The content of the invention
It is contemplated that at least solving one of technical problem in correlation technique to a certain extent.
Therefore, first purpose of the present invention is to propose a kind of certain content recognition methods, to realize for live
Certain content recognizes that manual identified certain content recognition effect is very poor in the prior art for solution, and the higher skill of Expenses Cost
Art problem.
Second object of the present invention is to propose a kind of certain content identifying device.
Third object of the present invention is to propose a kind of electronic equipment.
For up to above-mentioned purpose, first aspect present invention embodiment proposes a kind of certain content recognition methods, including:
Sectional drawing, generation picture sample storehouse are carried out to history live video;
Based on each samples pictures in picture sample storehouse, and each samples pictures whether be certain content mark, to machine
Learning model is trained;
Online live video is carried out to the test pictures obtained by sectional drawing, carried out using trained machine learning model
Content recognition, re -training is carried out according to recognition result to machine learning model;
Using the machine learning model by re -training, certain content identification is carried out to online live video.
Alternatively, as the first possible implementation of first aspect, machine learning model, including first order model
With second level model, the first order model and the second level model have identical model structure.
Alternatively, it is described to be based on each sample in picture sample storehouse as second of possible implementation of first aspect
Picture, and each samples pictures whether be certain content mark, machine learning model is trained, including:
Based on each samples pictures in the picture sample storehouse, and the picture sample storehouse whether be certain content mark
Note, is trained to the first order model of the machine learning model.
Alternatively, it is described that online live video is subjected to sectional drawing as the third possible implementation of first aspect
Resulting test pictures, carry out content recognition, according to recognition result to engineering using trained machine learning model
Practise model and carry out re -training, including:
By the test pictures, content recognition is carried out using trained first order model;
According to wrong part is recognized in the test pictures, second level model is trained, obtains trained
Second level model.
Alternatively, it is described to be recognized according in the test pictures as the 4th kind of possible implementation of first aspect
Wrong part, is trained to second level model, obtains trained second level model, including:
From the test pictures, the wrong misrecognition picture of inquiry identification, wherein, the misrecognition picture includes mark
For the particular picture comprising normal content, and/or it is labeled as including the normal picture of certain content;
According to the misrecognition picture, the picture sample storehouse is regenerated;
Based on the picture sample storehouse regenerated, the second level model of the machine learning model is trained.
Alternatively, as the 5th kind of possible implementation of first aspect, the machine using by re -training
Learning model, certain content identification is carried out to online live video, including:
Certain content identification is carried out to the sectional drawing of online live video using the first order model;
According to the confidence level of identification, the sectional drawing by the confidence level of identification less than the online live video of threshold value, using described
Second level model, carries out certain content identification.
Alternatively, as the 6th kind of possible implementation of first aspect, the machine learning model is convolutional Neural
Network;The certain content, including illegal contents.
The certain content recognition methods of the embodiment of the present invention, by carrying out sectional drawing to history live video, generates picture sample
This storehouse, based on each samples pictures in picture sample storehouse, and each samples pictures whether be certain content mark, to machine learning
Model is trained, and online live video is carried out to the test pictures obtained by sectional drawing, trained machine learning mould is utilized
Type carries out content recognition, carries out re -training to machine learning model according to recognition result, utilizes the machine by re -training
Learning model, certain content identification is carried out to online live video.Due to make use of machine learning model to carry out certain content knowledge
Not, identification process reduces human cost, improves recognition efficiency, solve people in the prior art without manually being participated in
The relatively low technical problem of work recognition efficiency.
For up to above-mentioned purpose, second aspect of the present invention embodiment proposes a kind of certain content identifying device, including:
Generation module, for carrying out sectional drawing, generation picture sample storehouse to history live video;
Training module, for based on each samples pictures in picture sample storehouse, and whether each samples pictures are certain content
Mark, machine learning model is trained;
Retraining module, for online live video to be carried out to the test pictures obtained by sectional drawing, using trained
Machine learning model carries out content recognition, and re -training is carried out to machine learning model according to recognition result;
Identification module, for using the machine learning model by re -training, being carried out to online live video in specific
Hold identification.
Alternatively, as the first possible implementation of second aspect, the machine learning model, including the first order
Model and second level model, the first order model and the second level model have identical model structure.
Alternatively, as second of possible implementation of second aspect, the training module, specifically for:
Based on each samples pictures in the picture sample storehouse, and the picture sample storehouse whether be certain content mark
Note, is trained to the first order model of the machine learning model.
Alternatively, as the third possible implementation of second aspect, the retraining module, including:
Recognition unit, for by the test pictures, content recognition to be carried out using trained first order model;
Retraining unit, for according to wrong part is recognized in the test pictures, being trained to second level model,
Obtain trained second level model.
Alternatively, as the 4th kind of possible implementation of second aspect, the retraining unit, specifically for:
From the test pictures, the wrong misrecognition picture of inquiry identification, wherein, the misrecognition picture includes mark
For the particular picture comprising normal content, and/or it is labeled as including the normal picture of certain content;
According to the misrecognition picture, the picture sample storehouse is regenerated;
Based on the picture sample storehouse regenerated, the second level model of the machine learning model is trained.
Alternatively, as the 5th kind of possible implementation of second aspect, the identification module, specifically for:
Certain content identification is carried out to the sectional drawing of online live video using the first order model;
According to the confidence level of identification, the sectional drawing by the confidence level of identification less than the online live video of threshold value, using described
Second level model, carries out certain content identification.
Alternatively, as the 6th kind of possible implementation of second aspect, the machine learning model is convolutional Neural
Network;The certain content, including illegal contents.
The certain content identifying device of the embodiment of the present invention, by carrying out sectional drawing to history live video, generates picture sample
This storehouse, based on each samples pictures in picture sample storehouse, and each samples pictures whether be certain content mark, to machine learning
Model is trained, and online live video is carried out to the test pictures obtained by sectional drawing, trained machine learning mould is utilized
Type carries out content recognition, carries out re -training to machine learning model according to recognition result, utilizes the machine by re -training
Learning model, certain content identification is carried out to online live video.Due to make use of machine learning model to carry out certain content knowledge
Not, identification process reduces human cost, improves recognition efficiency, solve people in the prior art without manually being participated in
The relatively low technical problem of work recognition efficiency.
For up to above-mentioned purpose, third aspect present invention embodiment proposes electronic equipment, including:Housing, processor, storage
Device, circuit board and power circuit, wherein, circuit board is placed in the interior volume that housing is surrounded, and processor and memory are arranged on
On circuit board;Power circuit, for being powered for each circuit or device of above-mentioned electronic equipment;Memory is used to store executable
Program code;The executable program code that processor is stored by reading in memory is corresponding with executable program code to run
Program, for performing the certain content recognition methods described in first aspect.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by the practice of the present invention.
Embodiment
Embodiments of the invention are described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or the element with same or like function.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and be not considered as limiting the invention.
Below with reference to the accompanying drawings certain content recognition methods and the device of the embodiment of the present invention described.
The live of internet has the features such as real-time, uncertainty degree is high, concurrency is big, is brought very to content monitoring
Big challenge.In particular in terms of protection of minors, it is desirable to be able to real-time monitor the certain content included in live.
Here certain content includes illegal contents, waste advertisements, vulgar content etc..
A kind of schematic flow sheet for certain content recognition methods that Fig. 1 is provided by the embodiment of the present invention, the present embodiment institute
The method of offer, make use of machine learning model to carry out certain content identification, and identification process is reduced without manually being participated in
Human cost, improves recognition efficiency, solves the less efficient technical problem of manual identified in the prior art.
As shown in figure 1, certain content recognition methods comprises the following steps:
Step 101, sectional drawing, generation picture sample storehouse are carried out to history live video.
Specifically, every default frame number, sectional drawing is carried out to history live video, or every predetermined period, it is straight to history
Broadcast video and carry out sectional drawing.The picture that sectional drawing is obtained is used as the samples pictures in picture sample storehouse.Using manual type, to sectional drawing
It is identified, it is particular picture or normal picture to mark each pictures.
Wherein, the picture comprising certain content can not be considered as normal picture.
Step 102, based on each samples pictures in picture sample storehouse, and each samples pictures whether be certain content mark
Note, is trained to machine learning model.
Specifically, by being trained to machine learning model so that each parameter value is arrived in machine learning model study, so that
Realize the identification to certain content.
Step 103, online live video is carried out to the test pictures obtained by sectional drawing, trained machine learning is utilized
Model carries out content recognition, and re -training is carried out to machine learning model according to recognition result.
Specifically, machine learning model is put on line and run, sectional drawing periodically is carried out to online live video, obtained
To test pictures.Content recognition is carried out to test pictures using trained machine learning model.
From the test pictures, the wrong misrecognition picture of inquiry identification, wherein, the misrecognition picture includes mark
For the particular picture comprising normal content, and/or it is labeled as including the normal picture of certain content;Schemed according to the misrecognition
Piece, regenerates the picture sample storehouse;Based on the picture sample storehouse regenerated, the machine learning model is carried out
Re -training.
Step 104, using the machine learning model by re -training, certain content knowledge is carried out to online live video
Not.
The method that the present embodiment is provided, can repeat step 103 and step 104, so that machine learning mould
Type is more perfect, and recognition accuracy is improved constantly.
In the present embodiment, sectional drawing, generation picture sample storehouse, based on various kinds in picture sample storehouse are carried out to history live video
This picture, and each samples pictures whether be certain content mark, machine learning model is trained, live regarded online
Frequency carries out the test pictures obtained by sectional drawing, and content recognition is carried out using trained machine learning model, is tied according to identification
Fruit carries out re -training to machine learning model, and using the machine learning model by re -training, online live video is entered
Row certain content is recognized.Due to make use of machine learning model to carry out certain content identification, identification process is without manually being joined
With reducing human cost, improving recognition efficiency, solve the less efficient technical problem of manual identified in the prior art.
For an embodiment in clear explanation, another certain content recognition methods is present embodiments provided, Fig. 2 is this hair
The schematic flow sheet for another certain content recognition methods that bright embodiment is provided, in the present embodiment, machine learning model
Specially two-level model, including first order model and second level model, first order model and the second level model have identical
Convolutional neural networks model structure.When the confidence level of first order Model Identification is less than threshold value, carried out using second level model special
Determine content recognition.Because second level model is to be trained what is obtained based on the wrong misrecognition picture of first order Model Identification,
Therefore, it is possible to obtain more accurate recognition result, recognition accuracy is improved.
The method of the present embodiment is applied particularly under application scenarios that illegal contents are identified, so that in the present embodiment
Certain content specifically include illegal contents, correspondingly, particular picture be the picture comprising certain content, be specifically as follows and include
The illegal picture of illegal contents.Wherein, certain content here can refer to waste advertisements, violence, pornographic etc..
As shown in Fig. 2 the certain content recognition methods, including:Training stage and cognitive phase.
Wherein, the training stage includes:
Step 201, sectional drawing, generation picture sample storehouse are carried out to history live video.
Specifically, every default frame number, sectional drawing is carried out to history live video, or every predetermined period, it is straight to history
Broadcast video and carry out sectional drawing.The picture that sectional drawing is obtained is used as the samples pictures in picture sample storehouse.Using manual type, to sectional drawing
It is identified, it is illegal picture or normal picture to mark each pictures.
Step 202, based on each samples pictures in picture sample storehouse, and picture sample storehouse whether be illegal contents mark
Note, is trained to the first order model of machine learning model.
Wherein, first order model and second level model refer to the convolutional neural networks model of Google
inception v3.From unlike inception v3, the first order model and second level model of the present embodiment are being connected entirely
Layer, softmax functions output recognition result is converted into by logit functions.So as to the numerical value for the 0-1 for directly using output,
It is used as the confidence level of this identification.
Step 203, online live video is carried out to the test pictures obtained by sectional drawing, trained first order mould is utilized
Type carries out content recognition.
Specifically, online live video is carried out to the test pictures obtained by sectional drawing, just with trained first
Level model carries out content recognition, without being identified using second level model.
Step 204, according to wrong part is recognized in test pictures, second level model is trained, obtained by instruction
Experienced second level model.
Specifically, from test pictures, the wrong misrecognition picture of inquiry identification, wherein, misrecognition picture includes mark
For the illegal picture comprising normal content, and/or it is labeled as including the normal picture of illegal contents.According to misrecognition picture, weight
The newly-generated picture sample storehouse.Based on the picture sample storehouse regenerated, the second level model of machine learning model is carried out
Training.
Wherein, cognitive phase includes:
Step 205, during ONLINE RECOGNITION illegal contents, the sectional drawing of online live video is carried out using first order model illegal
Content recognition, whether determine each sectional drawing is illegal picture, and recognition result confidence level.
Step 206, judge whether confidence level is less than threshold value, if less than threshold value, performing step 207, otherwise performing step
208。
Step 207, if confidence level is less than threshold value, using second level model, illegal contents identification is re-started to the sectional drawing,
Accept and believe the recognition result of second level model.
Step 208, if confidence level is not less than threshold value, the recognition result of first order model is accepted and believed.
In the present embodiment, sectional drawing, generation picture sample storehouse, based on various kinds in picture sample storehouse are carried out to history live video
This picture, and each samples pictures whether be illegal contents mark, machine learning model is trained, live regarded online
Frequency carries out the test pictures obtained by sectional drawing, and content recognition is carried out using trained machine learning model, is tied according to identification
Fruit carries out re -training to machine learning model, and using the machine learning model by re -training, online live video is entered
Row illegal contents are recognized.Due to make use of machine learning model to carry out illegal contents identification, identification process is without manually being joined
With reducing human cost, improving recognition efficiency, solve the less efficient technical problem of manual identified in the prior art.
In order to realize above-described embodiment, the present invention also proposes a kind of certain content identifying device.
Fig. 3 is a kind of structural representation of certain content identifying device provided in an embodiment of the present invention.
As shown in figure 3, the certain content identifying device includes:Generation module 31, training module 32, the and of retraining module 33
Identification module 34.
Generation module 31, for carrying out sectional drawing, generation picture sample storehouse to history live video.
Training module 32, for based on each samples pictures in picture sample storehouse, and whether each samples pictures are in specific
The mark of appearance, is trained to machine learning model.
Retraining module 33, for by the test pictures obtained by online live video progress sectional drawing, using by training
Machine learning model carry out content recognition, according to recognition result to machine learning model carry out re -training.
Identification module 34, for using the machine learning model by re -training, being carried out to online live video specific
Content recognition.
It should be noted that the foregoing explanation to embodiment of the method is also applied for the device of the embodiment, herein not
Repeat again.
Based on above-described embodiment, the embodiment of the present invention additionally provides the possible realization of another certain content identifying device
Mode, Fig. 4 is the structural representation of another certain content identifying device provided in an embodiment of the present invention, in a upper embodiment
On the basis of, machine learning model, including first order model and second level model, the first order model and the second level model
With identical model structure, for example:Convolutional neural networks.
Further, training module 32, specifically for:Based on each in the picture sample storehouse, and the picture sample storehouse
Samples pictures whether be certain content mark, the first order model of the machine learning model is trained.
Retraining module 33, including:Recognition unit 331 and retraining unit 332.
Recognition unit 331, for by the test pictures, content recognition to be carried out using trained first order model.
Retraining unit 332, for according to wrong part is recognized in the test pictures, being instructed to second level model
Practice, obtain trained second level model.
Specifically, retraining unit 332, specifically for:From the test pictures, the wrong misrecognition figure of inquiry identification
Piece, wherein, the misrecognition picture includes being labeled as including the particular picture of normal content, and/or is labeled as comprising specific interior
The normal picture of appearance;According to the misrecognition picture, the picture sample storehouse is regenerated;Based on the picture regenerated
Sample Storehouse, is trained to the second level model of the machine learning model.
Further, identification module 34, specifically for:The sectional drawing of online live video is carried out using the first order model
Certain content is recognized;According to the confidence level of identification, the sectional drawing by the confidence level of identification less than the online live video of threshold value is utilized
The second level model, carries out certain content identification.
In the embodiment of the present invention, sectional drawing, generation picture sample storehouse, based in picture sample storehouse are carried out to history live video
Each samples pictures, and each samples pictures whether be certain content mark, machine learning model is trained, will be straight online
The test pictures obtained by video progress sectional drawing are broadcast, content recognition are carried out using trained machine learning model, according to knowledge
Other result carries out re -training to machine learning model, using the machine learning model by re -training, is regarded to live online
Frequency carries out certain content identification.Due to make use of machine learning model to carry out certain content identification, identification process is without manually entering
Row is participated in, and is reduced human cost, is improved recognition efficiency, solves the less efficient technology of manual identified in the prior art and ask
Topic.
As a kind of possible application scenarios, the device that the present embodiment is provided is applied particularly to know illegal contents
Under other application scenarios so that the certain content in the present embodiment specifically includes illegal contents, correspondingly, particular picture be comprising
The picture of certain content, is specifically as follows the illegal picture comprising illegal contents.Wherein, certain content here can refer to rubbish
Rubbish advertisement, violence, pornographic etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means to combine specific features, structure, material or the spy that the embodiment or example are described
Point is contained at least one embodiment of the present invention or example.In this manual, to the schematic representation of above-mentioned term not
Identical embodiment or example must be directed to.Moreover, specific features, structure, material or the feature of description can be with office
Combined in an appropriate manner in one or more embodiments or example.In addition, in the case of not conflicting, the skill of this area
Art personnel can be tied the not be the same as Example or the feature of example and non-be the same as Example or example described in this specification
Close and combine.
In addition, term " first ", " second " are only used for describing purpose, and it is not intended that indicating or implying relative importance
Or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can express or
Implicitly include at least one this feature.In the description of the invention, " multiple " are meant that at least two, such as two, three
It is individual etc., unless otherwise specifically defined.
Any process described otherwise above or method description are construed as in flow chart or herein, represent to include
Module, fragment or the portion of the code of one or more executable instructions for the step of realizing custom logic function or process
Point, and the scope of the preferred embodiment of the present invention includes other realization, wherein can not be by shown or discussion suitable
Sequence, including according to involved function by it is basic simultaneously in the way of or in the opposite order, carry out perform function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Represent in flow charts or logic and/or step described otherwise above herein, for example, being considered use
In the order list for the executable instruction for realizing logic function, it may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system including the system of processor or other can be held from instruction
The system of row system, device or equipment instruction fetch and execute instruction) use, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium " can any can be included, store, communicate, propagate or pass
Defeated program is for instruction execution system, device or equipment or the dress for combining these instruction execution systems, device or equipment and using
Put.The more specifically example (non-exhaustive list) of computer-readable medium includes following:Electricity with one or more wirings
Connecting portion (electronic installation), portable computer diskette box (magnetic device), random access memory (RAM), read-only storage
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device, and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, can even is that can be in the paper of printing described program thereon or other are suitable for computer-readable medium
Medium, because can then enter edlin, interpretation or if necessary with it for example by carrying out optical scanner to paper or other media
His suitable method is handled electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each several part of the present invention can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In embodiment, the software that multiple steps or method can in memory and by suitable instruction execution system be performed with storage
Or firmware is realized.Such as, if realized with hardware with another embodiment, following skill well known in the art can be used
Any one of art or their combination are realized:With the logic gates for realizing logic function to data-signal from
Scattered logic circuit, the application specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can be compiled
Journey gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method is carried
Rapid to can be by program to instruct the hardware of correlation to complete, described program can be stored in a kind of computer-readable storage medium
In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
In addition, each functional unit in each embodiment of the invention can be integrated in a processing module, can also
That unit is individually physically present, can also two or more units be integrated in a module.Above-mentioned integrated mould
Block can both be realized in the form of hardware, it would however also be possible to employ the form of software function module is realized.The integrated module is such as
Fruit is realized using in the form of software function module and as independent production marketing or in use, can also be stored in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..Although having been shown and retouching above
Embodiments of the invention are stated, it is to be understood that above-described embodiment is exemplary, it is impossible to be interpreted as the limit to the present invention
System, one of ordinary skill in the art can be changed to above-described embodiment, change, replace and become within the scope of the invention
Type.
The embodiment of the present invention also provides a kind of electronic equipment, and electronic equipment includes the device described in foregoing any embodiment.
Fig. 5 is the structural representation of electronic equipment one embodiment of the present invention, it is possible to achieve implemented shown in Fig. 1-4 of the present invention
The flow of example, as shown in figure 5, above-mentioned electronic equipment can include:Housing 41, processor 42, memory 43, circuit board 44 and electricity
Source circuit 45, wherein, circuit board 44 is placed in the interior volume that housing 41 is surrounded, and processor 42 and memory 43 are arranged on circuit
On plate 44;Power circuit 45, for being powered for each circuit or device of above-mentioned electronic equipment;Memory 43 is used to store and can hold
Line program code;Processor 42 is run and executable program generation by reading the executable program code stored in memory 43
The corresponding program of code, for performing the certain content recognition methods described in foregoing any embodiment.
Processor 42 to the specific implementation procedure and processor 42 of above-mentioned steps by run executable program code come
The step of further performing, may refer to the description of Fig. 1-4 illustrated embodiments of the present invention, will not be repeated here.
The electronic equipment exists in a variety of forms, includes but is not limited to:
(1) mobile communication equipment:The characteristics of this kind equipment is that possess mobile communication function, and to provide speech, data
Communicate as main target.This Terminal Type includes:Smart mobile phone (such as iPhone), multimedia handset, feature mobile phone, and it is low
Hold mobile phone etc..
(2) super mobile personal computer equipment:This kind equipment belongs to the category of personal computer, there is calculating and processing work(
Can, typically also possess mobile Internet access characteristic.This Terminal Type includes:PDA, MID and UMPC equipment etc., such as iPad.
(3) portable entertainment device:This kind equipment can show and play content of multimedia.The kind equipment includes:Audio,
Video player (such as iPod), handheld device, e-book, and intelligent toy and portable car-mounted navigation equipment.
(4) server:The equipment for providing the service of calculating, the composition of server is total including processor, hard disk, internal memory, system
Line etc., server is similar with general computer architecture, but is due to need to provide highly reliable service, therefore in processing energy
Require higher in terms of power, stability, reliability, security, scalability, manageability.
(5) other electronic equipments with data interaction function.
One of ordinary skill in the art will appreciate that realize all or part of flow in above-described embodiment method, being can be with
The hardware of correlation is instructed to complete by computer program, described program can be stored in a computer read/write memory medium
In, the program is upon execution, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any
Those familiar with the art the invention discloses technical scope in, the change or replacement that can be readily occurred in, all should
It is included within the scope of the present invention.Therefore, protection scope of the present invention should be defined by scope of the claims.