CN109145991A - Image group generation method, image group generating means and electronic equipment - Google Patents
Image group generation method, image group generating means and electronic equipment Download PDFInfo
- Publication number
- CN109145991A CN109145991A CN201810973553.3A CN201810973553A CN109145991A CN 109145991 A CN109145991 A CN 109145991A CN 201810973553 A CN201810973553 A CN 201810973553A CN 109145991 A CN109145991 A CN 109145991A
- Authority
- CN
- China
- Prior art keywords
- image
- value
- similarity
- image group
- foreign peoples
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
Disclose a kind of image group generation method, image group generating means and electronic equipment.Described image group generation method includes: the first image that first category is obtained from data set；The class label of the image collected based on the data obtains the second image of the first category；Calculate the similarity of each image and the first image that foreign peoples's image of the first category is not belonging in the data set；And have the third image of larger similarity to form image group from foreign peoples's image selection and the first image based on the similarity.In this way, image recognition model can be made quickly and to effectively to restrain by having foreign peoples's image construction image group of larger similarity based on the similarity selection between image.
Description
Technical field
This application involves field of image processings, and more specifically, are related to a kind of image group generation method, image group generates
Device and electronic equipment.
Background technique
Image recognition refers to be handled image, analyzed and is understood using computer, to identify various different modes
The technology of target and object.
With the fast development of artificial intelligence technology, image recognition becomes a key areas of artificial intelligence technology.Example
Such as, pedestrian identifies that (Person re-identification) is referred to from from non-overlapping multiple camera field of view again
Target pedestrian is identified in pedestrian image library or video flowing.Here, different from pedestrian tracking common under single camera, Hang Renzai
Identification can realize the long-term follow and monitoring to specific pedestrian under different background environments and multi-cam setting, therefore
It has very big application prospect in monitoring field.The technology is widely used in intelligent video monitoring, intelligent security etc. at present
Field.
In artificial intelligence field, image recognition model is instructed by choosing the image group being made of multiple images
Practice, therefore, selected image group will the training process directly to image recognition model have an impact.
It is therefore desirable to be able to which the improved image group for being suitable for the training of image recognition model generates scheme.
Summary of the invention
In order to solve the above-mentioned technical problem, the application is proposed.Embodiments herein provides a kind of image group generation
Method, image group generating means and electronic equipment, by the way that there is larger similarity based on the similarity selection between image
Foreign peoples's image construction image group, can make image recognition model quickly and to effectively restrain.
According to the one aspect of the application, a kind of image group generation method is provided, comprising: first is obtained from data set
First image of classification；The class label of the image collected based on the data obtains the second image of the first category；It calculates
Each image of foreign peoples's image of the first category and the similarity of the first image are not belonging in the data set；With
And have the third image of larger similarity with shape from foreign peoples's image selection and the first image based on the similarity
At image group.
In above-mentioned image group generation method, the class label of the image collected based on the data obtains the first category
The second image include: that the similar image for belonging to the first category in the data set is determined based on the class label；With
And second image is randomly choosed from the similar image.
In above-mentioned image group generation method, the similarity is based on from foreign peoples's image selection and the first image
Third image with larger similarity includes: by foreign peoples's image according to the phase with the first image to form image group
It is sorted from large to small like degree；It is the normal state of fixed value that acquisition initial mean value, which is picture number, the initial variance of foreign peoples's image,
Distribution function；Obtain the first function value of each integral point of the normal distyribution function between zero to the initial mean value；
And foreign peoples's image after the first function value alternatively sorts as the probability value of the third image to select
Select the third image.
In above-mentioned image group generation method, the similarity is based on from foreign peoples's image selection and the first image
Third image with larger similarity further comprises to form image group: the initial mean value of the normal distyribution function is subtracted
Small is change mean；Obtain the second function of each integral point of the normal distyribution function between zero to the initial mean value
Value；And foreign peoples's image after the second function value alternatively sorts is as the probability value of the third image
To select the third image.
In above-mentioned image group generation method, the similarity is based on from foreign peoples's image selection and the first image
Third image with larger similarity further comprises to form image group: determining whether the change mean is zero；Response
It is zero in the change mean, the initial variance is reduced to change variance；The normal distyribution function is obtained zero to institute
State the third functional value of each integral point between initial mean value；And after the third functional value is alternatively sorted
Foreign peoples's image selects the third image as the probability value of the third image.
In above-mentioned image group generation method, the image in the data set is pedestrian image.
In above-mentioned image group generation method, described image group is for training pedestrian's weight identification model.
According to the another aspect of the application, a kind of image group generating means are provided, comprising: first acquisition unit is used for
The first image of first category is obtained from data set；Second acquisition unit, the classification of the image for collecting based on the data
Label obtains the second image of the first category；Similarity calculated, for calculate be not belonging in the data set it is described
Each image of foreign peoples's image of first category and the similarity of the first image；And image group forms unit, is used for base
Have the third image of larger similarity to form figure from foreign peoples's image selection and the first image in the similarity
As group.
In above-mentioned image group generating means, the second acquisition unit is used for: based on described in class label determination
Belong to the similar image of the first category in data set；And second image is randomly choosed from the similar image.
In above-mentioned image group generating means, described image group formed unit be used for: by foreign peoples's image according to institute
The similarity for stating the first image sorts from large to small；Obtain picture number, the initial variance that initial mean value is foreign peoples's image
For the normal distyribution function of fixed value；Obtain each integral point of the normal distyribution function between zero to the initial mean value
First function value；And foreign peoples's image after the first function value alternatively sorts is as the third figure
The probability value of picture is to select the third image.
In above-mentioned image group generating means, described image group forms unit and is further used for: by the normal distribution letter
Several initial mean values are reduced to change mean；It is whole each of between zero to the initial mean value to obtain the normal distyribution function
The second function value of several points；And foreign peoples's image after the second function value alternatively sorts is as described
The probability value of three images is to select the third image.
In above-mentioned image group generating means, described image group forms unit and is further used for: determining the change mean
It whether is zero；It is zero in response to the change mean, the initial variance is reduced to change variance；Obtain the normal distribution
The third functional value of each integral point of the function between zero to the initial mean value；And using the third functional value as
Foreign peoples's image after selected and sorted selects the third image as the probability value of the third image.
In above-mentioned image group generating means, the image in the data set is pedestrian image.
In above-mentioned image group generating means, described image group is for training pedestrian's weight identification model.
According to the application's in another aspect, providing a kind of electronic equipment, comprising: processor；And memory, in institute
It states and is stored with computer program instructions in memory, the computer program instructions make described when being run by the processor
Processor executes image group generation method as described above.
According to the another aspect of the application, a kind of computer-readable medium is provided, computer program is stored thereon with and refers to
It enables, the computer program instructions make the processor execute image group as described above generation side when being run by processor
Method.
Compared with prior art, image group generation method, image group generating means and electronics provided by the embodiments of the present application
Equipment can obtain the first image of first category from data set；The class label of the image collected based on the data obtains
Second image of the first category；Calculate each image that foreign peoples's image of the first category is not belonging in the data set
With the similarity of the first image；And had based on the similarity from foreign peoples's image selection and the first image
There is the third image of larger similarity to form image group.In this way, larger by being had based on the similarity selection between image
Foreign peoples's image construction image group of similarity, can make image recognition model quickly and to effectively restrain.
Detailed description of the invention
The embodiment of the present application is described in more detail in conjunction with the accompanying drawings, the above-mentioned and other purposes of the application,
Feature and advantage will be apparent.Attached drawing is used to provide to further understand the embodiment of the present application, and constitutes explanation
A part of book is used to explain the application together with the embodiment of the present application, does not constitute the limitation to the application.In the accompanying drawings,
Identical reference label typically represents same parts or step.
Fig. 1 illustrates the schematic diagrames according to the application scenarios of the image group generation method of the embodiment of the present application.
Fig. 2 illustrates the flow chart of the image group generation method according to the embodiment of the present application.
Fig. 3 illustrates the flow chart with normal distribution probability selection third image according to the embodiment of the present application.
Fig. 4 illustrates the process of the process of the difficulty for stepping up selected third image according to the embodiment of the present application
Figure.
Fig. 5 illustrates the block diagram of the image group generating means according to the embodiment of the present application.
Fig. 6 illustrates the block diagram of the electronic device according to the embodiment of the present application.
Specific embodiment
In the following, example embodiment according to the application will be described in detail by referring to the drawings.Obviously, described embodiment is only
It is only a part of the embodiment of the application, rather than the whole embodiments of the application, it should be appreciated that the application is not by described herein
The limitation of example embodiment.
Application is summarized
As described above, for image recognition model used in artificial intelligence field, need using acquired image as
Training set is trained image recognition model, reuses trained image recognition model then to carry out image recognition.
For example, in order to identify to pedestrian neural network need to be passed through using collected pedestrian's data as training set
(Neural Network, NN) trains pedestrian's weight identification model.Then, using the pedestrian's weight identification model pair trained
Pedestrian's picture extracts feature, and the similarity based on pedestrian's picture feature Yu target pedestrian picture feature, from corresponding pedestrian
Target pedestrian is identified in database.
Specifically, in order to train pedestrian's weight identification model, three images are usually chosen as a triple, currently most
Popular selection method is online difficult specimen sample (OHEM), is carried out in the sample being most difficult to of choosing at the very start of network training
Training.But because in this case, it is lower in the resolution for the three width images that the initial stage of model training is chosen, it is easy to lead
Cause model training failure.
In view of the above technical problems, the basic conception of the application be model training at the beginning when selection difficulty it is lower
Sample is to e-learning.Here, whether the difficulty of sample can be by being easy to differentiate and defining, that is, the feature between image is similar
The property the high then easier, on the contrary then be more difficult to.
Specifically, this application provides a kind of image group generation method, image group generating means and electronic equipments, first
The first image that first category is obtained from data set, the class label for being then based on the image of the data set obtain described the
A kind of other second image calculates each image that foreign peoples's image of the first category is not belonging in the data set later
With the similarity of the first image, and based on the similarity from foreign peoples's image selection and the first image have compared with
The third image of big similarity is to form image group.In this way, larger similar by being had based on the similarity selection between image
Foreign peoples's image construction image group of degree, can make image recognition model quickly and to effectively restrain.
Here, it will be understood by those skilled in the art that according to the image group of the embodiment of the present application can be comprising it is various to
Object is identified, for example, the group of the image of pedestrian, vehicle, special article etc..Also, correspondingly, described image identification model to
It identifies above-mentioned object to be identified, is not limited solely to pedestrian's weight identification model above-mentioned.But no matter for which kind of image recognition mould
Type can use and generate the lower image group of Scheme Choice difficulty according to the image group of the embodiment of the present application so that image is known
The training of other model is more easier.
After describing the basic principle of the application, carry out the various non-limits for specifically introducing the application below with reference to the accompanying drawings
Property embodiment processed.
Exemplary system
Fig. 1 illustrates the schematic diagrames according to the application scenarios of the image group generation method of the embodiment of the present application.
As shown in Figure 1, for image recognition model 100, choosing three images 101,102 and 103 conducts in the training stage
One triple is trained described image identification model 100.
Then, in cognitive phase, described image identification model 100 identifies input picture 104, and exports image knowledge
Other result.
In the following, will illustrate how to choose the lower image triple of difficulty to instruct to described image identification model 100
Practice.
Illustrative methods
Fig. 2 illustrates the flow chart of the image group generation method according to the embodiment of the present application.
As shown in Fig. 2, the image group generation method according to the embodiment of the present application includes: S210, is obtained from data set
A kind of other first image；S220, the class label of the image collected based on the data obtain the second figure of the first category
Picture；S230 calculates each image and the first image that foreign peoples's image of the first category is not belonging in the data set
Similarity；And S240, based on the similarity from foreign peoples's image selection and the first image with larger similar
The third image of degree is to form image group.
In step S210, the first image of first category is obtained from data set.Here, according to the embodiment of the present application
Image group generation method in, be primarily directed to the full supervision scene of label.That is, every in the data set
A image all has the label for classification, to determine classification that described image belongs to.For the data set of pedestrian image,
In each pedestrian image there is corresponding ID, that is, belonging to has identical ID with the image of a group traveling together, and belongs to different pedestrians
Image have different ID, therefore, it is possible to judge whether image belongs to same a group traveling together according to ID.
In step S220, the class label of the image collected based on the data obtains the second figure of the first category
Picture.It is, for class label possessed by each image in the data set, for example, the ID of pedestrian image, selection
Belong to the second image with the first image the same category, for example, another pedestrian image with the same ID of the first image, to form figure
As binary group.
Here, in the image group generation method according to the embodiment of the present application, it is true that the class label can be primarily based on
Belong to same category of similar image with the first image in the fixed data set, then is randomly choosed from the similar image
Second image.
In step S230, calculate be not belonging in the data set each image of foreign peoples's image of the first category with
The similarity of the first image.It specifically, can be true while the similar image as described above for determining the first image
Foreign peoples's image of first image described in the fixed data set, and calculate each image and the first image in these images
Similarity.For example, the similarity between image can be calculated in the distance in feature space by calculating image, wherein
The distance between described image is bigger, and similarity is smaller, conversely, distance therebetween is smaller, similarity is bigger.
In step S240, larger phase is had from foreign peoples's image selection and the first image based on the similarity
Like the third image of degree to form image group.
Here, it will be understood by those skilled in the art that can simply by foreign peoples's image according to first figure
The similarity of picture carries out sequence from big to small, and then selection comes the image of front as third image.
In addition, normal distyribution function value alternatively probability can be used in order to increase the randomness of image selection, thus
So that each image has the opportunity to be chosen to, the probability that only the lower image of difficulty is sampled is larger.
Fig. 3 illustrates the flow chart with normal distribution probability selection third image according to the embodiment of the present application.Such as Fig. 3 institute
Show, in the image group generation method according to the embodiment of the present application, is based on the similarity from foreign peoples's image selection and institute
Stating third image of first image with larger similarity to form image group includes: S310, by foreign peoples's image according to
The similarity of the first image sorts from large to small；S320, obtain initial mean value be foreign peoples's image picture number, just
Beginning variance is the normal distyribution function of fixed value；S330 obtains the normal distyribution function between zero to the initial mean value
Each integral point first function value；And S340, the foreign peoples figure after the first function value is alternatively sorted
Picture is as the probability value of the third image to select the third image.
In this way, the general trend of image selection is still that the lower image of difficulty is preferential, but simultaneously, selects the increasing of randomness
Adding can preferably be trained with assisted image recognition model, so that the image recognition model finally obtained is more robust.
In addition, it is further introduced into the thought of course formula sampling in the image group generation method according to the embodiment of the present application,
With the increase of network the number of iterations, the ability of image recognition model enhances, and is stepped up the difficulty of the sample of selection.
In the example of above-mentioned normal distyribution function, by gradually reduce mean value and the variance of the normal distyribution function come
Improve the difficulty of selected image.
Fig. 4 illustrates the process of the process of the difficulty for stepping up selected third image according to the embodiment of the present application
Figure.As shown in figure 4, being based on the similarity from the foreign peoples in the image group generation method according to the embodiment of the present application
After the third image of image selection and the first image with larger similarity is to form ternary image group, further wrap
It includes: S410 and the initial mean value of the normal distyribution function is reduced to change mean；S420 obtains the normal distyribution function
The second function value of each integral point between zero to the initial mean value；And S430, using the second function value as
Foreign peoples's image after selected and sorted selects the third image as the probability value of the third image.
That is, third image selected before reducing compared to mean value, mean value third image selected after reducing
Difficulty increase.
Also, after the third image after having selected mean value to reduce is to form new image triple, with image
The increase of the training time of identification model further increases the difficulty of selected image.As shown in figure 4, reducing the normal state
After the mean value of distribution function, in step S440, determine whether the change mean is zero, also, equal in response to the variation
Value is not zero, and returns to the step S420；It is zero in response to the change mean in step S450, by the normal distribution
The initial variance of function is reduced to change variance；In step S460, obtain the normal distyribution function zero to it is described it is initial
The third functional value of each integral point between value；And in step S470, after the third functional value is alternatively sorted
Foreign peoples's image the third image is selected as the probability value of the third image.
Therefore, at the initial stage of image recognition model training, simple sample is selected, and the phase after training, then select difficult sample
This.
Exemplary means
Fig. 5 illustrates the block diagram of the image group generating means according to the embodiment of the present application.
As shown in figure 5, the image group generating means 500 according to the embodiment of the present application include: first acquisition unit 510, use
In the first image for obtaining first category from data set；Second acquisition unit 520, image for collecting based on the data
Class label obtains the second image of the first category；Similarity calculated 530 does not belong to for calculating in the data set
In each image of foreign peoples's image of the first category and the similarity of the first image；And image group forms unit
540, for having the third figure of larger similarity from foreign peoples's image selection and the first image based on the similarity
As to form image group.
In one example, in above-mentioned image group generating means 500, the second acquisition unit 520 is used for: being based on institute
It states class label and determines the similar image for belonging to the first category in the data set；And from the similar image with
Machine selects second image.
In one example, in above-mentioned image group generating means 500, described image group forms unit 540 and is used for: by institute
Foreign peoples's image is stated to sort from large to small according to the similarity with the first image；Obtaining initial mean value is foreign peoples's image
Picture number, the normal distyribution function that initial variance is fixed value；Obtain the normal distyribution function zero to it is described it is initial
The first function value of each integral point between value；And the foreign peoples after the first function value alternatively sorts
Image selects the third image as the probability value of the third image.
In one example, in above-mentioned image group generating means 500, described image group forms unit 540 and further uses
In: the initial mean value of the normal distyribution function is reduced to change mean；The normal distyribution function is obtained zero to described
The second function value of each integral point between initial mean value；And the institute after the second function value alternatively sorts
Foreign peoples's image is stated as the probability value of the third image to select the third image.
In one example, in above-mentioned image group generating means 500, described image group forms unit 540 and further uses
In: determine whether the change mean is zero；It is zero in response to the change mean, the initial variance is reduced to variation side
Difference；Obtain the third functional value of each integral point of the normal distyribution function between zero to the initial mean value；And it will
Foreign peoples's image after the third functional value alternatively sorts is described to select as the probability value of the third image
Third image.
In one example, in above-mentioned image group generating means 500, the image in the data set is pedestrian image.
In one example, in above-mentioned image group generating means 500, described image group is for training pedestrian to identify mould again
Type.
Here, it will be understood by those skilled in the art that each unit and module in above-mentioned image group generating means 500
Concrete function and operation have been described above in the image group generation method with reference to figs. 2 to Fig. 4 description and are discussed in detail, and therefore, will
Omit its repeated description.
As described above, the image group generating means 500 according to the embodiment of the present application may be implemented in various terminal equipment,
Such as in the automatic Pilot auxiliary system of identification pedestrian.In one example, according to the image group generating means of the embodiment of the present application
500 can be used as a software module and/or hardware module and be integrated into the terminal device.For example, the device 500 can be with
It is a software module in the operating system of the terminal device, or can be and be directed to one that the terminal device is developed
Application program；Certainly, which equally can be one of numerous hardware modules of the terminal device.
Alternatively, in another example, which is also possible to discrete set with the terminal device
It is standby, and the device 500 can be connected to the terminal device by wired and or wireless network, and according to the data of agreement
Format transmits interactive information.
Example electronic device
In the following, being described with reference to Figure 6 the electronic equipment according to the embodiment of the present application.
Fig. 6 illustrates the block diagram of the electronic equipment according to the embodiment of the present application.
As shown in fig. 6, electronic equipment 10 includes one or more processors 11 and memory 12.
Processor 11 can be central processing unit (CPU) or have data-handling capacity and/or instruction execution capability
Other forms processing unit, and can control the other assemblies in electronic equipment 10 to execute desired function.
Memory 12 may include one or more computer program products, and the computer program product may include each
The computer readable storage medium of kind form, such as volatile memory and/or nonvolatile memory.The volatile storage
Device for example may include random access memory (RAM) and/or cache memory (cache) etc..It is described non-volatile to deposit
Reservoir for example may include read-only memory (ROM), hard disk, flash memory etc..It can be deposited on the computer readable storage medium
One or more computer program instructions are stored up, processor 11 can run described program instruction, to realize this Shen described above
The image group generation method of each embodiment please and/or other desired functions.In the computer-readable storage medium
The various contents such as image binary group, image similarity, image triple can also be stored in matter.
In one example, electronic equipment 10 can also include: input unit 13 and output device 14, these components pass through
The interconnection of bindiny mechanism's (not shown) of bus system and/or other forms.
For example, the input unit 13 can be the image collecting device for shooting image, such as camera etc..In addition,
The input unit 13 can also include such as keyboard, mouse etc..
The output device 14 can be output to the outside various information, including image triple generated etc..Output dress
Setting 14 may include such as display, loudspeaker, printer and communication network and its remote output devices connected
Deng.
Certainly, to put it more simply, illustrated only in Fig. 6 it is some in component related with the application in the electronic equipment 10,
The component of such as bus, input/output interface etc. is omitted.In addition to this, according to concrete application situation, electronic equipment 10 is also
It may include any other component appropriate.
Illustrative computer program product and computer readable storage medium
Other than the above method and equipment, embodiments herein can also be computer program product comprising meter
Calculation machine program instruction, it is above-mentioned that the computer program instructions make the processor execute this specification when being run by processor
According to the step in the image group generation method of the various embodiments of the application described in " illustrative methods " part.
The computer program product can be write with any combination of one or more programming languages for holding
The program code of row the embodiment of the present application operation, described program design language includes object oriented program language, such as
Java, C++ etc. further include conventional procedural programming language, such as " C " language or similar programming language.Journey
Sequence code can be executed fully on the user computing device, partly execute on a user device, be independent soft as one
Part packet executes, part executes on a remote computing or completely in remote computing device on the user computing device for part
Or it is executed on server.
In addition, embodiments herein can also be computer readable storage medium, it is stored thereon with computer program and refers to
It enables, the computer program instructions make the processor execute above-mentioned " the exemplary side of this specification when being run by processor
According to the step in the image group generation method of the various embodiments of the application described in method " part.
The computer readable storage medium can be using any combination of one or more readable mediums.Readable medium can
To be readable signal medium or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can include but is not limited to electricity, magnetic, light, electricity
Magnetic, the system of infrared ray or semiconductor, device or device, or any above combination.Readable storage medium storing program for executing it is more specific
Example (non exhaustive classification) includes: the electrical connection with one or more conducting wires, portable disc, hard disk, random access memory
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 above-mentioned any appropriate combination.
The basic principle of the application is described in conjunction with specific embodiments above, however, it is desirable to, it is noted that in this application
The advantages of referring to, advantage, effect etc. are only exemplary rather than limitation, must not believe that these advantages, advantage, effect etc. are the application
Each embodiment is prerequisite.In addition, detail disclosed above is merely to exemplary effect and the work being easy to understand
With, rather than limit, it is that must be realized using above-mentioned concrete details that above-mentioned details, which is not intended to limit the application,.
Device involved in the application, device, equipment, system block diagram only as illustrative example and be not intended to
It is required that or hint must be attached in such a way that box illustrates, arrange, configure.As those skilled in the art will appreciate that
, it can be connected by any way, arrange, configure these devices, device, equipment, system.Such as "include", "comprise", " tool
" etc. word be open vocabulary, refer to " including but not limited to ", and can be used interchangeably with it.Vocabulary used herein above
"or" and "and" refer to vocabulary "and/or", and can be used interchangeably with it, unless it is not such that context, which is explicitly indicated,.Here made
Vocabulary " such as " refers to phrase " such as, but not limited to ", and can be used interchangeably with it.
It may also be noted that each component or each step are can to decompose in the device of the application, device and method
And/or reconfigure.These decompose and/or reconfigure the equivalent scheme that should be regarded as the application.
The above description of disclosed aspect is provided so that any person skilled in the art can make or use this
Application.Various modifications in terms of these are readily apparent to those skilled in the art, and are defined herein
General Principle can be applied to other aspect without departing from scope of the present application.Therefore, the application is not intended to be limited to
Aspect shown in this, but according to principle disclosed herein and the consistent widest range of novel feature.
In order to which purpose of illustration and description has been presented for above description.In addition, this description is not intended to the reality of the application
It applies example and is restricted to form disclosed herein.Although already discussed above multiple exemplary aspects and embodiment, this field skill
Its certain modifications, modification, change, addition and sub-portfolio will be recognized in art personnel.
Claims (13)
1. a kind of image group generation method, comprising:
The first image of first category is obtained from data set；
The class label of the image collected based on the data obtains the second image of the first category；
Calculate each image and the phase of the first image that foreign peoples's image of the first category is not belonging in the data set
Like degree；And
Based on the similarity from foreign peoples's image selection and the first image have the third image of larger similarity with
Form image group.
2. image group generation method as described in claim 1, wherein the class label of the image collected based on the data obtains
Second image of the first category includes:
The similar image for belonging to the first category in the data set is determined based on the class label；And
Second image is randomly choosed from the similar image.
3. image group generation method as described in claim 1, wherein based on the similarity from foreign peoples's image selection with
The first image have larger similarity third image include: to form image group
Foreign peoples's image is sorted from large to small according to the similarity with the first image；
It is the normal distyribution function of fixed value that acquisition initial mean value, which is picture number, the initial variance of foreign peoples's image,；
Obtain the first function value of each integral point of the normal distyribution function between zero to the initial mean value；And
Foreign peoples's image after the first function value is alternatively sorted is as the probability value of the third image to select
Select the third image.
4. image group generation method as claimed in claim 3, wherein based on the similarity from foreign peoples's image selection with
Third image of the first image with larger similarity further comprises to form image group:
The initial mean value of the normal distyribution function is reduced to change mean；
Obtain the second function value of each integral point of the normal distyribution function between zero to the initial mean value；And
Foreign peoples's image after the second function value is alternatively sorted is as the probability value of the third image to select
Select the third image.
5. image group generation method as claimed in claim 4, wherein based on the similarity from foreign peoples's image selection with
Third image of the first image with larger similarity further comprises to form image group:
Determine whether the change mean is zero；
It is zero in response to the change mean, the initial variance is reduced to change variance；
Obtain the third functional value of each integral point of the normal distyribution function between zero to the initial mean value；And
Foreign peoples's image after the third functional value is alternatively sorted is as the probability value of the third image to select
Select the third image.
6. the image group generation method as described in any one of claims 1 to 5, wherein the image in the data set is
Pedestrian image.
7. image group generation method as claimed in claim 6, wherein described image group is for training pedestrian's weight identification model.
8. a kind of image group generating means, comprising:
First acquisition unit, for obtaining the first image of first category from data set；
Second acquisition unit, the class label of the image for collecting based on the data obtain the second figure of the first category
Picture；
Similarity calculated, for calculating each image for being not belonging to foreign peoples's image of the first category in the data set
With the similarity of the first image；And
Image group forms unit, larger for being had from foreign peoples's image selection with the first image based on the similarity
The third image of similarity is to form image group.
9. image group generating means as claimed in claim 8, wherein described image group forms unit and is used for:
Foreign peoples's image is sorted from large to small according to the similarity with the first image；
It is the normal distyribution function of fixed value that acquisition initial mean value, which is picture number, the initial variance of foreign peoples's image,；
Obtain the first function value of each integral point of the normal distyribution function between zero to the initial mean value；And
Foreign peoples's image after the first function value is alternatively sorted is as the probability value of the third image to select
Select the third image.
10. image group generating means as claimed in claim 9, wherein described image group forms unit and is further used for:
The initial mean value of the normal distyribution function is reduced to change mean；
Obtain the second function value of each integral point of the normal distyribution function between zero to the initial mean value；And
Foreign peoples's image after the second function value is alternatively sorted is as the probability value of the third image to select
Select the third image.
11. image group generating means as claimed in claim 10, wherein described image group forms unit and is further used for:
Determine whether the change mean is zero；
It is zero in response to the change mean, the initial variance is reduced to change variance；
Obtain the third functional value of each integral point of the normal distyribution function between zero to the initial mean value；And
Foreign peoples's image after the third functional value is alternatively sorted is as the probability value of the third image to select
Select the third image.
12. a kind of electronic equipment, comprising:
Processor；And
Memory is stored with computer program instructions in the memory, and the computer program instructions are by the processing
Device makes the processor execute such as image group generation method of any of claims 1-7 when running.
13. a kind of computer-readable medium is stored thereon with computer program instructions, the computer program instructions are processed
Device makes the processor execute the image group generation method as described in any one of claim 1-7 when running.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810973553.3A CN109145991B (en) | 2018-08-24 | 2018-08-24 | Image group generation method, image group generation device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810973553.3A CN109145991B (en) | 2018-08-24 | 2018-08-24 | Image group generation method, image group generation device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109145991A true CN109145991A (en) | 2019-01-04 |
| CN109145991B CN109145991B (en) | 2020-07-31 |
Family
ID=64827763
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810973553.3A Active CN109145991B (en) | 2018-08-24 | 2018-08-24 | Image group generation method, image group generation device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145991B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112863132A (en) * | 2021-04-23 | 2021-05-28 | 成都中轨轨道设备有限公司 | Natural disaster early warning system and early warning method |
| CN114205642A (en) * | 2020-08-31 | 2022-03-18 | 北京金山云网络技术有限公司 | Video image processing method and device |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1992911A (en) * | 2005-12-31 | 2007-07-04 | 中国科学院计算技术研究所 | Target tracking method of sports video |
| CN101873605A (en) * | 2010-05-27 | 2010-10-27 | 重庆邮电大学 | Adaptive method for classifying communication environments in network planning |
| CN103279746A (en) * | 2013-05-30 | 2013-09-04 | 苏州大学 | Method and system for identifying faces based on support vector machine |
| US20150161485A1 (en) * | 2009-12-15 | 2015-06-11 | Google Inc. | Learning Semantic Image Similarity |
| CN105975959A (en) * | 2016-06-14 | 2016-09-28 | 广州视源电子科技股份有限公司 | Face feature extraction modeling and face recognition method and device based on neural network |
| CN106202329A (en) * | 2016-07-01 | 2016-12-07 | 北京市商汤科技开发有限公司 | Sample data process, data identification method and device, computer equipment |
| CN107341497A (en) * | 2016-11-11 | 2017-11-10 | 东北大学 | The unbalanced weighting data streams Ensemble classifier Forecasting Methodology of sampling is risen with reference to selectivity |
| CN107577987A (en) * | 2017-08-01 | 2018-01-12 | 广州广电卓识智能科技有限公司 | Identity authentication method, system and device |
| CN107704890A (en) * | 2017-10-27 | 2018-02-16 | 北京旷视科技有限公司 | A kind of generation method and device of four-tuple image |
| CN108009528A (en) * | 2017-12-26 | 2018-05-08 | 广州广电运通金融电子股份有限公司 | Face authentication method, device, computer equipment and storage medium based on Triplet Loss |
| CN108108754A (en) * | 2017-12-15 | 2018-06-01 | 北京迈格威科技有限公司 | The training of identification network, again recognition methods, device and system again |
| CN108304864A (en) * | 2018-01-17 | 2018-07-20 | 清华大学 | Depth fights metric learning method and device |
| WO2018148493A1 (en) * | 2017-02-09 | 2018-08-16 | Painted Dog, Inc. | Methods and apparatus for detecting, filtering, and identifying objects in streaming video |
-
2018
- 2018-08-24 CN CN201810973553.3A patent/CN109145991B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1992911A (en) * | 2005-12-31 | 2007-07-04 | 中国科学院计算技术研究所 | Target tracking method of sports video |
| US20150161485A1 (en) * | 2009-12-15 | 2015-06-11 | Google Inc. | Learning Semantic Image Similarity |
| CN101873605A (en) * | 2010-05-27 | 2010-10-27 | 重庆邮电大学 | Adaptive method for classifying communication environments in network planning |
| CN103279746A (en) * | 2013-05-30 | 2013-09-04 | 苏州大学 | Method and system for identifying faces based on support vector machine |
| CN105975959A (en) * | 2016-06-14 | 2016-09-28 | 广州视源电子科技股份有限公司 | Face feature extraction modeling and face recognition method and device based on neural network |
| CN106202329A (en) * | 2016-07-01 | 2016-12-07 | 北京市商汤科技开发有限公司 | Sample data process, data identification method and device, computer equipment |
| CN107341497A (en) * | 2016-11-11 | 2017-11-10 | 东北大学 | The unbalanced weighting data streams Ensemble classifier Forecasting Methodology of sampling is risen with reference to selectivity |
| WO2018148493A1 (en) * | 2017-02-09 | 2018-08-16 | Painted Dog, Inc. | Methods and apparatus for detecting, filtering, and identifying objects in streaming video |
| CN107577987A (en) * | 2017-08-01 | 2018-01-12 | 广州广电卓识智能科技有限公司 | Identity authentication method, system and device |
| CN107704890A (en) * | 2017-10-27 | 2018-02-16 | 北京旷视科技有限公司 | A kind of generation method and device of four-tuple image |
| CN108108754A (en) * | 2017-12-15 | 2018-06-01 | 北京迈格威科技有限公司 | The training of identification network, again recognition methods, device and system again |
| CN108009528A (en) * | 2017-12-26 | 2018-05-08 | 广州广电运通金融电子股份有限公司 | Face authentication method, device, computer equipment and storage medium based on Triplet Loss |
| CN108304864A (en) * | 2018-01-17 | 2018-07-20 | 清华大学 | Depth fights metric learning method and device |
Non-Patent Citations (5)
| Title |
|---|
| ALEXANDER HERMANS 等: "In Defense of the Triplet Loss for Person Re-Identification", 《ARXIV》 * |
| DE CHENG 等: "Person Re-Identification by Multi-Channel Parts-Based CNN with Improved Triplet Loss Function", 《2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION（CVPR）》 * |
| SWAMI SANKARANARAYANAN 等: "TRIPLET SIMILARITY EMBEDDING FOR FACE VERIFICATION", 《ARXIV》 * |
| WEIHUA CHEN 等: "Beyond triplet loss: a deep quadruplet network for person re-identification", 《PROCEEDINGS OF IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
| YIQIANG CHEN 等: "Triplet CNN and pedestrian attribute recognition for improved person re-identification", 《14TH IEEE INTERNATIONAL CONFERENCE ON ADVANCED VIDEO AND SIGNAL BASED SURVEILLANCE (AVSS 2017)》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114205642A (en) * | 2020-08-31 | 2022-03-18 | 北京金山云网络技术有限公司 | Video image processing method and device |
| CN114205642B (en) * | 2020-08-31 | 2024-04-26 | 北京金山云网络技术有限公司 | Video image processing method and device |
| CN112863132A (en) * | 2021-04-23 | 2021-05-28 | 成都中轨轨道设备有限公司 | Natural disaster early warning system and early warning method |
| CN112863132B (en) * | 2021-04-23 | 2021-07-13 | 成都中轨轨道设备有限公司 | Natural disaster early warning system and early warning method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109145991B (en) | 2020-07-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110147726B (en) | Service quality inspection method and device, storage medium and electronic device | |
| CN108304882B (en) | Image classification method and device, server, user terminal and storage medium | |
| Richard et al. | Temporal action detection using a statistical language model | |
| CN110069709B (en) | Intention recognition method, device, computer readable medium and electronic equipment | |
| WO2019128426A1 (en) | Method for training model and information recommendation system | |
| CN109325148A (en) | The method and apparatus for generating information | |
| JP7172472B2 (en) | RULE GENERATION DEVICE, RULE GENERATION METHOD AND RULE GENERATION PROGRAM | |
| CN106372666B (en) | A kind of target identification method and device | |
| CN111860682B (en) | Sequence recognition method, sequence recognition device, image processing apparatus, and storage medium | |
| CN114064974B (en) | Information processing method, apparatus, electronic device, storage medium, and program product | |
| CN112527113A (en) | Method and apparatus for training gesture recognition and gesture recognition network, medium, and device | |
| CN109657534A (en) | The method, apparatus and electronic equipment analyzed human body in image | |
| CN109063790B (en) | Object recognition model optimization method and device and electronic equipment | |
| CN106156794B (en) | Character recognition method and device based on character style recognition | |
| CN113569968A (en) | Model training method, target detection method, device, equipment and storage medium | |
| CN111783812A (en) | Method and device for identifying forbidden images and computer readable storage medium | |
| CN111753862A (en) | Method and device for training neural network model and image recognition method | |
| CN109145991A (en) | Image group generation method, image group generating means and electronic equipment | |
| CN111222051A (en) | Training method and device of trend prediction model | |
| CN115062186A (en) | Video content retrieval method, device, equipment and storage medium | |
| Nekrasov et al. | Ugains: Uncertainty guided anomaly instance segmentation | |
| CN112416301A (en) | Deep learning model development method and device and computer readable storage medium | |
| CN112569591A (en) | Data processing method, device and equipment and readable storage medium | |
| CN111488400A (en) | Data classification method, device and computer readable storage medium | |
| CN113379895B (en) | Three-dimensional house model generation method and device and computer readable storage medium |
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 |