CN108830329A - Image processing method and device - Google Patents
Image processing method and device Download PDFInfo
- Publication number
- CN108830329A CN108830329A CN201810649563.1A CN201810649563A CN108830329A CN 108830329 A CN108830329 A CN 108830329A CN 201810649563 A CN201810649563 A CN 201810649563A CN 108830329 A CN108830329 A CN 108830329A
- Authority
- CN
- China
- Prior art keywords
- picture
- target photo
- training sample
- coordinate position
- marking error
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Abstract
The embodiment of the present application discloses image processing method and device.One specific embodiment of this method includes：By Target Photo input disaggregated model trained in advance；It is inconsistent in response to the classification results that determine that the Target Photo object that is included and the disaggregated model export, it is concentrated in the training sample of the disaggregated model, search the similar pictures of the Target Photo, wherein, the similarity of the similar pictures and the Target Photo is greater than the similarity that the training sample concentrates other samples pictures Yu the Target Photo；The similar pictures are determined as doubtful marking error picture.The embodiment of the present application filters out doubtful marking error picture from the samples pictures of training sample set, to reduce the quantity for marking error map piece in training sample set, improves the accuracy of model.
Description
Technical field
The invention relates to field of computer technology, and in particular at Internet technical field more particularly to picture
Manage method and apparatus.
Background technique
Picture recognition model refers to and is handled image, analyzed and understood using computer, to identify target and object
Model.Carrying out image recognition using the model has the advantages that fast and accurately, therefore, increasingly by picture recognition model
Mostly it is applied in the identification of object included in picture.
Before carrying out image recognition using picture recognition model, it usually needs be trained to model.It specifically, can be with
Training sample is taken out from sample set and inputs initial model, obtains preliminary classification result.And backpropagation is carried out based on loss function
To adjust model parameter.
The training sample set scale of construction of picture recognition model is huge, wherein there may be the pictures of marking error, i.e., in training
To the marking error of positive sample and negative sample in sample set.If be trained with the sample set, it be easy to cause the output of model
As a result mistake.
Summary of the invention
The embodiment of the present application proposes image processing method and device.
In a first aspect, the embodiment of the present application provides a kind of image processing method, including：Target Photo is inputted into instruction in advance
Experienced disaggregated model；It is inconsistent in response to the classification results that determine that the Target Photo object that is included and disaggregated model export,
The training sample of disaggregated model is concentrated, and searches the similar pictures of Target Photo, wherein the similarity of similar pictures and Target Photo
Greater than the similarity that training sample concentrates other samples pictures and Target Photo；Similar pictures are determined as doubtful marking error figure
Piece.
In some embodiments, Target Photo is the samples pictures of the training sample concentration of train classification models.
In some embodiments, it is concentrated in the training sample of picture classification model, searches the similar pictures of input picture, packet
It includes：The characteristic information of Target Photo is obtained, the characteristic information that training sample concentrates each picture is obtained；In hyperspace coordinate system
In, it determines that the coordinate position of the characteristic information of Target Photo is target coordinate position, determines the spy for the picture that training sample is concentrated
The coordinate position of reference breath is coordinate position to be selected；According to the sequence at a distance from target coordinate position from small to large, from each
Preset quantity coordinate position is determined in a coordinate position to be selected；Picture corresponding to preset quantity coordinate position is determined
For similar pictures.
In some embodiments, after similar pictures to be determined as to doubtful marking error picture, method further includes：It sends
Doubtful marking error picture determines whether doubtful marking error picture is marking error picture.
In some embodiments, after sending doubtful marking error picture, method further includes：Amendment picture is obtained,
In, amendment picture is the picture generated after being labeled corrigendum to marking error picture；The mark that training sample is concentrated is wrong
Accidentally picture replaces with amendment picture.
In some embodiments, method further includes：Based on the mark of amendment picture, re -training disaggregated model.
Second aspect, the embodiment of the present application provide a kind of picture processing unit, including：Taxon, be configured to by
Target Photo input disaggregated model trained in advance；Searching unit is configured in response to determine pair that Target Photo is included
As the classification results exported with disaggregated model are inconsistent, concentrated in the training sample of disaggregated model, search the similar of Target Photo
Picture, wherein the similarity of similar pictures and Target Photo is greater than training sample and concentrates other samples pictures and Target Photo
Similarity；Determination unit is configured to for similar pictures to be determined as doubtful marking error picture.
In some embodiments, Target Photo is the samples pictures of the training sample concentration of train classification models.
In some embodiments, searching unit is further configured to：The characteristic information of Target Photo is obtained, training is obtained
The characteristic information of each picture in sample set；In hyperspace coordinate system, the coordinate bit of the characteristic information of Target Photo is determined
It is set to target coordinate position, determines that the coordinate position of the characteristic information for the picture that training sample is concentrated is coordinate position to be selected；
According to the sequence at a distance from target coordinate position from small to large, determine that preset quantity is sat from each coordinate position to be selected
Cursor position；Picture corresponding to preset quantity coordinate position is determined as similar pictures.
In some embodiments, device further includes：Transmission unit is configured to send doubtful marking error picture, determines
Whether doubtful marking error picture is marking error picture.
In some embodiments, device further includes：Acquiring unit is configured to obtain amendment picture, wherein amendment picture
To be labeled the picture generated after corrigendum to marking error picture；Replacement unit is configured to concentrate training sample
Marking error picture replaces with amendment picture.
In some embodiments, device further includes：Training unit is configured to instruct again based on the mark of amendment picture
Practice disaggregated model.
The third aspect, the embodiment of the present application provide a kind of electronic equipment, including：One or more processors；Storage dress
It sets, for storing one or more programs, when one or more programs are executed by one or more processors, so that one or more
A processor realizes the method such as any embodiment in image processing method.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer journey
Sequence realizes the method such as any embodiment in image processing method when the program is executed by processor.
Picture processing scheme provided by the embodiments of the present application, firstly, the disaggregated model that Target Photo input is trained in advance.
Later, the object for being included in response to determining Target Photo is not belonging to the classification results of disaggregated model output, in disaggregated model
Training sample is concentrated, and the similar pictures of Target Photo are searched.Wherein, the similarity of similar pictures and Target Photo is greater than training sample
This concentrates the similarity of other samples pictures and Target Photo.Finally, similar pictures are determined as doubtful marking error picture.This
Application embodiment filters out doubtful marking error picture from the samples pictures of training sample set, to reduce training sample set
The quantity of middle mark error map piece, improves the accuracy of model.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon：
Fig. 1 is that this application can be applied to exemplary system architecture figures therein；
Fig. 2 is the flow chart according to one embodiment of the image processing method of the application；
Fig. 3 is the schematic diagram according to an application scenarios of the image processing method of the application；
Fig. 4 is the flow chart according to another embodiment of the image processing method of the application；
Fig. 5 is the structural schematic diagram according to one embodiment of the picture processing unit of the application；
Fig. 6 is adapted for the structural schematic diagram for the computer system for realizing the electronic equipment of the embodiment of the present application.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is shown can be using the exemplary system of the embodiment of the image processing method or picture processing unit of the application
System framework 100.
As shown in Figure 1, system architecture 100 may include terminal device 101,102,103, network 104 and server 105.
Network 104 between terminal device 101,102,103 and server 105 to provide the medium of communication link.Network 104 can be with
Including various connection types, such as wired, wireless communication link or fiber optic cables etc..
User can be used terminal device 101,102,103 and be interacted by network 104 with server 105, to receive or send out
Send message etc..Various telecommunication customer end applications can be installed on terminal device 101,102,103, such as image recognition application,
Shopping class application, searching class application, instant messaging tools, mailbox client, social platform software etc..
Here terminal 101,102,103 can be hardware, be also possible to software.When terminal 101,102,103 is hardware
When, can be the various electronic equipments with display screen, including but not limited to smart phone, tablet computer, E-book reader,
Pocket computer on knee and desktop computer etc..When terminal 101,102,103 is software, may be mounted at above-mentioned listed
In the electronic equipment of act.Multiple softwares or software module may be implemented into (such as providing the multiple soft of Distributed Services in it
Part or software module), single software or software module also may be implemented into.It is not specifically limited herein.
Server 105 can be to provide the server of various services, such as provide support to terminal device 101,102,103
Background server.Background server can carry out the data such as the Target Photo received the processing such as analyzing, and processing is tied
Fruit (such as doubtful marking error picture) feeds back to terminal device.
It should be noted that image processing method provided by the embodiment of the present application can be by server 105 or terminal
Equipment 101,102,103 executes, correspondingly, picture processing unit can be set in server 105 or terminal device 101,
102, in 103.
It should be understood that the number of terminal device, network and server in Fig. 1 is only schematical.According to realization need
It wants, can have any number of terminal device, network and server.
With continued reference to Fig. 2, the process 200 of one embodiment of the image processing method according to the application is shown.The figure
Piece processing method, includes the following steps：
Step 201, the disaggregated model that Target Photo input is trained in advance.
In the present embodiment, executing subject (such as server shown in FIG. 1 or the end of image processing method operation thereon
End equipment) Target Photo can be inputted to disaggregated model trained in advance.Above-mentioned disaggregated model can be two disaggregated models, can also
To be more disaggregated models, the object for being mainly used for being included to picture is classified.Target Photo refer to comprising some or it is certain
Any image of object.
In practice, disaggregated model can be by support vector machines (Support Vector Machine, SVM), simplicity
Bayesian model (Naive Bayesian Model, NBM), convolutional neural networks model (Convolutional Neural
Network, CNN) etc. classifiers (Classifier) training obtain.In addition, disaggregated model is also possible to by certain classification letters
Number (such as softmax function etc.) is in advance made of training.
The training process of disaggregated model can be as follows：Training sample set, each sample that training sample is concentrated are obtained first
It is included the picture of the type of object for mark.It can be using each picture as input, and the object that will included to picture
The type marked is trained preliminary classification model (such as above-mentioned support vector machines etc.) as output, is classified
Model.In the case where disaggregated model is two disaggregated model, two kinds are divided into the mark of sample.It can be wrapped at this time according to sample
The object contained, to one of above two mark of each sample mark.It is right in the case where disaggregated model is more disaggregated models
The mark of sample then can there are many.The object that can included according to sample at this time can mark each sample above-mentioned
One of a variety of marks are a variety of.
It is step 202, inconsistent in response to the classification results for determining that the Target Photo object that is included and disaggregated model export,
It is concentrated in the training sample of disaggregated model, searches the similar pictures of Target Photo, wherein similar pictures are similar to Target Photo
Degree is greater than the similarity that training sample concentrates other samples pictures and Target Photo.
It in the present embodiment, can be from disaggregated model after Target Photo is inputted disaggregated model by above-mentioned executing subject
Output in get classification results.Above-mentioned executing subject can be in response to determining the Target Photo object that is included and above-mentioned point
The classification results of class model output are inconsistent, concentrate in the training sample of disaggregated model, search the similar pictures of Target Photo.
Here the similarity of similar pictures and Target Photo is greater than training sample and concentrates other samples pictures and target figure
The similarity of piece.Namely the similarity between each samples pictures and Target Photo is ranked up, preset quantity similar diagram
The position of preset quantity before similarity between piece and Target Photo comes.
Above-mentioned executing subject can determine what the object that Target Photo is included and disaggregated model exported using various ways
Whether classification results are consistent.For example, above-mentioned executing subject can determine target if Target Photo is the picture by marking classification
Whether the classification that the object in picture is marked is consistent with classification results.If it is inconsistent, can determine that Target Photo is wrapped
The object contained and the classification results that disaggregated model exports are inconsistent, and vice versa.If Target Photo is without mark classification
Picture, above-mentioned executing subject can be sentenced determined by the object and classification results that user is included based on Target Photo by receiving
It is disconnected as a result, whether consistent come the classification results for determining that object that Target Photo is included and disaggregated model export.
In specific application scenarios, disaggregated model, point of disaggregated model will be inputted comprising the picture A of " cat " this object
Class is the result is that dog.Above-mentioned executing subject can be in the similar pictures of the training sample of disaggregated model concentration lookup picture A, these phases
It may include the picture for being labeled as the cat of dog like picture.
Similar pictures can be understood as and the higher picture of the similarity of Target Photo.Various ways can be used from training
The similar pictures of Target Photo are determined in sample set.For example, above-mentioned executing subject can determine the cryptographic Hash of picture.Figure
The similarity of the smaller picture of cryptographic Hash difference between piece is bigger, so similar pictures determined by the following ways and target
The similarity of picture is greater than the similarity between other samples pictures that training sample is concentrated and Target Photo.If Target Photo
The cryptographic Hash difference for certain samples pictures concentrated with training sample is less than threshold value, then can be using the samples pictures as similar diagram
Piece.Or when the sequence for carrying out cryptographic Hash difference between the samples pictures concentrated to Target Photo and training sample.Training sample
The smallest preset quantity samples pictures of difference are concentrated then to can be used as similar pictures.In addition it is also possible to using nearest neighbor algorithm into
Row is searched.Because when carrying out nearest neighbor algorithm calculating, the closer samples pictures of distance objective picture, the similarity with Target Photo
It is bigger, so, the several samples pictures closest with Target Photo can be taken as similar pictures.
Step 203, similar pictures are determined as doubtful marking error picture.
In the present embodiment, similar pictures can be determined as doubtful marking error picture by above-mentioned executing subject.Doubtful mark
Infuse the higher picture of a possibility that wrong picture is marking error.It is understood that herein, training sample is concentrated doubtful
The mark of marking error picture may be correct, it is also possible to mistake.
In some optional implementations of the present embodiment, Target Photo is that the training sample of train classification models is concentrated
Samples pictures.
In this implementation, Target Photo can be the samples pictures in training sample set, to disaggregated model into
Row training.Due to training sample concentrate picture have specific mark, can using training sample concentrate samples pictures as
Target Photo.The mistake of mark for the samples pictures that training sample is concentrated may be the same or similar, therefore be more advantageous to pair
The mistake that training sample is concentrated carries out batch corrigendum.
With continued reference to the schematic diagram that Fig. 3, Fig. 3 are according to the application scenarios of the image processing method of the present embodiment.?
In the application scenarios of Fig. 3, executing subject 301 can be inputted picture A 302 in disaggregated model 303 trained in advance.Execute master
Body 301 is inconsistent in response to the classification results " cat " for determining that the object Tiger that picture A is included is exported with disaggregated model, is dividing
The training sample of class model is concentrated, and the similar pictures X and similar pictures Y 304 of picture A are searched, wherein similar pictures X and similar
The similarity of picture Y and picture A is greater than the similarity that training sample concentrates other samples pictures Yu picture A；Similar pictures are true
It is set to doubtful marking error picture 305.
The method provided by the above embodiment of the application can filter out doubtful mark from the samples pictures of training sample set
Wrong picture is infused, reduces the quantity for marking error map piece in training sample set, improves the accuracy of model.
With further reference to Fig. 4, it illustrates the processes 400 of another embodiment of image processing method.Picture processing
The process 400 of method, includes the following steps：
Step 401, the disaggregated model that Target Photo input is trained in advance.
In the present embodiment, executing subject (such as server shown in FIG. 1 or the end of image processing method operation thereon
End equipment) Target Photo can be inputted to disaggregated model trained in advance.Above-mentioned disaggregated model can be two disaggregated models, can also
To be more disaggregated models, the object for being mainly used for being included to picture is classified.Target Photo refer to comprising some or it is certain
Any image of object.
It is step 402, inconsistent in response to the classification results for determining that the Target Photo object that is included and disaggregated model export,
Based on nearest neighbor algorithm, the similar pictures of Target Photo are searched from the picture of training sample set, wherein similar pictures and target figure
The similarity of piece is greater than the similarity that training sample concentrates other samples pictures and Target Photo.
It in the present embodiment, can be from disaggregated model after Target Photo is inputted disaggregated model by above-mentioned executing subject
Output in get classification results.Above-mentioned executing subject can be not belonging in response to the object for determining that Target Photo is included
The classification results for stating disaggregated model output are concentrated in the training sample of disaggregated model, search the similar pictures of Target Photo.
In some optional implementations of the present embodiment, it is based on nearest neighbor algorithm (k-Nearest Neighbor, kNN),
The similar pictures of Target Photo are searched from the picture of training sample set, including：
The characteristic information of Target Photo is obtained, the characteristic information that training sample concentrates each picture is obtained；
In hyperspace coordinate system, determine that the coordinate position of the characteristic information of Target Photo is target coordinate position, really
The coordinate position for determining the characteristic information of the picture of training sample concentration is coordinate position to be selected；
According to the sequence at a distance from target coordinate position from small to large, determined from each coordinate position to be selected default
Quantity coordinate position；
Picture corresponding to preset quantity coordinate position is determined as similar pictures.
In this implementation, characteristic information refers to the information for embodying picture feature.If disaggregated model is convolutional neural networks,
Characteristic information can be the information exported by full articulamentum, be embodied in the form of vector.The characteristic information of picture can be
It is presented in the form of coordinate points in hyperspace.Training sample concentrates the coordinate position of each picture and the coordinate bit of Target Photo
It the distance between sets and to be not quite similar.Picture is corresponding with the coordinate position of the characteristic information of the picture.Between coordinate position away from
From smaller, then the similarity between picture corresponding to the two coordinate positions is bigger.Therefore, can according to coordinates of targets position
The sequence of the distance set from small to large is ranked up each coordinate position to be selected.And the distance of the sequence obtained from sequence
Small one end determines coordinate position.
Step 403, similar pictures are determined as doubtful marking error picture.
In the present embodiment, similar pictures can be determined as doubtful marking error picture by above-mentioned executing subject.Doubtful mark
Infuse the higher picture of a possibility that wrong picture is marking error.It is understood that herein, training sample is concentrated doubtful
The mark of marking error picture may be correct, it is also possible to mistake.
Step 404, doubtful marking error picture is sent, determines whether doubtful marking error picture is marking error picture.
In the present embodiment, above-mentioned executing subject sends doubtful marking error picture to other electronic equipments, determines doubtful
Whether marking error picture is marking error picture.
In practice, whether above-mentioned executing subject can use various ways to determine doubtful marking error picture for mark mistake
Accidentally picture.For example, above-mentioned executing subject can send doubtful marking error picture to target terminal equipment, so that the target
Terminal device determines whether doubtful marking error picture is marking error picture.Specifically, target terminal equipment can will be doubtful
Marking error picture is shown to user.In this way, can judge the object and doubtful marking error figure that picture is included by user
Whether the mark of piece is consistent.If the result that user judges is inconsistent, target terminal can pass through the operation etc. of reception user
Form receives the judging result of user, and target terminal can then determine that doubtful marking error picture is marking error picture.This
Outside, above-mentioned executing subject can also show doubtful wrong picture.In this way, the user of above-mentioned executing subject is just it can be seen that this is doubtful
Mistake picture, and judge whether the doubtful wrong picture is marking error picture.Later, above-mentioned executing subject can utilize use
Family determines whether doubtful marking error picture is marking error picture to the judging result of local input.
In some optional implementations of the present embodiment, the above method further includes：Obtain amendment picture, wherein repair
Positive picture is the picture generated after being labeled corrigendum to marking error picture；
The marking error picture that training sample is concentrated is replaced with into amendment picture.
In this implementation, in the case where doubtful marking error picture is marking error picture, above-mentioned executing subject
Available amendment picture, and the replacement for carrying out samples pictures is concentrated in training sample.Specifically, the amendment picture got can
To be that above-mentioned target terminal equipment returns, it is also possible to what the user of above-mentioned executing subject inputted to the executing subject.Here
Be labeled corrigendum mean corrigendum picture mark.The process for correcting the mark of picture can be user and be set using target terminal
What standby or other equipment (such as above-mentioned executing subject) carried out.
The samples pictures that this implementation can be concentrated by replacement training sample, reduce error label, improve training sample
The accuracy of this concentration sample.
In the application scenes of above-mentioned implementation, the above method further includes：Based on the mark of amendment picture, again
Train classification models.
In the application scenarios, above-mentioned executing subject can further improve classification by re -training disaggregated model
The accuracy of model.
In the present embodiment, mark of the above-mentioned executing subject based on amendment picture, can be with the above-mentioned disaggregated model of re -training.
Specifically, above-mentioned executing subject can be using each picture as input, and the object for being included to amendment picture is marked
Type re-starts training to disaggregated model as output.
The present embodiment can be obtained and the most similar picture of Target Photo using nearest neighbor algorithm.And it can be wrong to mark
Accidentally the mark of picture is corrected, and in order to be trained the accuracy for improving model using the picture for having corrected mark, is reduced
The probability of model output error.
With further reference to Fig. 5, as the realization to method shown in above-mentioned each figure, this application provides a kind of processing of picture to fill
The one embodiment set, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which specifically can be applied to respectively
In kind electronic equipment.
As shown in figure 5, the picture processing unit 500 of the present embodiment includes：Taxon 501, searching unit 502 and determination
Unit 503.Wherein, taxon 501 are configured to inputting Target Photo into disaggregated model trained in advance；Searching unit
502, it is inconsistent to be configured in response to the classification results for determining that the Target Photo object that is included and disaggregated model export, and is dividing
The training sample of class model is concentrated, and searches the similar pictures of Target Photo, wherein the similarity of similar pictures and Target Photo is big
The similarity of other samples pictures and Target Photo is concentrated in training sample；Determination unit 503 is configured to similar pictures are true
It is set to doubtful marking error picture.
In some embodiments, Target Photo can be inputted training in advance by the taxon 501 of picture processing unit 500
Disaggregated model.Above-mentioned disaggregated model can be two disaggregated models, be also possible to more disaggregated models, be mainly used for wrapping picture
The object contained is classified.Target Photo refers to comprising any image of some or certain objects.
It in some embodiments, can be from disaggregated model after Target Photo is inputted disaggregated model by searching unit 502
Classification results are got in output.Above-mentioned executing subject can be in response to determining the Target Photo object that is included and above-mentioned classification
The classification results of model output are inconsistent, concentrate in the training sample of disaggregated model, search the similar pictures of Target Photo.
In some embodiments, similar pictures can be determined as doubtful marking error picture by determination unit 503.Doubtful mark
Infuse the picture that wrong picture is doubtful marking error, that is, the mark of doubtful marking error picture may be correctly, can also
It can be wrong.
In some optional implementations of the present embodiment, Target Photo is that the training sample of train classification models is concentrated
Samples pictures.
In some optional implementations of the present embodiment, searching unit 502 is further configured to：Obtain target figure
The characteristic information of piece obtains the characteristic information that training sample concentrates each picture；In hyperspace coordinate system, target figure is determined
The coordinate position of the characteristic information of piece is target coordinate position, determines the coordinate bit of the characteristic information for the picture that training sample is concentrated
It is set to coordinate position to be selected；According to the sequence at a distance from target coordinate position from small to large, from each coordinate bit to be selected
Set middle determining preset quantity coordinate position；Picture corresponding to preset quantity coordinate position is determined as similar pictures.
In some optional implementations of the present embodiment, picture processing unit further includes：Transmission unit is configured to
Doubtful marking error picture is sent, determines whether doubtful marking error picture is marking error picture.
In some optional implementations of the present embodiment, picture processing unit further includes：Acquiring unit is configured to
Obtain amendment picture, wherein amendment picture is the picture generated after being labeled corrigendum to marking error picture；Replacement is single
Member is configured to the marking error picture that training sample is concentrated replacing with amendment picture.
In some optional implementations of the present embodiment, picture processing unit further includes：Training unit is configured to
Based on the mark of amendment picture, re -training disaggregated model.
Below with reference to Fig. 6, it illustrates the computer systems 600 for the electronic equipment for being suitable for being used to realize the embodiment of the present application
Structural schematic diagram.Electronic equipment shown in Fig. 6 is only an example, function to the embodiment of the present application and should not use model
Shroud carrys out any restrictions.
As shown in fig. 6, computer system 600 includes central processing unit (CPU) 601, it can be read-only according to being stored in
Program in memory (ROM) 602 or be loaded into the program in random access storage device (RAM) 603 from storage section 608 and
Execute various movements appropriate and processing.In RAM 603, also it is stored with system 600 and operates required various programs and data.
CPU 601, ROM 602 and RAM 603 are connected with each other by bus 604.Input/output (I/O) interface 605 is also connected to always
Line 604.
I/O interface 605 is connected to lower component：Importation 606 including keyboard, mouse etc.；It is penetrated including such as cathode
The output par, c 607 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.；Storage section 608 including hard disk etc.；
And the communications portion 609 of the network interface card including LAN card, modem etc..Communications portion 609 via such as because
The network of spy's net executes communication process.Driver 610 is also connected to I/O interface 605 as needed.Detachable media 611, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 610, in order to read from thereon
Computer program be mounted into storage section 608 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 609, and/or from detachable media
611 are mounted.When the computer program is executed by central processing unit (CPU) 601, limited in execution the present processes
Above-mentioned function.It should be noted that the computer-readable medium of the application can be computer-readable signal media or calculating
Machine readable storage medium storing program for executing either the two any combination.Computer readable storage medium for example can be --- but it is unlimited
In system, device or the device of --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or any above combination.It calculates
The more specific example of machine readable storage medium storing program for executing can include but is not limited to：It is electrical connection with one or more conducting wires, portable
Formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable programmable read only memory
(EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device or
The above-mentioned any appropriate combination of person.In this application, computer readable storage medium can be it is any include or storage program
Tangible medium, which can be commanded execution system, device or device use or in connection.And in this Shen
Please in, computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
In carry computer-readable program code.The data-signal of this propagation can take various forms, including but not limited to
Electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be computer-readable
Any computer-readable medium other than storage medium, the computer-readable medium can send, propagate or transmit for by
Instruction execution system, device or device use or program in connection.The journey for including on computer-readable medium
Sequence code can transmit with any suitable medium, including but not limited to：Wirelessly, electric wire, optical cable, RF etc. or above-mentioned
Any appropriate combination.
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit also can be set in the processor, for example, can be described as：A kind of processor packet
Include taxon, searching unit and determination unit.Wherein, the title of these units is not constituted under certain conditions to the unit
The restriction of itself, for example, taxon is also described as " by the list of Target Photo input disaggregated model trained in advance
Member ".
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in device described in above-described embodiment；It is also possible to individualism, and without in the supplying device.Above-mentioned calculating
Machine readable medium carries one or more program, when said one or multiple programs are executed by the device, so that should
Device：By Target Photo input disaggregated model trained in advance；In response to determining the object that Target Photo is included and mould of classifying
The classification results of type output are inconsistent, concentrate in the training sample of disaggregated model, search the similar pictures of Target Photo, wherein
The similarity of similar pictures and Target Photo is greater than the similarity that training sample concentrates other samples pictures and Target Photo；By phase
It is determined as doubtful marking error picture like picture.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (14)
1. a kind of image processing method, including：
By Target Photo input disaggregated model trained in advance；
The object for being included in response to the determination Target Photo and the classification results that the disaggregated model exports are inconsistent, in institute
The training sample for stating disaggregated model is concentrated, and the similar pictures of the Target Photo are searched, wherein the similar pictures and the mesh
The similarity of piece of marking on a map is greater than the training sample and concentrates the similarities of other samples pictures Yu the Target Photo；
The similar pictures are determined as doubtful marking error picture.
2. according to the method described in claim 1, wherein, the Target Photo is the training sample set of the training disaggregated model
In samples pictures.
3. being searched according to the method described in claim 1, wherein, the training sample in the picture classification model is concentrated
The similar pictures of the input picture, including：
The characteristic information for obtaining the Target Photo obtains the characteristic information that the training sample concentrates each picture；
In hyperspace coordinate system, determine that the coordinate position of the characteristic information of the Target Photo is target coordinate position, really
The coordinate position of the characteristic information for the picture that the fixed training sample is concentrated is coordinate position to be selected；
According to the sequence at a distance from the target coordinate position from small to large, from each coordinate position to be selected described in determination
Preset quantity coordinate position；
Picture corresponding to the preset quantity coordinate position is determined as the similar pictures.
4. according to the method described in claim 1, wherein, the similar pictures are determined as doubtful marking error picture described
Later, the method also includes：
The doubtful marking error picture is sent, determines whether the doubtful marking error picture is marking error picture.
5. according to the method described in claim 4, wherein, it is described send the doubtful marking error picture after, the side
Method further includes：
Obtain amendment picture, wherein the amendment picture is to generate after being labeled corrigendum to the marking error picture
Picture；
The marking error picture that the training sample is concentrated replaces with the amendment picture.
6. according to the method described in claim 5, wherein, the method also includes：
Based on the mark of the amendment picture, disaggregated model described in re -training.
7. a kind of picture processing unit, including：
Taxon is configured to inputting Target Photo into disaggregated model trained in advance；
Searching unit is configured in response to determine point of the object that the Target Photo is included and disaggregated model output
Class result is inconsistent, concentrates in the training sample of the disaggregated model, searches the similar pictures of the Target Photo, wherein institute
The similarity for stating similar pictures and the Target Photo is greater than the other samples pictures of training sample concentration and the target figure
The similarity of piece；
Determination unit is configured to for the similar pictures to be determined as doubtful marking error picture.
8. device according to claim 7, wherein the Target Photo is the training sample set of the training disaggregated model
In samples pictures.
9. device according to claim 7, wherein the searching unit is further configured to：
The characteristic information for obtaining the Target Photo obtains the characteristic information that the training sample concentrates each picture；
In hyperspace coordinate system, determine that the coordinate position of the characteristic information of the Target Photo is target coordinate position, really
The coordinate position of the characteristic information for the picture that the fixed training sample is concentrated is coordinate position to be selected；
According to the sequence at a distance from the target coordinate position from small to large, from each coordinate position to be selected described in determination
Preset quantity coordinate position；
Picture corresponding to the preset quantity coordinate position is determined as the similar pictures.
10. device according to claim 7, wherein described device further includes：
Transmission unit is configured to send the doubtful marking error picture, determine the doubtful marking error picture whether be
Marking error picture.
11. device according to claim 10, wherein described device further includes：
Acquiring unit is configured to obtain amendment picture, wherein the amendment picture is to mark to the marking error picture
The picture generated after note corrigendum；
Replacement unit, the marking error picture for being configured to concentrate the training sample replace with the amendment picture.
12. device according to claim 11, wherein described device further includes：
Training unit is configured to the mark based on the amendment picture, disaggregated model described in re -training.
13. a kind of electronic equipment, including：
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
Now such as method as claimed in any one of claims 1 to 6.
14. a kind of computer readable storage medium, is stored thereon with computer program, wherein when the program is executed by processor
Realize such as method as claimed in any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810649563.1A CN108830329B (en) | 2018-06-22 | 2018-06-22 | Picture processing method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810649563.1A CN108830329B (en) | 2018-06-22 | 2018-06-22 | Picture processing method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108830329A true CN108830329A (en) | 2018-11-16 |
| CN108830329B CN108830329B (en) | 2020-07-10 |
Family
ID=64143294
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810649563.1A Active CN108830329B (en) | 2018-06-22 | 2018-06-22 | Picture processing method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108830329B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726755A (en) * | 2018-12-26 | 2019-05-07 | 北京云测信息技术有限公司 | A kind of picture mask method, device and electronic equipment |
| CN109948684A (en) * | 2019-03-12 | 2019-06-28 | 百度在线网络技术（北京）有限公司 | Quality detecting method, device and its relevant device of point cloud data mark quality |
| CN110110749A (en) * | 2019-03-29 | 2019-08-09 | 广州思德医疗科技有限公司 | Image processing method and device in a kind of training set |
| CN110399933A (en) * | 2019-07-31 | 2019-11-01 | 北京字节跳动网络技术有限公司 | Data mark modification method, device, computer-readable medium and electronic equipment |
| CN110414588A (en) * | 2019-07-23 | 2019-11-05 | 广东小天才科技有限公司 | Picture mask method, device, computer equipment and storage medium |
| CN110705596A (en) * | 2019-09-04 | 2020-01-17 | 北京三快在线科技有限公司 | White screen detection method and device, electronic equipment and storage medium |
| CN111143925A (en) * | 2019-12-18 | 2020-05-12 | 万翼科技有限公司 | Drawing labeling method and related product |
| CN111259200A (en) * | 2020-01-16 | 2020-06-09 | 北京达佳互联信息技术有限公司 | Video category classification method and device, electronic equipment and storage medium |
| CN111444945A (en) * | 2020-03-20 | 2020-07-24 | 北京每日优鲜电子商务有限公司 | Sample information filtering method and device, computer equipment and storage medium |
| CN112861962A (en) * | 2021-02-03 | 2021-05-28 | 北京百度网讯科技有限公司 | Sample processing method, sample processing device, electronic device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
| CN103971342A (en) * | 2014-05-21 | 2014-08-06 | 厦门美图之家科技有限公司 | Image noisy point detection method based on convolution neural network |
| CN105468760A (en) * | 2015-12-01 | 2016-04-06 | 北京奇虎科技有限公司 | Method and apparatus for labeling face images |
| CN106951484A (en) * | 2017-03-10 | 2017-07-14 | 百度在线网络技术（北京）有限公司 | Picture retrieval method and device, computer equipment and computer-readable medium |
-
2018
- 2018-06-22 CN CN201810649563.1A patent/CN108830329B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
| CN103971342A (en) * | 2014-05-21 | 2014-08-06 | 厦门美图之家科技有限公司 | Image noisy point detection method based on convolution neural network |
| CN105468760A (en) * | 2015-12-01 | 2016-04-06 | 北京奇虎科技有限公司 | Method and apparatus for labeling face images |
| CN106951484A (en) * | 2017-03-10 | 2017-07-14 | 百度在线网络技术（北京）有限公司 | Picture retrieval method and device, computer equipment and computer-readable medium |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726755A (en) * | 2018-12-26 | 2019-05-07 | 北京云测信息技术有限公司 | A kind of picture mask method, device and electronic equipment |
| CN109948684A (en) * | 2019-03-12 | 2019-06-28 | 百度在线网络技术（北京）有限公司 | Quality detecting method, device and its relevant device of point cloud data mark quality |
| CN110110749A (en) * | 2019-03-29 | 2019-08-09 | 广州思德医疗科技有限公司 | Image processing method and device in a kind of training set |
| CN110414588A (en) * | 2019-07-23 | 2019-11-05 | 广东小天才科技有限公司 | Picture mask method, device, computer equipment and storage medium |
| CN110399933A (en) * | 2019-07-31 | 2019-11-01 | 北京字节跳动网络技术有限公司 | Data mark modification method, device, computer-readable medium and electronic equipment |
| CN110399933B (en) * | 2019-07-31 | 2021-05-07 | 北京字节跳动网络技术有限公司 | Data annotation correction method and device, computer readable medium and electronic equipment |
| CN110705596A (en) * | 2019-09-04 | 2020-01-17 | 北京三快在线科技有限公司 | White screen detection method and device, electronic equipment and storage medium |
| CN111143925A (en) * | 2019-12-18 | 2020-05-12 | 万翼科技有限公司 | Drawing labeling method and related product |
| CN111259200A (en) * | 2020-01-16 | 2020-06-09 | 北京达佳互联信息技术有限公司 | Video category classification method and device, electronic equipment and storage medium |
| CN111444945A (en) * | 2020-03-20 | 2020-07-24 | 北京每日优鲜电子商务有限公司 | Sample information filtering method and device, computer equipment and storage medium |
| CN112861962A (en) * | 2021-02-03 | 2021-05-28 | 北京百度网讯科技有限公司 | Sample processing method, sample processing device, electronic device and storage medium |
| CN112861962B (en) * | 2021-02-03 | 2024-04-09 | 北京百度网讯科技有限公司 | Sample processing method, device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108830329B (en) | 2020-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108830329A (en) | Image processing method and device | |
| CN108898185A (en) | Method and apparatus for generating image recognition model | |
| CN108446387A (en) | Method and apparatus for updating face registration library | |
| CN107919129A (en) | Method and apparatus for controlling the page | |
| CN108984399A (en) | Detect method, electronic equipment and the computer-readable medium of interface difference | |
| CN109308681A (en) | Image processing method and device | |
| CN108830235A (en) | Method and apparatus for generating information | |
| CN109063653A (en) | Image processing method and device | |
| CN109903112A (en) | Information output method and device | |
| CN108628830A (en) | A kind of method and apparatus of semantics recognition | |
| CN109242801A (en) | Image processing method and device | |
| CN109344762A (en) | Image processing method and device | |
| CN109947989A (en) | Method and apparatus for handling video | |
| CN109377508A (en) | Image processing method and device | |
| CN108280422A (en) | The method and apparatus of face for identification | |
| CN108960110A (en) | Method and apparatus for generating information | |
| CN109214501A (en) | The method and apparatus of information for identification | |
| CN109118456A (en) | Image processing method and device | |
| CN108182457A (en) | For generating the method and apparatus of information | |
| CN109684624A (en) | A kind of method and apparatus in automatic identification Order Address road area | |
| CN109389182A (en) | Method and apparatus for generating information | |
| CN110516099A (en) | Image processing method and device | |
| CN111160410B (en) | Object detection method and device | |
| CN110335061B (en) | Transaction mode portrait establishing method, device, medium and electronic equipment | |
| CN110633717A (en) | Training method and device for target detection model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee after: Tiktok vision (Beijing) Co.,Ltd. Address before: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee before: BEIJING BYTEDANCE NETWORK TECHNOLOGY Co.,Ltd. Address after: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee after: Douyin Vision Co.,Ltd. Address before: 100041 B-0035, 2 floor, 3 building, 30 Shixing street, Shijingshan District, Beijing. Patentee before: Tiktok vision (Beijing) Co.,Ltd. |
|
| CP01 | Change in the name or title of a patent holder |