determined that the category label of the singing category is shown in the playing interface of the target live data stream, and the category label of the singing category may be the category label 401 shown in fig. 4. Similarly, assuming that after the service model 1 and the service model 2 are adopted to identify the target live broadcast data stream 2, the decision value of the target live broadcast data stream 2 belonging to the music category is 60 (for example, the decision threshold of the music category is 50), and the decision value of the target live broadcast data stream belonging to the sports category is 70 (for example, the decision threshold of the sports category is 50), it is determined that the target live broadcast data stream simultaneously belongs to the music category and the dance category, but since the decision value 70 of the target live broadcast data stream belonging to the sports category is greater than the decision value of the target live broadcast data stream belonging to the music category, it is determined that the category label of the sports category is displayed in the playing interface of the target live broadcast data stream, and the category label of the sports category can be the category label 402 shown in fig. 4.
S204: and testing the service model based on the decision result.
It is understood that the decision result of the target live broadcast data stream is obtained by identifying and processing the target live broadcast data stream by using the service model, and then the service model can be tested based on the decision result to measure the identification performance of the service model. Specifically, the decision result may be compared with an actual tag display condition (or an expected tag display condition) of the target live broadcast data stream to obtain a comparison result; if the comparison result indicates that the decision result is matched with (or the same as) the target live broadcast data stream expected label display condition, determining that the service model correctly identifies the target live broadcast data stream; if the comparison result indicates that the decision result is not matched with the target live broadcast data stream expected label display condition, determining that the business model has an error in identifying the target live broadcast data stream; therefore, the recognition performance of the business model can be tested by adopting the decision result of the target live broadcast data stream. For example, assuming that the service model performs recognition processing and decision processing on the target live broadcast data stream, and the obtained decision result indicates that a category label of a dancing category is not displayed in a playing interface of the target live broadcast data stream; however, the type to which the target live broadcast data stream actually belongs is a dancing type, namely a dancing label is displayed in a playing interface of the target live broadcast data stream, and then it is determined that the business model identifies the target live broadcast data stream wrongly, and further it is determined that the business model needs to be optimized. The expected tag display condition of the target live data stream can be fed back by a main broadcast or observed by a tester after the target live data stream is broadcast.
Based on the above description, the embodiment of the application can also help the tester to quickly locate the problem of the anchor or the feedback of the operator. For example, assume that the anchor feedback: and if the dance tag is not displayed in the playing interface of the live broadcast data stream, namely the anchor actually executes the dance action, but the dance tag is not displayed in the playing interface of the live broadcast data stream, the live broadcast data stream can be input to the identification service, and the live broadcast data stream is identified and decided by the service model. If the decision result indicates that the dance label of the dance category is displayed in a playing interface of the live broadcast data stream, determining that the service model accurately identifies the live broadcast data stream, namely that the identification of the service model has no error; at this time, the problem can be located in the link for transmitting the decision result to the terminal, that is, the problem exists in the link for returning the decision result to the terminal, so that the category label of the dancing category cannot be displayed in the playing interface of the live data stream. And if the decision result indicates that the dance label of the dance category is not displayed in the playing interface of the live broadcast data stream, determining that the live broadcast data stream is identified incorrectly (or the identification effect is poor) by the service model, namely positioning the problem to an identification service so as to optimize the service model. Through the process, the problem of the feedback of the anchor or the operator can be rapidly positioned, and the problem troubleshooting efficiency is improved.
In summary, the embodiment of the application can obtain the target live broadcast data stream in the live broadcast service, and call the service model to perform recognition processing and decision processing on the target live broadcast data stream, so as to test the service model according to the decision result of the decision processing. In the scheme, the target live broadcast data stream can be directly pulled from the live broadcast service for identification processing and decision processing, and the test personnel do not need to manually start broadcasting; the method for testing the service model by directly pulling the target live broadcast data stream from the live broadcast service shortens the test link length of the service model and improves the test efficiency of the service model.
Fig. 5 is a flowchart illustrating a business processing method according to an exemplary embodiment of the present application; the service processing method may be executed by a server, and the service processing method may include, but is not limited to, steps S501 to S507:
s501: and acquiring a target live broadcast data stream in the live broadcast service.
S502: and calling a service model to identify the target live broadcast data stream to obtain an identification result of the target live broadcast data stream.
It should be noted that, for specific implementation manners of steps S501 to S502, reference may be made to related descriptions of specific implementation manners of steps S201 to S202 in the embodiment shown in fig. 2, and details are not described herein.
In addition, the live broadcast service may include N stream segments, where N is an integer greater than 1, and the target live broadcast data stream is any one of the N stream segments. The live broadcast service mentioned in the embodiment of the present application may be a service generated when any anchor broadcasts live through an application program, and in this implementation manner, N stream segments included in the live broadcast service are generated when the same anchor (or terminals used by the same anchor) broadcasts directly; or, the live broadcast service may also be a service generated when different anchor broadcasts through an application program, and in this implementation, the N stream segments included in the live broadcast service may be generated when different anchor broadcasts (or terminals used by different anchor broadcasts, respectively) broadcast.
The categories to which the N stream segments included in the live service belong may be the same or different. For example: the live broadcasting service is generated by a main broadcasting live broadcasting, and when the main broadcasting performs dancing action in a first time period, the category of the stream segment generated based on the first time period is a dancing category; when the host performs a singing action in a second time period, the category to which the stream segment generated based on the second time period belongs is a singing category; the first time period and the second time period are different time periods. For another example: the live broadcast service is generated by two anchor broadcasts during live broadcast, and when the actions executed by the two anchor broadcasts are dancing, the categories of two stream segments generated based on the two anchor broadcasts are dancing categories; when the actions performed by the two anchor broadcasters are dance and singing, respectively, the categories to which the streaming segments generated based on the two anchor broadcasters belong are a dance category and a singing category, respectively. The present embodiment does not limit the category to which the N stream segments included in the multicast service belong.
The embodiment of the application supports visual output of the recognized recognition result. Specifically, the method can identify N stream segments included in the live broadcast service to obtain N identification results of the N stream segments; obtaining an identification result curve graph according to the N identification results of the N flow segments, realizing visual output of the identification results, and facilitating a tester to directly use the N identification results; of course, when the live broadcast service only includes the target live broadcast data stream, an identification result graph may also be generated based on the identification result of the target live broadcast data stream, but the identification result graph at this time only includes the decision value that the target live broadcast data stream belongs to the target category. In the specific implementation, a service model can be called to identify and process N-1 stream segments except a target live broadcast data stream in the N stream segments to obtain identification results of the N-1 stream segments; and generating a recognition result curve graph based on the recognition results of the N-1 stream segments and the recognition result of the target live broadcast data stream. It can be understood that the process of invoking the service model to identify and process N-1 stream segments is similar to the process of invoking the service model to identify and process a target live broadcast data stream described in the embodiment shown in fig. 2, and is not described herein again.
An exemplary description is given below with reference to fig. 6 to generate the recognition result graph, and assuming that the service model supports recognition of a target category and supports recognition of a live broadcast data stream with a broadcast length of 10 seconds, and a broadcast time of a live broadcast data stream to be recognized in the currently acquired live broadcast service is 120 seconds, the live broadcast data stream of 120 seconds may be divided into 12 stream segments with a broadcast time of 10 seconds, and the target live broadcast data stream is any one of the 12 stream segments; and calling a service model to sequentially identify the 12 stream segments according to the playing time sequence of the 12 stream segments, and sequentially marking the obtained 12 identification results on an identification result curve graph to obtain the identification result curve graph of the 12 stream segments.
As shown in fig. 6, it is assumed that the first stream segment (i.e. the playing time period is 00:00-00:10) of the data stream is identified, and the identification result of the first stream segment includes that the decision value of the first stream segment belonging to the target category is 28; identifying a second stream segment (namely the playing time period is 00:10-00:20) of the data stream, wherein the obtained identification result of the second stream segment comprises that the decision value of the second stream segment belonging to the target class is 50; and a decision value of 80 for the third stream segment (i.e., the playing period of 00:20-00:30), 92 for the fourth stream segment (i.e., the playing period of 00:30-00:40), 91 for the fifth stream segment (i.e., the playing period of 00:40-00:50), 72 for the sixth stream segment (i.e., the playing period of 00:50-01:00), 40 for the seventh stream segment (i.e., the playing period of 01:00-01:10), 44 for the eighth stream segment (i.e., the playing period of 01:10-01:20), 78 for the ninth stream segment (i.e., the playing period of 01:20-01:30), 89 for the tenth stream segment (i.e., the playing period of 01:30-01:40), 89 for the eleventh stream segment (i.e., the playing period of 01:40-01:50), and a twelfth stream segment (i.e., the playing period of 01:40-01:50) The decision value for the segment (i.e., the playing time period is 01:50-02:00) is 89. Assuming that the decision threshold of the target class is 60, the flow segments with decision values larger than the decision threshold, i.e. the flow segments corresponding to the decision values with decision values above the decision threshold 60, in the 12 flow segments, can be quickly determined on the graph of the recognition result, and referring to fig. 6, the flow segments with decision values larger than the decision threshold include: a third stream segment, a fourth stream segment, a fifth stream segment, a sixth stream segment, a ninth stream segment, a tenth stream segment, an eleventh stream segment, and a twelfth stream segment.
If the received feedback information of the anchor is: the category label of the target category is not shown in the playing time period 01:20-01:30, that is, the anchor performs the action of the target category in the playing time period 01:20-01:30, but does not show the category label of the target category in the playing interface of the stream segment. The tester can directly find the decision value of the ninth stream segment corresponding to the playing time period 01:20-01:30 according to the recognition result graph, and eliminate the problem according to the comparison result of the decision value of the ninth stream segment and the decision threshold. For example, if the decision value of the ninth flow segment is 74 and greater than the decision threshold 60, it is determined that the service model correctly identifies the ninth flow segment, that is, the service model correctly identifies that the ninth flow segment belongs to the target class, and it can be determined that the problem occurs in the link where the decision result is returned to the terminal. In conclusion, the problem can be better positioned by the tester by identifying the result curve graph, and the problem troubleshooting efficiency is improved.
It should be noted that fig. 6 is only an exemplary graph of the recognition result, and in practical applications, the expression form of the recognition result curve may also be adaptively changed; for example, the specific data of the decision value is displayed at the adjacent position of each decision value on the recognition result graph; and the like; the embodiment of the present application does not limit the expression form of the graph of the recognition result. In addition, in fig. 6, 12 stream segments included in the live broadcast service are obtained by splitting a complete live broadcast data stream, but it can be understood that N stream segments included in the live broadcast service may also be independent stream segments, and a determination method of the stream segments included in the live broadcast service in the embodiment of the present application is not limited.
S503: and performing decision processing on the label display of the target live broadcast data stream according to the recognition result to obtain a decision result.
It should be noted that, for a specific implementation manner of step S503, reference may be made to related descriptions of a specific implementation manner of step S203 in the embodiment shown in fig. 2, which is not described herein again.
S504: and acquiring reference labels of the target live data streams.
The reference label of the target live data stream can be used for indicating that a category label of a target category should be displayed or not be displayed in a playing interface of the target live data stream; in other words, the reference annotation of the target live data stream includes a decision result that the target live data stream is correct. The reference label of the target live data stream can be from the anchor, for example, the anchor actively feeds back the action category executed by the anchor; or the reference label can be labeled for the target live broadcast data stream after the target live broadcast data stream is played by a tester; and the determination method of the reference label of the target live data stream is not limited in the embodiment of the present application, and is described herein.
S505: the decision result is compared with the reference label.
S506: and if the decision result is matched with the reference label, determining that the identification performance of the service model meets the test requirement.
S507: and if the decision result is not matched with the reference label, determining that the identification performance of the service model does not meet the test requirement.
In steps S505 to S507, after the reference label of the target live broadcast data stream is obtained, the decision result of the identified target live broadcast data stream may be compared with the reference label, so as to determine whether the service model identifies the target live broadcast data stream accurately. And if the obtained decision result is matched with the reference label (if the decision result is the same) after the business model is called to perform recognition processing and decision processing on the target live broadcast data stream, namely the business model accurately recognizes the target category to which the target live broadcast data stream belongs, determining that the recognition performance of the business model meets the test requirement. Otherwise, if the obtained decision result is not matched with (if different from) the reference label after the business model is called to perform the identification processing and decision processing on the target live broadcast data stream, that is, the business model identifies the category to which the target live broadcast data stream belongs incorrectly, it is determined that the identification performance of the business model does not meet the test requirement.
The implementation manner of testing the service model includes that the service model is tested according to the decision result of the target live broadcast data stream shown in steps S505 to S507, and the embodiment of the present application also supports the test of the service model by using a test set. In the specific implementation, the test set is used for testing the service model, the target live broadcast data stream is a stream segment in the test set, the test set comprises M stream segments and a reference label corresponding to each stream segment, and M is an integer greater than 1; the implementation manner of testing the service model by using the test set may include:
calling a service model to identify M-1 stream segments except a target live broadcast data stream in the M stream segments to obtain identification results of the M-1 stream segments; the recognition result of any one of the stream segments includes a decision value that the any one of the stream segments belongs to the target category. And secondly, carrying out decision processing on each of the M-1 stream segments according to the identification results of the M-1 stream segments to obtain decision results of the M-1 stream segments, wherein the decision result of any one stream segment indicates whether a category label of a target category is displayed in a playing interface of the any one stream segment. Comparing the decision results of the M-1 stream segments and the decision result of the target live data stream with corresponding reference labels; in other words, comparing the decision result of each stream segment with the corresponding reference label can obtain whether the decision result of each stream segment matches with the corresponding reference label. Fourthly, counting the target number of the flow segments of which the decision results are matched with the reference labels, and generating performance test results for the service model test based on the target number and the M flow segments; if the performance test result is larger than the test result threshold value, determining that the identification performance of the service model meets the test requirement; and if the performance test result is not greater than the test result threshold value, determining that the identification performance of the service model does not meet the test requirement. The performance test result is obtained by dividing the target number of the stream segments of which the decision result is matched with the reference annotation by the total number (namely M) of the stream segments contained in the test set, and the performance test result can reflect the accuracy of the service model identification.
For example, assuming that the test set contains 6 (i.e., M ═ 6) flow segments, the traffic model supports identifying a target class whose decision threshold is 50; calling a service model to identify each flow segment contained in the test set, so as to obtain 6 identification results; and performing decision processing on each stream segment according to the 6 identification results to obtain 6 decision results. The 6 recognition results and the decision result corresponding to each stream segment can be shown in table 1.
TABLE 1
Stream fragment Recognition results (decision value) Decision results Reference label
Stream segment 1 20 Not showing category label Not showing category label
Stream fragment 2 55 Display category label Not showing category label
Stream fragment 3 60 Display category label Display category label
Stream segment 4 46 Not showing category label Not showing category label
Stream segment 5 70 Display category label Display category label
Stream fragment 6 80 Display category label Display category label
As shown in table 1, the stream segments with decision results matching with the corresponding reference labels in the 6 stream segments can be obtained, including: flow segment 1, flow segment 3, flow segment 4, flow segment 5, and flow segment 6, the process of calculating the performance test result (or accuracy) of the service model can be seen in table 2.
TABLE 2
Figure BDA0003156422540000201
If the threshold value of the test result is 80%, the performance test result 83.33% of the service model is determined to be greater than the threshold value of the test result 80%, that is, the recognition performance of the service model is determined to meet the test requirement, which indicates that the recognition performance of the service model is better. The test result threshold may be preset by a service person according to a service requirement, the test result threshold of each service model may not be the same, and the determination manner and the specific numerical value of the test result threshold are not limited in the embodiments of the present application, which is described herein.
In addition, after the recognition performance of the service model is determined not to meet the test requirements, the embodiment of the application also supports the optimization of the service model so as to improve the recognition performance of the service model. Specifically, a training set can be adopted to train the business model, so as to obtain an optimized business model; the training set is used for training the service model, and may include at least one training stream segment and a preset label of each training stream segment, where the at least one training stream segment may include: and testing the stream fragments of which the decision results in the set are not matched with the corresponding reference labels. That is to say, after the service model is tested by using the test set, the test stream segments whose decision results are not matched with the corresponding reference labels can be added to the training set, so that the number of the training stream segments in the training set can be enriched, and the service model can be trained in a targeted manner, for example, if the recognition effect of the service model on a certain type of test data stream (for example, an animation cartoon including a motion played by a main broadcaster) is found to be not good in the test process, the type of test data stream is added to the training set for training, so that the optimized service model can better recognize the type of data stream.
An implementation manner for finding a stream segment in a test set, which can be added to a training set as a training stream segment, may include: firstly, screening out a stream segment with a reference label indicating that a category label should be displayed in a playing interface from a test set, wherein the stream segment which should display the category label in the playing interface is called a positive category stream segment, and the stream segment which should not display the category label in the playing interface is called a negative category stream segment in the embodiment of the application; secondly, identifying the screened positive type stream segments by adopting a service model to obtain an identification result of each positive type stream segment, and further obtaining an identification result curve chart corresponding to the positive type stream segments in the test set; and finally, detecting the positive flow segments with the decision values lower than the decision threshold in the recognition result graph, wherein the positive flow segments with the decision values lower than the decision threshold are the flow segments of which the service model cannot recognize the categories, and adding the positive flow segments with the decision values lower than the decision threshold to the training set. Through the process, the tester can conveniently and quickly search the flow segments which can be used for training the service model directly based on the recognition result curve graph, and compared with the one-by-one analysis of massive recognition results, the speed of searching the flow segments which can be used for training the service model can be improved.
With reference to fig. 7, the above implementation is exemplarily described, and as shown in fig. 7, it is assumed that the test set includes 12 segments of the forward class stream, and the service model supports identifying the target class, where the decision threshold of the target class is 50; then 12 recognition results are obtained after the 12 positive class stream segments are subjected to recognition processing, and the 12 recognition results form a curve in a recognition result curve graph; the positive class stream segments with the decision values lower than the decision threshold include a positive class stream segment 2, a positive class stream segment 5 and a positive class stream segment 8, that is, it is determined that the service model cannot accurately identify the classes to which the positive class stream segment 2, the positive class stream segment 5 and the positive class stream segment 8 belong, and then it is determined that the positive class stream segment 2, the positive class stream segment 5 and the positive class stream segment 8 are used as training stream segments added to the training set. The method for searching the training stream segment used for training the business model from the recognition result graph improves the speed of searching the training stream segment.
In the embodiment of the application, the target live broadcast data stream can be directly pulled from the live broadcast service for identification processing and decision processing, and the test personnel does not need to manually broadcast; the method for testing the service model by directly pulling the target live broadcast data stream from the live broadcast service shortens the test link length of the service model and improves the test efficiency of the service model. Or, a test set can be used for testing the service model, and the test set comprises a target live broadcast data stream, so that the accuracy of the service model can be obtained through more test data streams, and the identification performance of the service model can be further determined. In addition, the test data stream which cannot be identified by the service model in the test set can be added to the training set for carrying out secondary training optimization on the service model, so that the number of the sample data streams in the training set is enriched, and the service model can be tested in a targeted manner.
While the method of the embodiments of the present application has been described in detail above, to facilitate better implementation of the above-described aspects of the embodiments of the present application, the apparatus of the embodiments of the present application is provided below accordingly.
Fig. 8 is a schematic structural diagram of a service processing apparatus according to an exemplary embodiment of the present application; the service processing apparatus may be used as a computer program (comprising program code) running in a server; the service processing apparatus may be configured to perform some or all of the steps in the method embodiments shown in fig. 2 and fig. 5. Referring to fig. 8, the service processing apparatus includes the following units:
an obtaining unit 801, configured to obtain a target live broadcast data stream in a live broadcast service;
the processing unit 802 is configured to invoke a service model to perform recognition processing on a target live broadcast data stream to obtain a recognition result of the target live broadcast data stream, where the recognition result includes a decision value that the target live broadcast data stream belongs to a target category;
the processing unit 802 is further configured to perform decision processing on the tag display of the target live broadcast data stream according to the recognition result to obtain a decision result; and the number of the first and second groups,
the processing unit 802 is further configured to test the service model based on the decision result.
In one implementation, a live broadcast service is accessed to a micro service interface; the processing unit 802 is configured to, when acquiring a target live broadcast data stream in a live broadcast service, specifically:
and calling a micro-service interface to receive a target live broadcast data stream of the live broadcast service.
In an implementation manner, the processing unit 802 is configured to, when invoking a service model to perform recognition processing on a target live broadcast data stream to obtain a recognition result of the target live broadcast data stream, specifically:
performing framing processing on a target live broadcast data stream to obtain one or more image frames of the target live broadcast data stream;
calling a service model to identify each image frame to obtain an identification sub-result of each image frame;
and calculating to obtain the identification result of the target live broadcast data stream based on the identification sub-result of each image frame.
In one implementation, the decision result comprises a first decision result; the processing unit 802 is configured to perform decision processing on the tag display of the target live broadcast data stream according to the recognition result, and when a decision result is obtained, specifically configured to:
comparing the decision value with a decision threshold for the target class;
and if the decision value is greater than the decision threshold value, obtaining a first decision result, wherein the first decision result is used for indicating a category label of a target category displayed in a playing interface of the target live broadcast data stream.
In one implementation, the decision result comprises a second decision result; a processing unit 802, further configured to:
and if the decision value is not greater than the decision threshold value, obtaining a second decision result, wherein the second decision result is used for indicating that the category label of the target category is not displayed in the playing interface of the target live broadcast data stream.
In one implementation, the live broadcast service includes N stream segments, where N is an integer greater than 1, and the target live broadcast data stream is any one of the N stream segments; a processing unit 802, further configured to:
calling a service model to identify N-1 stream segments except the target live broadcast data stream in the N stream segments to obtain identification results of the N-1 stream segments;
and generating a recognition result curve graph based on the recognition results of the N-1 stream segments and the recognition result of the target live broadcast data stream.
In an implementation manner, the processing unit 802 is configured to, when testing the service model based on the decision result, specifically:
acquiring a reference label of a target live broadcast data stream, wherein the reference label indicates that a category label of a target category should be displayed or not displayed in a playing interface of the target live broadcast data stream;
comparing the decision result with the reference label;
if the decision result is matched with the reference label, determining that the identification performance of the service model meets the test requirement;
and if the decision result is not matched with the reference label, determining that the identification performance of the service model cannot meet the test requirement.
In one implementation, the target live broadcast data stream is a stream segment in a test set, the test set includes M stream segments and a reference label corresponding to each stream segment, and M is an integer greater than 1; the test set is used for testing the service model; the processing unit 802 is configured to, when testing the service model based on the decision result, specifically:
calling a service model to identify M-1 stream segments except the target live broadcast data stream in the M stream segments to obtain identification results of the M-1 stream segments;
performing decision processing on each stream segment in the M-1 stream segments according to the identification results of the M-1 stream segments to obtain decision results of the M-1 stream segments;
comparing the decision results of the M-1 stream segments and the decision result of the target live data stream with corresponding reference labels;
and counting the target number of the flow segments of which the decision results are matched with the reference labels, and generating a performance test result for the service model test based on the target number and the M flow segments.
In one implementation, the processing unit 802 is further configured to:
if the performance test result is greater than the test result threshold value, determining that the identification performance of the service model meets the test requirement;
and if the performance test result is not greater than the test result threshold value, determining that the identification performance of the service model does not meet the test requirement.
In one implementation, if the recognition performance of the service model does not meet the test requirement, the processing unit 802 is further configured to:
adding the stream fragments of which the decision results are not matched with the reference labels into a training set, wherein the training set is used for training the service model;
and training the business model by adopting a training set to obtain the optimized business model.
According to an embodiment of the present application, the units in the service processing apparatus shown in fig. 8 may be respectively or entirely combined into one or several other units to form the service processing apparatus, or some unit(s) therein may be further split into multiple units with smaller functions to form the service processing apparatus, which may implement the same operation without affecting implementation of technical effects of the embodiment of the present application. The units are divided based on logic functions, and in practical application, the functions of one unit can be realized by a plurality of units, or the functions of a plurality of units can be realized by one unit. In other embodiments of the present application, the service processing apparatus may also include other units, and in practical applications, these functions may also be implemented by being assisted by other units, and may be implemented by cooperation of multiple units. According to another embodiment of the present application, the business processing apparatus as shown in fig. 8 may be constructed by running a computer program (including program codes) capable of executing the steps involved in the respective methods as shown in fig. 2 and fig. 5 on a general-purpose computing device such as a computer including a central processing unit (C PU), a random access storage medium (RAM), a read-only storage medium (ROM), and the like, and a storage element, and the business processing method of the embodiment of the present application may be implemented. The computer program may be recorded on a computer-readable recording medium, for example, and loaded and executed in the above-described computing apparatus via the computer-readable recording medium.
In this embodiment of the application, the obtaining unit 801 may be configured to obtain a target live broadcast data stream in a live broadcast service, and the processing unit 802 invokes a service model to perform recognition processing and decision processing on the target live broadcast data stream, so as to test the service model according to a decision result of the decision processing. In the scheme, the target live broadcast data stream can be directly pulled from the live broadcast service for identification processing and decision processing, and the test personnel do not need to manually start broadcasting; and the service model is tested by adopting a decision result of the target live broadcast data stream, and the mode of directly pulling the target live broadcast data stream from the live broadcast service to test the service model shortens the test link length of the service model and improves the test efficiency of the service model.
Fig. 9 shows a schematic structural diagram of a service processing device according to an exemplary embodiment of the present application. Referring to fig. 9, the service processing device includes at least a processor 901, a communication interface 902 and a computer-readable storage medium 903. The processor 901, the communication interface 902, and the computer-readable storage medium 903 may be connected by a bus or other means. The communication interface 902 is used for receiving and transmitting data, among other things. A computer readable storage medium 903 may be stored in the memory of the service processing device, the computer readable storage medium 903 being used to store a computer program, the computer program comprising program instructions, the processor 901 being used to execute the program instructions stored by the computer readable storage medium 903. The processor 901 (or CPU) is a computing core and a control core of the service Processing device, and is adapted to implement one or more instructions, and specifically, adapted to load and execute the one or more instructions so as to implement a corresponding method flow or a corresponding function.
An embodiment of the present application further provides a computer-readable storage medium (Memory), which is a Memory device in a service processing device and is used to store programs and data. It is understood that the computer readable storage medium herein may include a built-in storage medium in the service processing device, and may also include an extended storage medium supported by the service processing device. The computer readable storage medium provides a memory space that stores the processing system of the business processing device. Also stored in this memory space are one or more instructions, which may be one or more computer programs (including program code), suitable for loading and execution by processor 901. It should be noted that the computer-readable storage medium may be a high-speed RAM memory, or may be a non-volatile memory (non-volatile memory), such as at least one disk memory; optionally, at least one computer readable storage medium located remotely from the aforementioned processor is also possible.
In one embodiment, the business processing device may be a graphics processor as mentioned in the previous embodiments; the computer-readable storage medium has one or more instructions stored therein; one or more instructions stored in the computer-readable storage medium are loaded and executed by the processor 901 to implement the corresponding steps in the foregoing service processing method embodiment; in particular implementations, one or more instructions in the computer-readable storage medium are loaded by the processor 901 and perform the following steps:
acquiring a target live broadcast data stream in a live broadcast service;
calling a service model to identify a target live broadcast data stream to obtain an identification result of the target live broadcast data stream, wherein the identification result comprises a decision value of the target live broadcast data stream belonging to a target category;
performing decision processing on the label display of the target live broadcast data stream according to the recognition result to obtain a decision result; and the number of the first and second groups,
and testing the service model based on the decision result.
In one implementation, a live broadcast service is accessed to a micro service interface; one or more instructions in the computer-readable storage medium are loaded by the processor 901, and when acquiring a target live data stream in a live service, the instructions are specifically configured to execute the following steps:
and calling a micro-service interface to receive a target live broadcast data stream of the live broadcast service.
In one implementation, one or more instructions in the computer-readable storage medium are loaded by the processor 901, and when a service model is called to perform recognition processing on a target live broadcast data stream to obtain a recognition result of the target live broadcast data stream, the following steps are specifically performed:
performing framing processing on a target live broadcast data stream to obtain one or more image frames of the target live broadcast data stream;
calling a service model to identify each image frame to obtain an identification sub-result of each image frame;
and calculating to obtain the identification result of the target live broadcast data stream based on the identification sub-result of each image frame.
In one implementation, the decision result comprises a first decision result; one or more instructions in the computer-readable storage medium are loaded by the processor 901, and when the decision processing is performed on the tag display of the target live data stream according to the recognition result, and the decision result is obtained, the following steps are specifically executed:
comparing the decision value with a decision threshold for the target class;
and if the decision value is greater than the decision threshold value, obtaining a first decision result, wherein the first decision result is used for indicating a category label of a target category displayed in a playing interface of the target live broadcast data stream.
In one implementation, the decision result comprises a second decision result; one or more instructions in the computer readable storage medium are loaded by processor 901 and are further configured to perform the steps of:
and if the decision value is not greater than the decision threshold value, obtaining a second decision result, wherein the second decision result is used for indicating that the category label of the target category is not displayed in the playing interface of the target live broadcast data stream.
In one implementation, the live broadcast service includes N stream segments, where N is an integer greater than 1, and the target live broadcast data stream is any one of the N stream segments; one or more instructions in the computer readable storage medium are loaded by processor 901 and are further configured to perform the steps of:
calling a service model to identify N-1 stream segments except the target live broadcast data stream in the N stream segments to obtain identification results of the N-1 stream segments;
and generating a recognition result curve graph based on the recognition results of the N-1 stream segments and the recognition result of the target live broadcast data stream.
In one implementation, one or more instructions in the computer-readable storage medium are loaded by the processor 901 and, when testing the business model based on the decision result, are specifically configured to perform the following steps:
acquiring a reference label of a target live broadcast data stream, wherein the reference label indicates that a category label of a target category should be displayed or not displayed in a playing interface of the target live broadcast data stream;
comparing the decision result with the reference label;
if the decision result is matched with the reference label, determining that the identification performance of the service model meets the test requirement;
and if the decision result is not matched with the reference label, determining that the identification performance of the service model cannot meet the test requirement.
In one implementation, the target live broadcast data stream is a stream segment in a test set, the test set includes M stream segments and a reference label corresponding to each stream segment, and M is an integer greater than 1; the test set is used for testing the service model; one or more instructions in the computer-readable storage medium are loaded by the processor 901 and, when testing the service model based on the decision result, are specifically configured to perform the following steps:
calling a service model to identify M-1 stream segments except the target live broadcast data stream in the M stream segments to obtain identification results of the M-1 stream segments;
performing decision processing on each stream segment in the M-1 stream segments according to the identification results of the M-1 stream segments to obtain decision results of the M-1 stream segments;
comparing the decision results of the M-1 stream segments and the decision result of the target live data stream with corresponding reference labels;
and counting the target number of the flow segments of which the decision results are matched with the reference labels, and generating a performance test result for the service model test based on the target number and the M flow segments.
In one implementation, one or more instructions in the computer-readable storage medium are loaded by the processor 901 and are further configured to perform the steps of:
if the performance test result is greater than the test result threshold value, determining that the identification performance of the service model meets the test requirement;
and if the performance test result is not greater than the test result threshold value, determining that the identification performance of the service model does not meet the test requirement.
In one implementation, if the recognition performance of the business model does not meet the testing requirements, one or more instructions in the computer readable storage medium are loaded by the processor 901 and are further configured to perform the steps of:
adding the stream fragments of which the decision results are not matched with the reference labels into a training set, wherein the training set is used for training the service model;
and training the business model by adopting a training set to obtain the optimized business model.
In this embodiment of the application, the communication interface 902 may obtain a target live broadcast data stream in a live broadcast service, and the processor 901 calls a service model to perform recognition processing and decision processing on the target live broadcast data stream, so as to test the service model according to a decision result of the decision processing. In the scheme, the target live broadcast data stream can be directly pulled from the live broadcast service for identification processing and decision processing, and the test personnel do not need to manually start broadcasting; and the service model is tested by adopting a decision result of the target live broadcast data stream, and the mode of directly pulling the target live broadcast data stream from the live broadcast service to test the service model shortens the test link length of the service model and improves the test efficiency of the service model.
Embodiments of the present application also provide a computer program product or a computer program comprising computer instructions stored in a computer-readable storage medium. The processor of the service processing device reads the computer instruction from the computer readable storage medium, and the processor executes the computer instruction, so that the service processing device executes the service processing method.
Those of ordinary skill in the art would appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. The procedures or functions described in accordance with the embodiments of the invention are all or partially effected when the computer program instructions are loaded and executed on a computer. The computer may be a general purpose computer, a special purpose computer, a network of computers, or other programmable device. The computer instructions may be stored on or transmitted over a computer-readable storage medium. The computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center by wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device, such as a server, a data center, etc., that incorporates one or more of the available media. The available media may be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), or semiconductor media (e.g., Solid State Disks (SSDs)), among others.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present disclosure, and all the changes or substitutions should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (13)

1. A method for processing a service, comprising:
acquiring a target live broadcast data stream in a live broadcast service;
calling a service model to identify the target live broadcast data stream to obtain an identification result of the target live broadcast data stream, wherein the identification result comprises a decision value of the target live broadcast data stream belonging to a target category;
performing decision processing on the label display of the target live broadcast data stream according to the identification result to obtain a decision result; and the number of the first and second groups,
and testing the service model based on the decision result.
2. The method of claim 1, wherein the live service accesses a micro-service interface; the acquiring of the target live broadcast data stream in the live broadcast service includes:
and calling the micro-service interface to receive the target live broadcast data stream of the live broadcast service.
3. The method of claim 1, wherein the invoking the service model to perform recognition processing on the target live data stream to obtain a recognition result of the target live data stream comprises:
performing framing processing on the target live broadcast data stream to obtain one or more image frames of the target live broadcast data stream;
calling the service model to identify each image frame to obtain an identification sub-result of each image frame;
and calculating the identification result of the target live broadcast data stream based on the identification sub-result of each image frame.
4. The method of claim 3, wherein the decision result comprises a first decision result; the step of performing decision processing on the tag display of the target live broadcast data stream according to the identification result to obtain a decision result comprises the following steps:
comparing the decision value to a decision threshold for the target class;
and if the decision value is larger than the decision threshold value, obtaining the first decision result, wherein the first decision result is used for indicating that the category label of the target category is displayed in a playing interface of the target live broadcast data stream.
5. The method of claim 4, wherein the decision result comprises a second decision result; the method further comprises the following steps:
and if the decision value is not greater than the decision threshold value, obtaining a second decision result, wherein the second decision result is used for indicating that the category label of the target category is not displayed in a playing interface of the target live broadcast data stream.
6. The method of claim 1, wherein the live service includes N stream segments, N being an integer greater than 1, the target live data stream being any one of the N stream segments; the method further comprises the following steps:
calling the service model to identify N-1 stream segments of the N stream segments except the target live broadcast data stream to obtain identification results of the N-1 stream segments;
and generating a recognition result curve graph based on the recognition results of the N-1 stream segments and the recognition result of the target live data stream.
7. The method of claim 1, wherein the testing the business model based on the decision result comprises:
acquiring a reference label of the target live broadcast data stream, wherein the reference label indicates that a category label of the target category should be displayed or not displayed in a playing interface of the target live broadcast data stream;
comparing the decision result with the reference label;
if the decision result is matched with the reference label, determining that the identification performance of the service model meets the test requirement;
and if the decision result is not matched with the reference label, determining that the identification performance of the service model cannot meet the test requirement.
8. The method of claim 1, wherein the target live data stream is one stream segment in a test set, the test set includes M stream segments and a reference label corresponding to each stream segment, M is an integer greater than 1; the test set is used for testing the service model; the testing the service model based on the decision result comprises:
calling the service model to identify M-1 stream segments except the target live broadcast data stream in the M stream segments to obtain identification results of the M-1 stream segments;
performing decision processing on each stream segment in the M-1 stream segments according to the identification results of the M-1 stream segments to obtain decision results of the M-1 stream segments;
comparing the decision results of the M-1 stream segments and the decision result of the target live data stream with corresponding reference labels;
and counting the target number of the flow segments with the decision results matched with the reference labels, and generating performance test results for the service model test based on the target number and the M flow segments.
9. The method of claim 8, wherein the method further comprises:
if the performance test result is larger than the test result threshold value, determining that the identification performance of the service model meets the test requirement;
and if the performance test result is not greater than the test result threshold value, determining that the identification performance of the service model does not meet the test requirement.
10. The method of any of claims 7 or 8, wherein if the recognition performance of the business model does not meet the test requirements, the method further comprises:
adding the stream fragments of which the decision results are not matched with the reference labels to a training set, wherein the training set is used for training the business model;
and training the business model by adopting the training set to obtain the optimized business model.
11. A traffic processing apparatus, comprising:
the acquisition unit is used for acquiring a target live broadcast data stream in a live broadcast service;
the processing unit is used for calling a service model to identify the target live broadcast data stream to obtain an identification result of the target live broadcast data stream, wherein the identification result comprises a decision value of the target live broadcast data stream belonging to a target category;
the processing unit is further configured to perform decision processing on the tag display of the target live broadcast data stream according to the identification result to obtain a decision result; and the number of the first and second groups,
the processing unit is further configured to test the service model based on the decision result.
12. A traffic processing device, comprising:
a processor adapted to execute a computer program;
computer readable storage medium, in which a computer program is stored which, when executed by the processor, implements a business process method according to any one of claims 1-10.
13. A computer-readable storage medium, characterized in that it stores a computer program adapted to be loaded by a processor and to execute the traffic processing method according to any of claims 1-10.
CN202110780207.5A 2021-07-09 2021-07-09 Service processing method, device, equipment and medium Pending CN113824950A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110780207.5A CN113824950A (en) 2021-07-09 2021-07-09 Service processing method, device, equipment and medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110780207.5A CN113824950A (en) 2021-07-09 2021-07-09 Service processing method, device, equipment and medium

Publications (1)

Publication Number Publication Date
CN113824950A true CN113824950A (en) 2021-12-21

Family

ID=78924146

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110780207.5A Pending CN113824950A (en) 2021-07-09 2021-07-09 Service processing method, device, equipment and medium

Country Status (1)

Country Link
CN (1) CN113824950A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115242880A (en) * 2022-07-14 2022-10-25 湖南三湘银行股份有限公司 Micro-service framework access method based on network request bridging

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115242880A (en) * 2022-07-14 2022-10-25 湖南三湘银行股份有限公司 Micro-service framework access method based on network request bridging

Similar Documents

Publication Publication Date Title
EP3855753B1 (en) Method and apparatus for locating video playing node, device and storage medium
WO2022022152A1 (en) Video clip positioning method and apparatus, and computer device and storage medium
CN110784759B (en) Bullet screen information processing method and device, electronic equipment and storage medium
CN110390033B (en) Training method and device for image classification model, electronic equipment and storage medium
CN110837579A (en) Video classification method, device, computer and readable storage medium
CN110012356A (en) Video recommendation method, device and equipment and computer storage medium
CN108989882B (en) Method and apparatus for outputting music pieces in video
CN109121022B (en) Method and apparatus for marking video segments
CN109145828B (en) Method and apparatus for generating video category detection model
CN103299324A (en) Learning tags for video annotation using latent subtags
CN111798879A (en) Method and apparatus for generating video
CN111611436A (en) Label data processing method and device and computer readable storage medium
CN110347866B (en) Information processing method, information processing device, storage medium and electronic equipment
CN112995690B (en) Live content category identification method, device, electronic equipment and readable storage medium
CN108062377A (en) The foundation of label picture collection, definite method, apparatus, equipment and the medium of label
CN113069769B (en) Cloud game interface display method and device, electronic equipment and storage medium
KR20210087001A (en) Image auditing method and device, electronic equipment and storage medium
CN110674349B (en) Video POI (Point of interest) identification method and device and electronic equipment
CN113515997B (en) Video data processing method and device and readable storage medium
CN113962965B (en) Image quality evaluation method, device, equipment and storage medium
CN113515998A (en) Video data processing method and device and readable storage medium
CN113515669A (en) Data processing method based on artificial intelligence and related equipment
CN111597446B (en) Content pushing method and device based on artificial intelligence, server and storage medium
CN113901263B (en) Label generation method and device for video material
CN109816023B (en) Method and device for generating picture label model

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination