CN108419091A - A kind of verifying video content method and device based on machine learning - Google Patents
A kind of verifying video content method and device based on machine learning Download PDFInfo
- Publication number
- CN108419091A CN108419091A CN201810174046.3A CN201810174046A CN108419091A CN 108419091 A CN108419091 A CN 108419091A CN 201810174046 A CN201810174046 A CN 201810174046A CN 108419091 A CN108419091 A CN 108419091A
- Authority
- CN
- China
- Prior art keywords
- violation content
- video
- recognition
- violation
- audio
- 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
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/234—Processing of video elementary streams, e.g. splicing of video streams, manipulating MPEG-4 scene graphs
- H04N21/23418—Processing of video elementary streams, e.g. splicing of video streams, manipulating MPEG-4 scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/233—Processing of audio elementary streams
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/23—Processing of content or additional data; Elementary server operations; Server middleware
- H04N21/235—Processing of additional data, e.g. scrambling of additional data or processing content descriptors
Abstract
The present invention proposes that a kind of verifying video content method and device based on machine learning, this method include：Extract video information, audio-frequency information and the text message in pending video resource；Violation content recognition processing is carried out to the video information by trained convolutional neural networks, and violation content recognition processing is carried out to the audio-frequency information and the text message by trained Recognition with Recurrent Neural Network respectively, judge in the video information, the audio-frequency information and the text message whether to include violation content；If including violation content in the video information, the audio-frequency information or the text message, the violation content type for including in the video information, the audio-frequency information or the text message is confirmed and exported.Above-mentioned technical proposal realizes the violation content for including in automatic identification video resource, can substitute manual examination and verification video content, improves verifying video content efficiency.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of verifying video content methods based on machine learning
And device.
Background technology
With the development of Internet technology, internet video flow is increased significantly in recent years, short-sighted frequency, net cast etc.
Various novel user original contents (User Generated Content, UGC) promote internet video more and more abundant.Largely
Video resource propagated in the channels such as internet, mobile radio communication, broadcasting and TV wired network.
The Devoting Major Efforts To Developing of the development of internet and internet video application facilitates user by video transmission information, simultaneously
Be also convenient for criminal and disseminated the violation video resource comprising flame, for example, be related to racial discrimination, political sensitivity topic,
The video resource of the information such as violence, pornographic, terror.Therefore, internet video rapid development background under, to video content into
Row audit is the requisite measure for safeguarding good internet environment.Conventional method usually relies on manually is uploaded to internet to audit
Whether the content of video resource includes violation content.Since internet video stock number is increasing, by manually to regarding
Frequency content, which carries out audit, must consume a large amount of human costs, and less efficient.
Invention content
Defect based on the above-mentioned prior art and deficiency, the present invention propose a kind of verifying video content based on machine learning
Method and device may be implemented automatically to audit video content.
In order to achieve the above object, the following technical solutions are proposed by the present invention：
A kind of verifying video content method based on machine learning, including：
Extract video information, audio-frequency information and the text message in pending video resource；
Violation content recognition processing is carried out to the video information by trained convolutional neural networks, and is passed through
Trained Recognition with Recurrent Neural Network carries out violation content recognition processing to the audio-frequency information and the text message respectively, sentences
Whether include violation content in the disconnected video information, the audio-frequency information and the text message；
If including violation content in the video information, the audio-frequency information or the text message, confirm and defeated
Go out the violation content type for including in the video information, the audio-frequency information or the text message.
Preferably, this method further includes：
When it includes violation content to confirm in the video information, the audio-frequency information or the text message, confirm simultaneously
Export time location of the violation content in the pending video resource.
Preferably, violation content recognition processing is carried out to the video information by trained convolutional neural networks,
Judge in the video information whether to include violation content, including：
The video information is divided into multiple video frame；
Violation content recognition processing is carried out to the multiple video frame respectively by trained convolutional neural networks, point
Do not judge in each video frame whether to include violation content；
If including violation content in any one video frame, confirm in the video information to include violation content.
Preferably, to the training process of the convolutional neural networks, including：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The image for having marked violation content type is inputted into convolutional neural networks, makes the convolutional neural networks according to pre-
If parameter identification described image in violation content；
The violation content type that the convolutional neural networks are identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the convolutional neural networks；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the convolution god is adjusted
Parameter through the violation content in Network Recognition image.
Preferably, to the training process of the Recognition with Recurrent Neural Network, including：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The audio and/or text input Recognition with Recurrent Neural Network of violation content type will be marked, has made the cycle nerve
Network identifies the violation content in the audio and/or text according to preset parameter；
The violation content type that the Recognition with Recurrent Neural Network is identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the Recognition with Recurrent Neural Network；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the cycle god is adjusted
Parameter through the violation content in Network Recognition audio and/or text.
A kind of verifying video content device based on machine learning, including：
Information extraction unit, for extracting video information, audio-frequency information and text message in pending video resource；
Content recognition unit, for carrying out violation content to the video information by trained convolutional neural networks
Identifying processing, and the audio-frequency information and the text message are disobeyed respectively by trained Recognition with Recurrent Neural Network
Content recognition processing is advised, judges in the video information, the audio-frequency information and the text message whether to include violation content；
Output unit includes violation content for working as in the video information, the audio-frequency information or the text message
When, confirm and export the violation content type for including in the video information, the audio-frequency information or the text message.
Preferably, the output unit is additionally operable to：
When it includes violation content to confirm in the video information, the audio-frequency information or the text message, confirm simultaneously
Export time location of the violation content in the pending video resource.
Preferably, the content recognition unit disobeys the video information by trained convolutional neural networks
Content recognition processing is advised, when whether judging in the video information comprising violation content, is specifically used for：
The video information is divided into multiple video frame；
Violation content recognition processing is carried out to the multiple video frame respectively by trained convolutional neural networks, point
Do not judge in each video frame whether to include violation content；
If including violation content in any one video frame, confirm in the video information to include violation content.
Preferably, the content recognition unit is additionally operable to be trained the convolutional neural networks；
The content recognition unit when being trained, is specifically used for the convolutional neural networks：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The image for having marked violation content type is inputted into convolutional neural networks, makes the convolutional neural networks according to pre-
If parameter identification described image in violation content；
The violation content type that the convolutional neural networks are identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the convolutional neural networks；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the convolution god is adjusted
Parameter through the violation content in Network Recognition image.
Preferably, the content recognition unit is additionally operable to be trained the Recognition with Recurrent Neural Network；
The content recognition unit when being trained, is specifically used for the Recognition with Recurrent Neural Network：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The audio and/or text input Recognition with Recurrent Neural Network of violation content type will be marked, has made the cycle nerve
Network identifies the violation content in the audio and/or text according to preset parameter；
The violation content type that the Recognition with Recurrent Neural Network is identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the Recognition with Recurrent Neural Network；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the cycle god is adjusted
Parameter through the violation content in Network Recognition audio and/or text.
Verifying video content method proposed by the present invention extracts video information, audio-frequency information from pending video resource
And text message, violation content recognition processing then is carried out to video information by trained convolutional neural networks, is passed through
Trained Recognition with Recurrent Neural Network carries out violation content recognition processing to audio-frequency information and text message, judge video information,
Whether include violation content in audio-frequency information and text message, if including violation content, confirms and export violation content class
Type.Above-mentioned technical proposal realizes the violation content for including in automatic identification video resource, can substitute in manual examination and verification video
Hold, improves verifying video content efficiency.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is a kind of flow signal of verifying video content method based on machine learning provided in an embodiment of the present invention
Figure；
Fig. 2 is the structural schematic diagram of cyclic convolution neural network provided in an embodiment of the present invention；
Fig. 3 is the flow diagram provided in an embodiment of the present invention that violation content auditing is carried out to video information；
Fig. 4 is the flow diagram provided in an embodiment of the present invention being trained to convolutional neural networks；
Fig. 5 is a kind of structural representation of verifying video content device based on machine learning provided in an embodiment of the present invention
Figure.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a kind of verifying video content method based on machine learning, shown in Figure 1, the party
Method includes：
Video information, audio-frequency information and text message in the pending video resource of S101, extraction；
Specifically, above-mentioned pending video resource, can be any type of video resource being present in internet,
Such as traditional film, TV, variety video etc., it can also be novel video resource, such as live video, small video, short-sighted
Frequency etc..Above-mentioned pending video resource can refer to that user is uploaded to the video money mutually propagated between internet or user
Source may also mean that the video resource that users from networks obtains.Theoretically, the video resource of Any Digit form, can be by
It is set as pending video resource, content auditing can be carried out to it using technical solution of the embodiment of the present invention.
Video resource is usually to be made of three kinds of video, audio and text media resources, and the present invention is to pending
When video resource carries out violation content auditing, the video that includes to video resource respectively, three kinds of media resources of audio and text into
Row audit, to judge whether include violation content in video resource.
Parsing is decoded to pending video resource, can therefrom extract the video information for including, audio-frequency information
And text message.It should be noted that the video information extracted is only to include the data information of video content, audio-frequency information is
Only include the data information of audio content, and text message is exactly the data information for only including content of text.It is another optional
Mode is made of a series of video frame being connected according to time order and function due to video resource, is carried out to video resource
Parsing can obtain a series of video frame, audio frame and text frame, that is to say, that in a frame of video resource, simultaneously
Including video information, audio-frequency information and text message.When carrying out parsing extraction to pending video resource in this way, it can extract
A series of video frame, a series of audio frame and a series of text frame are obtained, it then can be to above-mentioned a series of video
Frame, a series of audio frame and a series of text frame are audited respectively.It should be understood that if not carried out to video resource
Frame divides, then can directly extract continuous video, audio and text from video resource, can be used for auditing.
S102, violation content recognition processing is carried out to the video information by trained convolutional neural networks, with
And violation content recognition is carried out to the audio-frequency information and the text message by trained Recognition with Recurrent Neural Network respectively
Whether processing judges in the video information, the audio-frequency information and the text message to include violation content；
Specifically, above-mentioned convolutional neural networks and Recognition with Recurrent Neural Network, are all to belong to the common mathematical model of machine learning,
Such mathematical model has learning ability, can learn the data structure of input data and inherent pattern, constantly adjust network ginseng
Number, to improve the confidence level of the efficiency and recognition result of study.
Convolutional neural networks add feature learning part on the basis of multilayer neural network, this part can imitate people
Brain is to the classification on signal processing.Concrete operations are exactly to add different volumes before the layer of multilayer neural network connected entirely
Lamination and pond layer, constitute input layer-convolutional layer-pond layer-convolutional layer-pond layer-...-hidden layer-output layer.It is simple next
It says, the work that multilayer neural network is done is Feature Mapping to value, is characterized in hand picking.And depth machine learning model
The work that convolutional neural networks are done is that signal is mapped to feature again by Feature Mapping to value, is characterized in being selected by network oneself.
Convolutional neural networks essence is the neural network of a multilayer, and basic operation includes successively：Convolution algorithm, Chi Hua
Operation, full connection operation and identification operation.
Convolution algorithm：The characteristic pattern of preceding layer and a convolution kernel that can learn carry out convolution algorithm, the result warp of convolution
The output crossed after activation primitive forms the neuron of later layer, to constitute this layer of characteristic pattern, also referred to as feature extraction layer, each god
Input through member is connected with the local experiences of preceding layer, and extracts the feature of the part, once the local feature is extracted, it
Position relationship between other feature is just determined.
Pond operation：Input signal is divided into nonoverlapping region by it, passes through pond (down-sampling) for each region
Operation reduces the spatial resolution of network, for example maximum value pond is the maximum value in selection region, and mean value pond is to calculate
Average value in region.The offset and distortion of signal are eliminated by the operation.
Full connection operation：Input signal exports as multigroup signal, after the pond operation of multiple convolution core by connecting entirely
Multigroup signal is combined as one group of signal by operation successively.
Identify operation：Above-mentioned calculating process is characterized study operation, (need to be divided according to business demand on above-mentioned operating basis
Class or regression problem) increase by a layer network for classifying or returning calculating.
Convolutional neural networks are inherently a kind of mapping being input to output, it can learn largely to input and export
Between mapping relations, without the accurate mathematic(al) representation between any output and input, as long as with known pattern
Convolutional network is trained, network is just with the mapping ability between inputoutput pair.What convolutional neural networks executed is to have
Supervised training, thus its sample set be by shaped like：The vector of (input signal, label value) is to composition.
Fig. 2 show the structural schematic diagram of above-mentioned Recognition with Recurrent Neural Network, is used in the loop body of Recognition with Recurrent Neural Network
The neural network structure of one similar full articulamentum.What the state in Recognition with Recurrent Neural Network was indicated by a vector, this
The dimension of vector is the size of Recognition with Recurrent Neural Network hidden layer, it is assumed that it is h.The input x of neural network in loop body has two
Part, a part are the state of last moment, and another part is the input sample at current time.It is defeated for language model
It can be that the corresponding word of current word is vectorial (word embedding) to enter sample.Recognition with Recurrent Neural Network has " Memorability ", institute
Recognition with Recurrent Neural Network is used in embodiments of the present invention in natural language processing (NLP) and voice field with mainly applying
Audio and text are handled.
It should be noted that, the present invention different with the optimal suitable application area of Recognition with Recurrent Neural Network based on convolutional neural networks
Convolutional neural networks for being audited to video information, Recognition with Recurrent Neural Network are used for audio-frequency information and text by embodiment
Information is audited.Again since there is learning ability, suitable application area to have for convolutional neural networks and Recognition with Recurrent Neural Network again
There is versatility, therefore convolutional neural networks or Recognition with Recurrent Neural Network can also be used simultaneously to video information, audio-frequency information and text
This information is audited, i.e., using same network carry out video information, audio-frequency information and text message violation content auditing,
Or any other available deep learning algorithm realization can also be used to disobey video information, audio-frequency information and text message
Advise content auditing.
Before the above-mentioned convolutional neural networks of application and Recognition with Recurrent Neural Network, the embodiment of the present invention is first to above-mentioned convolutional Neural
Network and Recognition with Recurrent Neural Network are trained, and above-mentioned convolutional neural networks is made to have the violation content in the video of identification input,
And the ability for identifying violation content type, the audio for making above-mentioned Recognition with Recurrent Neural Network have identification input or the violation in text
Content, and identify the ability of violation content type.Then, in the video information input for being included by pending video resource
Trained convolutional neural networks are stated, can identify in the video information of input whether include violation content；It will be pending
The audio-frequency information and text message that video resource is included input above-mentioned trained Recognition with Recurrent Neural Network, can identify input
Audio-frequency information and text message in whether include violation content.
If including violation content in the video information, the audio-frequency information or the text message, then follow the steps
S103, confirmation simultaneously export the violation content type for including in the video information, the audio-frequency information or the text message.
Specifically, when above-mentioned convolutional neural networks recognize violation video content, Huo Zheshang from the video information of input
When stating Recognition with Recurrent Neural Network and recognizing violation audio content or violation content of text from the audio-frequency information or text message of input,
Then can be confirmed in above-mentioned pending video resource to include violation content.
Further, convolutional neural networks and Recognition with Recurrent Neural Network used in the embodiment of the present invention to video information,
When audio-frequency information and text message carry out violation content recognition, can not only identify whether to include violation content, can also identify
The type of violation content, such as relate to the types such as political affairs, violence, pornographic, terror.Therefore, when above-mentioned convolutional neural networks are from input
Violation video content or above-mentioned Recognition with Recurrent Neural Network are recognized in video information from the audio-frequency information or text message of input
When recognizing violation audio content or violation content of text, the type of included violation content is also further confirmed that.
Then, setting of the embodiment of the present invention exports recognized violation content type.It should be noted that the present invention is real
Apply the violation content type that example includes in video information, audio-frequency information or the text message for exporting pending video resource
When, in a manner of enumerating, all violation content types recognized are sequentially output, i.e., enumerates and knows from video resource respectively
Each the violation content type being clipped to.
By above-mentioned introduction as it can be seen that the verifying video content method that the embodiment of the present invention is provided is provided from pending video
Video information, audio-frequency information and text message are extracted in source, then by trained convolutional neural networks to video information
Violation content recognition processing is carried out, audio-frequency information and text message are carried out in violation by trained Recognition with Recurrent Neural Network
Hold identifying processing, judges in video information, audio-frequency information and text message whether to include violation content, if including in violation
Hold, then confirms and export violation content type.Above-mentioned technical proposal realizes in the violation for including in automatic identification video resource
Hold, manual examination and verification video content can be substituted, improves verifying video content efficiency.
Optionally, in another embodiment of the present invention, this method further includes：
When it includes violation content to confirm in the video information, the audio-frequency information or the text message, confirm simultaneously
Export time location of the violation content in the pending video resource.
Specifically, above-mentioned pending video resource is by a series of video frame arranged according to chronological order, sound
What frequency frame and text frame formed.Therefore, according to video resource duration can determine each video frame, audio frame and text frame when
Between position.When identifying violation content from the video information of above-mentioned pending video resource, audio-frequency information and text message,
Violation content is identified actually from the video frame of above-mentioned pending video resource, audio frame and text frame.Further,
It, can be according to regarding where the violation content identified when recognizing violation content from video frame, audio frame or text frame
Frequency frame, audio frame or the time location of text frame determine the violation content identified in above-mentioned pending video resource
Time location.It is appreciated that time location of the violation content in above-mentioned pending video resource, is exactly violation content place
Video resource where it of video frame, audio frame or text frame in time location.
Based on above-mentioned theory, the embodiment of the present invention is passing through trained convolutional neural networks and Recognition with Recurrent Neural Network pair
The video frame of pending video resource, audio frame and text frame carry out violation content recognition, and when identifying violation content, also
It further determines that time location of the violation content in above-mentioned pending video resource, is further checked in violation for auditor
Hold and facility is provided.
Optionally, in another embodiment of the present invention, shown in Figure 3, pass through trained convolutional Neural net
Whether network carries out violation content recognition processing to the video information, judge in the video information to include violation content, specifically
Including：
S301, the video information is divided into multiple video frame；
Specifically, the video information in video resource, is combined by multiple video frame arranged according to chronological order
It forms, therefore be decoded parsing to video information to obtain multiple video frame that video information is included.It is appreciated that every
A video frame is all piece image.
S302, the multiple video frame is carried out at violation content recognition respectively by trained convolutional neural networks
Whether reason judges in each video frame to include violation content respectively；
Specifically, multiple video frame that video information divides are sequentially input into trained convolutional neural networks,
So that convolutional neural networks is carried out violation content recognition processing respectively to each video frame, whether judges in each video frame comprising separated
Advise content.
If including violation content in any one video frame, S303 is thened follow the steps, confirms and is wrapped in the video information
Content containing violation.
Specifically, if in any one video frame in multiple video frame of video information including violation content,
Can determine in the video information to include violation content.
Further, it when convolutional neural networks recognize violation content from a certain video frame, can also further identify
Then violation content type exports recognized violation video content types.
In known video information duration, and then determine time of each video frame of video information in above-mentioned video information
When position, the violation identified can also be exported by confirming the time location of the video frame where the violation content that identifies
Time location of the content in video information, i.e. time location of the video frame where it in video information.
The processing procedure of violation content recognition, Recognition with Recurrent Neural Network are carried out to video information according to above-mentioned convolutional neural networks
The processing procedure that violation content recognition is carried out to audio-frequency information and text message is referred to above-mentioned processing procedure execution.
Optionally, in another embodiment of the present invention, shown in Figure 4, to the training process of convolutional neural networks,
It specifically includes：
S401, the image for having marked violation content type is inputted into convolutional neural networks, makes the convolutional neural networks
The violation content in described image is identified according to preset parameter；
Specifically, the above-mentioned image for having marked violation content type, refers to being included in advance by manually accurately marking image
Violation content type image.For example, the figure of the violations content type such as mark " relating to political affairs ", " pornographic ", " violence ", " terror "
Picture.
After putting up convolutional neural networks, network parameter is initialized, has then been marked above-mentioned in violation
The image for holding type inputs the convolutional neural networks, make the convolutional neural networks according to initialization parameter to the image of input into
Row violation content recognition, the violation content type identified.
S402, the violation content type for identifying the convolutional neural networks carry out pair with the violation content type of mark
Than obtaining the cost value that the convolutional neural networks carry out violation content recognition；
Specifically, above-mentioned convolutional neural networks carry out the cost value of violation content recognition, convolutional neural networks pair are illustrated
The image of input carries out the error of violation content recognition, can simply be interpreted as the violation content type and mark of neural network recognization
The difference of the violation content type of note.
The embodiment of the present invention can be disobeyed by calculating above-mentioned convolutional neural networks to what image progress violation content recognition obtained
Content type is advised at a distance from the violation content type of mark, the generation of violation content recognition is carried out as above-mentioned convolutional neural networks
Value.
The method for the above-mentioned calculating cost value that the embodiment of the present invention proposes, only a kind of optional computational methods are not
Unique computational methods.Above-mentioned convolutional neural networks actually can arbitrarily be calculated and carry out what violation content recognition obtained
The computational methods of the difference degree of the violation content type of violation content type and mark, can be adopted by the embodiment of the present invention
With.For example, can set above-mentioned cost value to Boolean, cost value is that " 1 " then illustrates that above-mentioned convolutional neural networks are disobeyed
The violation content type that rule content recognition obtains is different from the violation content type of mark, and cost value is that " 0 " then illustrates above-mentioned volume
It is identical as the violation content type of mark that product neural network carries out the violation content type that violation content recognition obtains.
Whether the cost value that S403, judgement obtain is less than given threshold；
Specifically, only when above-mentioned convolutional neural networks carry out the violation content type and mark that violation content recognition obtains
Violation content type difference it is sufficiently small, that is, when the above-mentioned cost value being calculated is sufficiently small, just illustrate above-mentioned convolution
The precision that neural network carries out violation content recognition reaches sets requirement.
In embodiments of the present invention, by judge above-mentioned convolutional neural networks carry out violation content recognition cost value whether
Less than given threshold, to judge whether the precision of above-mentioned convolutional neural networks progress violation content recognition reaches requirement.
If the cost value that above-mentioned convolutional neural networks carry out violation content recognition is less than given threshold, illustrate above-mentioned volume
The precision that product neural network carries out violation content recognition reaches requirement, can be used for carrying out violation content recognition to video frame.Such as
The cost value that the above-mentioned convolutional neural networks of fruit carry out violation content recognition is not less than given threshold, then illustrates above-mentioned convolutional Neural net
The precision that network carries out violation content recognition also not up to requires.
If obtained cost value is not less than given threshold, thens follow the steps S404, according to the cost value being calculated, adjust
The parameter of violation content in the whole convolutional neural networks identification image.
Specifically, if the cost value that above-mentioned convolutional neural networks carry out violation content recognition is not less than given threshold,
Illustrate that above-mentioned convolutional neural networks carry out the precision of violation content recognition and also not up to require, at this point, the embodiment of the present invention according to
Convolutional neural networks carry out the cost value of violation content recognition, and the parameter that convolutional neural networks are carried out with violation content recognition carries out
Adjustment, so that convolutional neural networks carry out the precision higher of violation content recognition.
After the parameter for adjusting above-mentioned convolutional neural networks progress violation content recognition, return to step S401 will have been marked again
It notes violation content type and inputs convolutional neural networks, make image of the convolutional neural networks according to the parameter after adjustment to input
Violation content recognition is carried out, and the parameter of convolutional neural networks is adjusted according to recognition result.It is, cycle executes
Step S401~S404 is stated, until the cost value that is calculated is less than given threshold, at this time to the instruction of above-mentioned convolutional neural networks
Practice and completes.
Similarly, to the training process of above-mentioned Recognition with Recurrent Neural Network, it is referred to the instruction shown in Fig. 4 to convolutional neural networks
Practice process to execute：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The audio and/or text input Recognition with Recurrent Neural Network of violation content type will be marked, has made the cycle nerve
Network identifies the violation content in the audio and/or text according to preset parameter；
The violation content type that the Recognition with Recurrent Neural Network is identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the Recognition with Recurrent Neural Network；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the cycle god is adjusted
Parameter through the violation content in Network Recognition audio and/or text.
The specific implementation procedure of above-mentioned processing procedure, reference can be made to the training process shown in Fig. 4 to convolutional neural networks is held
Row, details are not described herein again.
Another embodiment of the present invention also discloses a kind of verifying video content device based on machine learning, referring to Fig. 5 institutes
Show, which includes：
Information extraction unit 100, for extracting video information, audio-frequency information and text envelope in pending video resource
Breath；
Content recognition unit 110, for being carried out in violation of rules and regulations to the video information by trained convolutional neural networks
Content recognition processing, and by trained Recognition with Recurrent Neural Network respectively to the audio-frequency information and the text message into
Whether the processing of row violation content recognition judges in the video information, the audio-frequency information and the text message comprising in violation of rules and regulations
Content；
Output unit 120, for working as in the video information, the audio-frequency information or the text message comprising in violation
Rong Shi confirms and exports the violation content type for including in the video information, the audio-frequency information or the text message.
Optionally, in another embodiment of the present invention, the output unit 120 is additionally operable to：
When it includes violation content to confirm in the video information, the audio-frequency information or the text message, confirm simultaneously
Export time location of the violation content in the pending video resource.
Optionally, in another embodiment of the present invention, the content recognition unit 110 passes through trained convolution
Whether neural network carries out violation content recognition processing to the video information, judge in the video information to include violation content
When, it is specifically used for：
The video information is divided into multiple video frame；
Violation content recognition processing is carried out to the multiple video frame respectively by trained convolutional neural networks, point
Do not judge in each video frame whether to include violation content；
If including violation content in any one video frame, confirm in the video information to include violation content.
Optionally, in another embodiment of the present invention, the content recognition unit 110 is additionally operable to convolution god
It is trained through network；
The content recognition unit when being trained, is specifically used for the convolutional neural networks：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The image for having marked violation content type is inputted into convolutional neural networks, makes the convolutional neural networks according to pre-
If parameter identification described image in violation content；
The violation content type that the convolutional neural networks are identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the convolutional neural networks；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the convolution god is adjusted
Parameter through the violation content in Network Recognition image.
Optionally, in another embodiment of the present invention, the content recognition unit 110 is additionally operable to cycle god
It is trained through network；
The content recognition unit when being trained, is specifically used for the Recognition with Recurrent Neural Network：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The audio and/or text input Recognition with Recurrent Neural Network of violation content type will be marked, has made the cycle nerve
Network identifies the violation content in the audio and/or text according to preset parameter；
The violation content type that the Recognition with Recurrent Neural Network is identified is compared with the violation content type of mark, is obtained
The cost value of violation content recognition is carried out to the Recognition with Recurrent Neural Network；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the cycle god is adjusted
Parameter through the violation content in Network Recognition audio and/or text.
Specifically, each unit of the verifying video content device based on machine learning disclosed in the various embodiments described above
Specific works content refers to the content of above method embodiment, and details are not described herein again.
Finally, it is to be noted that, herein, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment including a series of elements includes not only that
A little elements, but also include other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Each embodiment is described by the way of progressive in this specification, the highlights of each of the examples are with other
The difference of embodiment, just to refer each other for identical similar portion between each embodiment.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest range caused.
Claims (10)
1. a kind of verifying video content method based on machine learning, which is characterized in that including：
Extract video information, audio-frequency information and the text message in pending video resource；
By trained convolutional neural networks to the video information carry out violation content recognition processing, and pass through by
Trained Recognition with Recurrent Neural Network carries out violation content recognition processing to the audio-frequency information and the text message respectively, judges institute
Whether state in video information, the audio-frequency information and the text message includes violation content；
If including violation content in the video information, the audio-frequency information or the text message, confirms and export institute
State the violation content type for including in video information, the audio-frequency information or the text message.
2. according to the method described in claim 1, it is characterized in that, this method further includes：
When it includes violation content to confirm in the video information, the audio-frequency information or the text message, confirms and export
Time location of the violation content in the pending video resource.
3. according to the method described in claim 1, it is characterized in that, by trained convolutional neural networks to the video
Whether information carries out violation content recognition processing, judge in the video information to include violation content, including：
The video information is divided into multiple video frame；
Violation content recognition processing is carried out to the multiple video frame respectively by trained convolutional neural networks, is sentenced respectively
Whether include violation content in each video frame of breaking；
If including violation content in any one video frame, confirm in the video information to include violation content.
4. the method according to any claim in claims 1 to 3, which is characterized in that the convolutional neural networks
Training process, including：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The image for having marked violation content type is inputted into convolutional neural networks, makes the convolutional neural networks according to preset
Parameter identifies the violation content in described image；
The violation content type that the convolutional neural networks are identified is compared with the violation content type of mark, obtains institute
State the cost value that convolutional neural networks carry out violation content recognition；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the convolutional Neural net is adjusted
Network identifies the parameter of the violation content in image.
5. the method according to any claim in claims 1 to 3, which is characterized in that the Recognition with Recurrent Neural Network
Training process, including：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The audio and/or text input Recognition with Recurrent Neural Network of violation content type will be marked, has made the Recognition with Recurrent Neural Network
The violation content in the audio and/or text is identified according to preset parameter；
The violation content type that the Recognition with Recurrent Neural Network is identified is compared with the violation content type of mark, obtains institute
State the cost value that Recognition with Recurrent Neural Network carries out violation content recognition；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the cycle nerve net is adjusted
Network identifies the parameter of audio and/or the violation content in text.
6. a kind of verifying video content device based on machine learning, which is characterized in that including：
Information extraction unit, for extracting video information, audio-frequency information and text message in pending video resource；
Content recognition unit, for carrying out violation content recognition to the video information by trained convolutional neural networks
Processing, and the audio-frequency information and the text message are carried out in violation respectively by trained Recognition with Recurrent Neural Network
Hold identifying processing, judges in the video information, the audio-frequency information and the text message whether to include violation content；
Output unit is used for when in the video information, the audio-frequency information or the text message including violation content, really
Recognize and export the violation content type for including in the video information, the audio-frequency information or the text message.
7. device according to claim 6, which is characterized in that the output unit is additionally operable to：
When it includes violation content to confirm in the video information, the audio-frequency information or the text message, confirms and export
Time location of the violation content in the pending video resource.
8. device according to claim 6, which is characterized in that the content recognition unit passes through trained convolution god
Violation content recognition processing is carried out to the video information through network, judges in the video information whether to include violation content
When, it is specifically used for：
The video information is divided into multiple video frame；
Violation content recognition processing is carried out to the multiple video frame respectively by trained convolutional neural networks, is sentenced respectively
Whether include violation content in each video frame of breaking；
If including violation content in any one video frame, confirm in the video information to include violation content.
9. the device according to any claim in claim 6 to 8, which is characterized in that the content recognition unit is also
For being trained to the convolutional neural networks；
The content recognition unit when being trained, is specifically used for the convolutional neural networks：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The image for having marked violation content type is inputted into convolutional neural networks, makes the convolutional neural networks according to preset
Parameter identifies the violation content in described image；
The violation content type that the convolutional neural networks are identified is compared with the violation content type of mark, obtains institute
State the cost value that convolutional neural networks carry out violation content recognition；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the convolutional Neural net is adjusted
Network identifies the parameter of the violation content in image.
10. the device according to any claim in claim 6 to 8, which is characterized in that the content recognition unit is also
For being trained to the Recognition with Recurrent Neural Network；
The content recognition unit when being trained, is specifically used for the Recognition with Recurrent Neural Network：
Cycle executes following operation, until the cost value being calculated is less than given threshold；
The audio and/or text input Recognition with Recurrent Neural Network of violation content type will be marked, has made the Recognition with Recurrent Neural Network
The violation content in the audio and/or text is identified according to preset parameter；
The violation content type that the Recognition with Recurrent Neural Network is identified is compared with the violation content type of mark, obtains institute
State the cost value that Recognition with Recurrent Neural Network carries out violation content recognition；
Judge whether obtained cost value is less than given threshold；
If obtained cost value is not less than given threshold, according to the cost value being calculated, the cycle nerve net is adjusted
Network identifies the parameter of audio and/or the violation content in text.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810174046.3A CN108419091A (en) | 2018-03-02 | 2018-03-02 | A kind of verifying video content method and device based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810174046.3A CN108419091A (en) | 2018-03-02 | 2018-03-02 | A kind of verifying video content method and device based on machine learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108419091A true CN108419091A (en) | 2018-08-17 |
Family
ID=63129497
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810174046.3A Pending CN108419091A (en) | 2018-03-02 | 2018-03-02 | A kind of verifying video content method and device based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108419091A (en) |
Cited By (35)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109168051A (en) * | 2018-09-11 | 2019-01-08 | 天津理工大学 | A kind of network direct broadcasting platform supervision evidence-obtaining system based on blue-ray storage |
| CN109168024A (en) * | 2018-09-26 | 2019-01-08 | 平安科技（深圳）有限公司 | A kind of recognition methods and equipment of target information |
| CN109284784A (en) * | 2018-09-29 | 2019-01-29 | 北京数美时代科技有限公司 | A kind of content auditing model training method and device for live scene video |
| CN109523127A (en) * | 2018-10-17 | 2019-03-26 | 平安科技（深圳）有限公司 | Staffs training evaluating method and relevant device based on big data analysis |
| CN109660828A (en) * | 2018-09-27 | 2019-04-19 | 深圳壹账通智能科技有限公司 | Video resource management method, equipment and computer readable storage medium |
| CN109656141A (en) * | 2019-01-11 | 2019-04-19 | 武汉天喻聚联网络有限公司 | Violation identification and machine behaviour control method, equipment, storage medium based on artificial intelligence technology |
| CN109754306A (en) * | 2018-11-13 | 2019-05-14 | 北京码牛科技有限公司 | Information processing method and device, electronic equipment and computer-readable medium |
| CN109766807A (en) * | 2018-12-28 | 2019-05-17 | 广州华多网络科技有限公司 | Machine audits processing method, device, electronic equipment and storage medium |
| CN109831697A (en) * | 2018-12-28 | 2019-05-31 | 广州华多网络科技有限公司 | The detection method and system of violation handling duration |
| CN109831696A (en) * | 2018-12-28 | 2019-05-31 | 广州华多网络科技有限公司 | Handle method, apparatus, electronic equipment and the storage medium of violation video content |
| CN110418161A (en) * | 2019-08-02 | 2019-11-05 | 广州虎牙科技有限公司 | Video reviewing method and device, electronic equipment and readable storage medium storing program for executing |
| CN110796098A (en) * | 2019-10-31 | 2020-02-14 | 广州市网星信息技术有限公司 | Method, device, equipment and storage medium for training and auditing content auditing model |
| CN110837615A (en) * | 2019-11-05 | 2020-02-25 | 福建省趋普物联科技有限公司 | Artificial intelligent checking system for advertisement content information filtering |
| CN110852231A (en) * | 2019-11-04 | 2020-02-28 | 云目未来科技(北京)有限公司 | Illegal video detection method and device and storage medium |
| CN110956123A (en) * | 2019-11-27 | 2020-04-03 | 中移（杭州）信息技术有限公司 | Rich media content auditing method and device, server and storage medium |
| CN111079816A (en) * | 2019-12-11 | 2020-04-28 | 北京金山云网络技术有限公司 | Image auditing method and device and server |
| CN111125539A (en) * | 2019-12-31 | 2020-05-08 | 武汉市烽视威科技有限公司 | CDN harmful information blocking method and system based on artificial intelligence |
| CN111126373A (en) * | 2019-12-23 | 2020-05-08 | 北京中科神探科技有限公司 | Internet short video violation judgment device and method based on cross-modal identification technology |
| CN111143612A (en) * | 2019-12-27 | 2020-05-12 | 广州市百果园信息技术有限公司 | Video auditing model training method, video auditing method and related device |
| CN111225234A (en) * | 2019-12-23 | 2020-06-02 | 广州市百果园信息技术有限公司 | Video auditing method, video auditing device, equipment and storage medium |
| CN111368071A (en) * | 2018-12-07 | 2020-07-03 | 北京奇虎科技有限公司 | Video detection method and device based on video related text and electronic equipment |
| CN111382623A (en) * | 2018-12-28 | 2020-07-07 | 广州市百果园信息技术有限公司 | Live broadcast auditing method, device, server and storage medium |
| CN112235632A (en) * | 2020-09-09 | 2021-01-15 | 北京达佳互联信息技术有限公司 | Video processing method and device and server |
| CN112507884A (en) * | 2020-12-10 | 2021-03-16 | 北京有竹居网络技术有限公司 | Live content detection method and device, readable medium and electronic equipment |
| CN112668559A (en) * | 2021-03-15 | 2021-04-16 | 冠传网络科技(南京)有限公司 | Multi-mode information fusion short video emotion judgment device and method |
| CN112749608A (en) * | 2020-06-08 | 2021-05-04 | 腾讯科技（深圳）有限公司 | Video auditing method and device, computer equipment and storage medium |
| CN112860943A (en) * | 2021-01-04 | 2021-05-28 | 浙江诺诺网络科技有限公司 | Teaching video auditing method, device, equipment and medium |
| CN112990182A (en) * | 2021-05-10 | 2021-06-18 | 北京轻松筹信息技术有限公司 | Finance information auditing method and system and electronic equipment |
| CN113221845A (en) * | 2021-06-07 | 2021-08-06 | 北京猎豹移动科技有限公司 | Advertisement auditing method, device, equipment and storage medium |
| CN113613035A (en) * | 2021-07-30 | 2021-11-05 | 广州繁星互娱信息科技有限公司 | Sensitive information processing method and device, electronic equipment and storage medium |
| CN114157829A (en) * | 2020-09-08 | 2022-03-08 | 顺丰科技有限公司 | Model training optimization method and device, computer equipment and storage medium |
| WO2022134591A1 (en) * | 2020-12-23 | 2022-06-30 | 深圳壹账通智能科技有限公司 | Stage-based quality inspection data classification method, apparatus, and device, and storage medium |
| CN115052173A (en) * | 2022-06-07 | 2022-09-13 | 北京胜视京基数字文化产业发展有限公司 | Content analysis method and system for network short video |
| CN115761598A (en) * | 2022-12-20 | 2023-03-07 | 昆明思碓网络科技有限公司 | Big data analysis method and system based on cloud service platform |
| US11948595B2 (en) | 2018-10-10 | 2024-04-02 | Bigo Technology Pte. Ltd. | Method for detecting audio, device, and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010075311A2 (en) * | 2008-12-26 | 2010-07-01 | Five Apes, Inc. | Multi-stage image pattern recognizer |
| CN105844239A (en) * | 2016-03-23 | 2016-08-10 | 北京邮电大学 | Method for detecting riot and terror videos based on CNN and LSTM |
| CN105844238A (en) * | 2016-03-23 | 2016-08-10 | 乐视云计算有限公司 | Method and system for discriminating videos |
| CN106250837A (en) * | 2016-07-27 | 2016-12-21 | 腾讯科技（深圳）有限公司 | The recognition methods of a kind of video, device and system |
| CN107480785A (en) * | 2017-07-04 | 2017-12-15 | 北京小米移动软件有限公司 | The training method and device of convolutional neural networks |
| CN107682719A (en) * | 2017-09-05 | 2018-02-09 | 广州数沃信息科技有限公司 | A kind of monitoring and assessing method and device of live content health degree |
-
2018
- 2018-03-02 CN CN201810174046.3A patent/CN108419091A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2010075311A2 (en) * | 2008-12-26 | 2010-07-01 | Five Apes, Inc. | Multi-stage image pattern recognizer |
| CN105844239A (en) * | 2016-03-23 | 2016-08-10 | 北京邮电大学 | Method for detecting riot and terror videos based on CNN and LSTM |
| CN105844238A (en) * | 2016-03-23 | 2016-08-10 | 乐视云计算有限公司 | Method and system for discriminating videos |
| CN106250837A (en) * | 2016-07-27 | 2016-12-21 | 腾讯科技（深圳）有限公司 | The recognition methods of a kind of video, device and system |
| CN107480785A (en) * | 2017-07-04 | 2017-12-15 | 北京小米移动软件有限公司 | The training method and device of convolutional neural networks |
| CN107682719A (en) * | 2017-09-05 | 2018-02-09 | 广州数沃信息科技有限公司 | A kind of monitoring and assessing method and device of live content health degree |
Cited By (45)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109168051B (en) * | 2018-09-11 | 2021-02-09 | 天津理工大学 | Network live broadcast platform supervision and evidence obtaining system based on blue light storage |
| CN109168051A (en) * | 2018-09-11 | 2019-01-08 | 天津理工大学 | A kind of network direct broadcasting platform supervision evidence-obtaining system based on blue-ray storage |
| CN109168024A (en) * | 2018-09-26 | 2019-01-08 | 平安科技（深圳）有限公司 | A kind of recognition methods and equipment of target information |
| CN109168024B (en) * | 2018-09-26 | 2022-05-27 | 平安科技（深圳）有限公司 | Target information identification method and device |
| CN109660828A (en) * | 2018-09-27 | 2019-04-19 | 深圳壹账通智能科技有限公司 | Video resource management method, equipment and computer readable storage medium |
| CN109660828B (en) * | 2018-09-27 | 2022-04-22 | 深圳壹账通智能科技有限公司 | Video resource management method, device and computer readable storage medium |
| CN109284784A (en) * | 2018-09-29 | 2019-01-29 | 北京数美时代科技有限公司 | A kind of content auditing model training method and device for live scene video |
| US11948595B2 (en) | 2018-10-10 | 2024-04-02 | Bigo Technology Pte. Ltd. | Method for detecting audio, device, and storage medium |
| CN109523127A (en) * | 2018-10-17 | 2019-03-26 | 平安科技（深圳）有限公司 | Staffs training evaluating method and relevant device based on big data analysis |
| CN109754306A (en) * | 2018-11-13 | 2019-05-14 | 北京码牛科技有限公司 | Information processing method and device, electronic equipment and computer-readable medium |
| CN111368071A (en) * | 2018-12-07 | 2020-07-03 | 北京奇虎科技有限公司 | Video detection method and device based on video related text and electronic equipment |
| CN109831696B (en) * | 2018-12-28 | 2021-03-12 | 广州华多网络科技有限公司 | Method and device for processing illegal video content, electronic equipment and storage medium |
| CN111382623A (en) * | 2018-12-28 | 2020-07-07 | 广州市百果园信息技术有限公司 | Live broadcast auditing method, device, server and storage medium |
| CN109831697A (en) * | 2018-12-28 | 2019-05-31 | 广州华多网络科技有限公司 | The detection method and system of violation handling duration |
| CN109831697B (en) * | 2018-12-28 | 2021-05-21 | 广州华多网络科技有限公司 | Method and system for detecting violation processing duration |
| CN109831696A (en) * | 2018-12-28 | 2019-05-31 | 广州华多网络科技有限公司 | Handle method, apparatus, electronic equipment and the storage medium of violation video content |
| CN109766807A (en) * | 2018-12-28 | 2019-05-17 | 广州华多网络科技有限公司 | Machine audits processing method, device, electronic equipment and storage medium |
| CN109656141A (en) * | 2019-01-11 | 2019-04-19 | 武汉天喻聚联网络有限公司 | Violation identification and machine behaviour control method, equipment, storage medium based on artificial intelligence technology |
| CN110418161A (en) * | 2019-08-02 | 2019-11-05 | 广州虎牙科技有限公司 | Video reviewing method and device, electronic equipment and readable storage medium storing program for executing |
| CN110796098B (en) * | 2019-10-31 | 2021-07-27 | 广州市网星信息技术有限公司 | Method, device, equipment and storage medium for training and auditing content auditing model |
| CN110796098A (en) * | 2019-10-31 | 2020-02-14 | 广州市网星信息技术有限公司 | Method, device, equipment and storage medium for training and auditing content auditing model |
| CN110852231A (en) * | 2019-11-04 | 2020-02-28 | 云目未来科技(北京)有限公司 | Illegal video detection method and device and storage medium |
| CN110837615A (en) * | 2019-11-05 | 2020-02-25 | 福建省趋普物联科技有限公司 | Artificial intelligent checking system for advertisement content information filtering |
| CN110956123B (en) * | 2019-11-27 | 2024-02-27 | 中移（杭州）信息技术有限公司 | Method, device, server and storage medium for auditing rich media content |
| CN110956123A (en) * | 2019-11-27 | 2020-04-03 | 中移（杭州）信息技术有限公司 | Rich media content auditing method and device, server and storage medium |
| CN111079816A (en) * | 2019-12-11 | 2020-04-28 | 北京金山云网络技术有限公司 | Image auditing method and device and server |
| CN111126373A (en) * | 2019-12-23 | 2020-05-08 | 北京中科神探科技有限公司 | Internet short video violation judgment device and method based on cross-modal identification technology |
| CN111225234A (en) * | 2019-12-23 | 2020-06-02 | 广州市百果园信息技术有限公司 | Video auditing method, video auditing device, equipment and storage medium |
| CN111143612A (en) * | 2019-12-27 | 2020-05-12 | 广州市百果园信息技术有限公司 | Video auditing model training method, video auditing method and related device |
| CN111125539B (en) * | 2019-12-31 | 2024-02-02 | 武汉市烽视威科技有限公司 | CDN harmful information blocking method and system based on artificial intelligence |
| CN111125539A (en) * | 2019-12-31 | 2020-05-08 | 武汉市烽视威科技有限公司 | CDN harmful information blocking method and system based on artificial intelligence |
| CN112749608A (en) * | 2020-06-08 | 2021-05-04 | 腾讯科技（深圳）有限公司 | Video auditing method and device, computer equipment and storage medium |
| CN112749608B (en) * | 2020-06-08 | 2023-10-17 | 腾讯科技（深圳）有限公司 | Video auditing method, device, computer equipment and storage medium |
| CN114157829A (en) * | 2020-09-08 | 2022-03-08 | 顺丰科技有限公司 | Model training optimization method and device, computer equipment and storage medium |
| CN112235632A (en) * | 2020-09-09 | 2021-01-15 | 北京达佳互联信息技术有限公司 | Video processing method and device and server |
| CN112507884A (en) * | 2020-12-10 | 2021-03-16 | 北京有竹居网络技术有限公司 | Live content detection method and device, readable medium and electronic equipment |
| WO2022134591A1 (en) * | 2020-12-23 | 2022-06-30 | 深圳壹账通智能科技有限公司 | Stage-based quality inspection data classification method, apparatus, and device, and storage medium |
| CN112860943A (en) * | 2021-01-04 | 2021-05-28 | 浙江诺诺网络科技有限公司 | Teaching video auditing method, device, equipment and medium |
| CN112668559A (en) * | 2021-03-15 | 2021-04-16 | 冠传网络科技(南京)有限公司 | Multi-mode information fusion short video emotion judgment device and method |
| CN112990182A (en) * | 2021-05-10 | 2021-06-18 | 北京轻松筹信息技术有限公司 | Finance information auditing method and system and electronic equipment |
| CN113221845A (en) * | 2021-06-07 | 2021-08-06 | 北京猎豹移动科技有限公司 | Advertisement auditing method, device, equipment and storage medium |
| CN113613035A (en) * | 2021-07-30 | 2021-11-05 | 广州繁星互娱信息科技有限公司 | Sensitive information processing method and device, electronic equipment and storage medium |
| CN115052173A (en) * | 2022-06-07 | 2022-09-13 | 北京胜视京基数字文化产业发展有限公司 | Content analysis method and system for network short video |
| CN115761598A (en) * | 2022-12-20 | 2023-03-07 | 昆明思碓网络科技有限公司 | Big data analysis method and system based on cloud service platform |
| CN115761598B (en) * | 2022-12-20 | 2023-09-08 | 易事软件(厦门)股份有限公司 | Big data analysis method and system based on cloud service platform |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108419091A (en) | A kind of verifying video content method and device based on machine learning | |
| CN106790019B (en) | Encryption method for recognizing flux and device based on feature self study | |
| CN104636501B (en) | A kind of artificial instant translation system of multi-media network and method | |
| CN104732978B (en) | The relevant method for distinguishing speek person of text based on combined depth study | |
| CN110489755A (en) | Document creation method and device | |
| CN109522304A (en) | Exception object recognition methods and device, storage medium | |
| CN109118091A (en) | A kind of artistic accomplishment evaluation system | |
| CN108229485A (en) | For testing the method and apparatus of user interface | |
| CN109065027A (en) | Speech differentiation model training method, device, computer equipment and storage medium | |
| CN106104674A (en) | Mixing voice identification | |
| CN107251033A (en) | System and method for carrying out active user checking in online education | |
| CN109147765A (en) | Audio quality comprehensive evaluating method and system | |
| CN106161209B (en) | A kind of method for filtering spam short messages and system based on depth self study | |
| CN106897746A (en) | Data classification model training method and device | |
| CN107343223A (en) | The recognition methods of video segment and device | |
| CN106897559A (en) | A kind of symptom and sign class entity recognition method and device towards multi-data source | |
| CN109657038A (en) | The method for digging, device and electronic equipment of a kind of question and answer to data | |
| CN111160606B (en) | Test question difficulty prediction method and related device | |
| CN111182162A (en) | Telephone quality inspection method, device, equipment and storage medium based on artificial intelligence | |
| CN110363233A (en) | A kind of the fine granularity image-recognizing method and system of the convolutional neural networks based on block detector and Fusion Features | |
| CN107633225A (en) | Information obtaining method and device | |
| CN109741734A (en) | A kind of speech evaluating method, device and readable medium | |
| CN109992781A (en) | Processing, device, storage medium and the processor of text feature | |
| CN106653020A (en) | Multi-business control method and system for smart sound and video equipment based on deep learning | |
| CN109686382A (en) | A kind of speaker clustering method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180817 |