CN108647720A - Iterative cycles recognition methods, system, equipment and the storage medium of commodity image - Google Patents
Iterative cycles recognition methods, system, equipment and the storage medium of commodity image Download PDFInfo
- Publication number
- CN108647720A CN108647720A CN201810442387.4A CN201810442387A CN108647720A CN 108647720 A CN108647720 A CN 108647720A CN 201810442387 A CN201810442387 A CN 201810442387A CN 108647720 A CN108647720 A CN 108647720A
- Authority
- CN
- China
- Prior art keywords
- commodity
- commodity image
- confirmed
- image
- matching result
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- 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
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2178—Validation; Performance evaluation; Active pattern learning techniques based on feedback of a supervisor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/751—Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/759—Region-based matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention provides iterative cycles recognition methods, system, equipment and the storage mediums of a kind of commodity image, including:Step S1:Several commodity images are acquired, commodity image is identified by the commodity identification model pre-seted, commodity identification model has a recognition accuracy threshold value;Step S2:The commodity image that will be less than recognition accuracy threshold value matches merchandise classification title of the recognition accuracy more than pre-set threshold and corresponding matching result to be confirmed;Step S3:Matching result to be confirmed is confirmed, when matching result to be confirmed is correct, then executes step S2 again, when matching result to be confirmed is mistake, then triggers step S4;Step S4:Another recognition accuracy is matched more than the merchandise classification title of pre-set threshold and corresponding matching result to be confirmed by merchandise items identification model to the commodity image and executes step S3 again.The present invention can reduce the quantity of artificial mark picture, improve work efficiency.
Description
Technical field
The present invention relates to artificial intelligence, and in particular, to a kind of iterative cycles recognition methods of commodity image, is set system
Standby and storage medium.
Background technology
An important application of the image recognition as artificial intelligence, in recent years an always very active field.
Many researchers have carried out a large amount of research in image recognition, it was also proposed that many algorithms, such as convolutional neural networks, cluster are calculated
Method, support vector machine method and random forests algorithm etc..Most of image recognition algorithm is all built upon the base of statistical model
On plinth, user is needed manually to mark great amount of images sample, then there is the training sample with category label to pass through training
Obtain image recognition model.
In today of big data development, one that unlabelled sample data is very common is marked by marked sample
Kind method.Since handmarking's sample is limited, and it cannot be guaranteed that the data sample of all people's work label is all correct
's.Moreover, in practical problem, unmarked sample size can be considerably beyond the data of marked sample.Therefore, how people is reduced
The quantity of work labeled data, and ensure such Active Learning Method of the accuracy of artificial labeled data as possible, become in recent years
One of the research hotspot come.
Invention content
For the defects in the prior art, the object of the present invention is to provide a kind of iterative cycles identification sides of commodity image
Method, system, equipment and storage medium.
According to the iterative cycles recognition methods of commodity image provided by the invention, include the following steps:
Step S1:Acquire several commodity images, by the commodity identification model that pre-sets to the commodity image into
Row identification, the commodity identification model have a recognition accuracy threshold value;
Step S2:It will be less than a commodity image of the recognition accuracy threshold value or the commodity image area in commodity image
Domain matches a recognition accuracy and is more than the merchandise classification title of pre-set threshold and corresponding matching result to be confirmed;
Step S3:The matching result to be confirmed is confirmed, when the matching result to be confirmed is correct, then again
Secondary execution step S2 then triggers step S4 when the matching result to be confirmed is mistake;
Step S4:Another recognition accuracy is matched by merchandise items identification model to the commodity image and is more than presetting threshold
The merchandise classification title of value and corresponding matching result to be confirmed simultaneously execute step S3 again.
Preferably, further include following steps:
When to the commodity image Region Matching in the commodity image or commodity image less than the recognition accuracy threshold value just
After really, it will be less than the commodity image of the recognition accuracy threshold value or the quotient be added in the commodity image region in commodity image
The training image collection of product identification model, and then the commodity identification model is trained again;
Step S1 to step S4 is repeated, realizes the circuit training of the commodity identification model.
Preferably, training generates the commodity identification model pre-seted as follows:
Step M1:Multigroup commodity image is acquired, every group of commodity image includes the commodity image of multiple same merchandise classifications;
Step M2:Multiple commodity recognition and verification models are generated by the training of multigroup commodity image.
Preferably, the matching result to be confirmed is confirmed, is included the following steps:
Step N1:Other multiple commodity identification models are passed sequentially through to one less than the recognition accuracy threshold value
A commodity image in commodity image or commodity image is identified, until identifying in the commodity image or commodity image
One commodity image;
Step N2:When identifying the commodity image in the commodity image or commodity image, to described to be confirmed
Confirmed with result.
Preferably, training generates the commodity identification model pre-seted as follows:
Step M1:Calculate the distribution of color figure of multiple commodity images in pre-stored data library;
Step M2:According between the distribution of color figure of multiple commodity images merchandise classification title corresponding with commodity image
The commodity identification model pre-seted described in contact foundation.
Preferably, described that matching result to be confirmed is confirmed, specifically, passing through matching to be confirmed described in manual confirmation
As a result whether correct, and the matching result to be confirmed is fed back;
The feedback, specifically, in the matching result to be confirmed correct labeled button or error identification button into
Row triggering input.
Preferably, one identification of commodity image matching that will be less than in the commodity image of the recognition accuracy threshold value is accurate
Degree is realized more than the merchandise classification title of pre-set threshold by following either type:
The commodity image extracted region is gone out, and matches the merchandise classification that a recognition accuracy is more than pre-set threshold
Title;
The commodity image region is identified in the commodity image, and matches a recognition accuracy more than default
Determine the merchandise classification title of threshold value.
The iterative cycles identifying system of commodity image provided by the invention, the iterative cycles for the commodity image are known
Other method, including:
Commodity image identification module, for several collected commodity images, being identified by the commodity pre-seted
The commodity image is identified in model, and the commodity identification model has a recognition accuracy threshold value;
Similitude matching module, in the commodity image or commodity image for will be less than the recognition accuracy threshold value
One commodity image Region Matching, one recognition accuracy is more than the merchandise classification title of pre-set threshold and corresponding to be confirmed
Matching result;
Matching result confirmation module, for confirming to the matching result to be confirmed, when the matching to be confirmed is tied
When fruit is correct, then similitude matching module is triggered again, when the matching result to be confirmed is mistake, then by described similar
Property matching module another recognition accuracy is matched more than pre-set threshold by merchandise items identification model to the commodity image
Merchandise classification title and corresponding matching result to be confirmed simultaneously trigger matching result confirmation module again.
The iterative cycles identification equipment of commodity image provided by the invention, including:
Processor;
Memory, wherein being stored with the executable instruction of the processor;
Wherein, the processor is configured to follow to execute the iteration of the commodity image via the execution executable instruction
The step of ring recognition methods.
Computer readable storage medium provided by the invention, for storing program, described program is performed described in realization
The step of iterative cycles recognition methods of commodity image.
Compared with prior art, the present invention has following advantageous effect:
The present invention is identified the commodity image by the commodity identification model pre-seted, and will be less than institute successively
One recognition accuracy of a commodity image Region Matching stated in a commodity image or commodity image for recognition accuracy threshold value is more than
The merchandise classification title of pre-set threshold and corresponding matching result to be confirmed, then identify mould by other multiple commodity
The commodity image region in the commodity image or commodity image less than the recognition accuracy threshold value is identified in type, i.e.,
By cross-training between model, identification and the mark of commodity image are realized, can not only constantly reinforce commodity identification model energy
Power, and the quantity of artificial mark picture is reduced, it improves work efficiency;
The present invention is identified the commodity image by the commodity identification model pre-seted, and will be less than institute successively
One recognition accuracy of a commodity image Region Matching stated in a commodity image or commodity image for recognition accuracy threshold value is more than
The merchandise classification title of pre-set threshold and corresponding matching result to be confirmed, and by manually being tied to the matching to be confirmed
Fruit is fed back, and determines whether matched merchandise classification title is correct, therefore only need by manually to the matching result of machine
Confirmed, manually every pictures need not be labeled successively, reduce the artificial amount of labour.
Description of the drawings
Upon reading the detailed description of non-limiting embodiments with reference to the following drawings, other feature of the invention,
Objects and advantages will become more apparent upon:
Fig. 1 is the step flow chart of the iterative cycles recognition methods of commodity image in the present invention;
Fig. 2 is the step flow chart of the commodity identification model pre-seted in the present invention;
Fig. 3 is the step flow chart of the commodity identification model pre-seted in variation of the present invention;
Fig. 4 is the step flow chart that matching result to be confirmed is confirmed in the present invention;
Fig. 5 is the module diagram of the iterative cycles identifying system of commodity image in the present invention;
Fig. 6 is the structural schematic diagram of the iterative cycles identification equipment of commodity image in the present invention;And
Fig. 7 is the structural schematic diagram of Computer readable storage medium storing program for executing of the present invention.
Specific implementation mode
With reference to specific embodiment, the present invention is described in detail.Following embodiment will be helpful to the technology of this field
Personnel further understand the present invention, but the invention is not limited in any way.It should be pointed out that the ordinary skill of this field
For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made.These belong to the present invention
Protection domain.
In the present embodiment, Fig. 1 is the step flow chart of the iterative cycles recognition methods of commodity image in the present invention, is such as schemed
Shown in 1, the iterative cycles recognition methods of commodity image provided by the invention includes the following steps:
Step S1:Acquire several commodity images, by the commodity identification model that pre-sets to the commodity image into
Row identification, the commodity identification model have a recognition accuracy threshold value;
Step S2:It will be less than a commodity image of the recognition accuracy threshold value or the commodity figure in commodity image successively
As one recognition accuracy of Region Matching is more than the merchandise classification title of pre-set threshold and corresponding matching result to be confirmed;
Step S3:The matching result to be confirmed is confirmed, when the matching result to be confirmed is correct, then again
Secondary execution step S2 then triggers step S4 when the matching result to be confirmed is mistake;
Step S4:Another recognition accuracy is matched by merchandise items identification model to the commodity image and is more than presetting threshold
The merchandise classification title of value and corresponding matching result to be confirmed simultaneously execute step S3 again.
In the present embodiment, the pre-set threshold is accurate for indicating recognition accuracy, the identification of the pre-set threshold
Exactness is less than the recognition accuracy threshold value.By recognition accuracy the pre-set threshold and the recognition accuracy threshold value it
Between commodity image or commodity image on a commodity image region sorted successively according to recognition accuracy, first will identification
The higher merchandise classification name-matches of accuracy go out.
The adopted name of the entitled commodity of merchandise classification refers to as the generic name of the commodity known to the public, such as:Electricity
Depending on machine, computer, desk, clothes ... etc..The adopted name of commodity is only referred to the title of same class commodity, cannot be used for area
Not same kind of different commodity.
In the present embodiment, the pre-set threshold is configured as needed, and in commodity image or commodity image
Matched recognition accuracy is calculated by the commodity identification model between one commodity image Region Matching and merchandise classification title
It generates.
The iterative cycles recognition methods of commodity image provided by the invention, further includes following steps:
When to the commodity image Region Matching in the commodity image or commodity image less than the recognition accuracy threshold value just
After really, it will be less than the commodity image of the recognition accuracy threshold value or the quotient be added in the commodity image region in commodity image
The training image collection of product identification model, and then the commodity identification model is trained again;
Step S1 to step S4 is repeated, realizes the circuit training of the commodity identification model.
In the present embodiment, the commodity image or the commodity image in commodity image that will be above the recognition accuracy threshold value
The training image collection of the commodity identification model is also added in Region Matching.
Fig. 2 is the step flow chart of commodity identification model pre-seted in the present invention, is pre-seted as shown in Fig. 2, described
Training generates commodity identification model as follows:
Step M1:Multigroup commodity image is acquired, every group of commodity image includes the commodity image of multiple same merchandise classifications;
Step M2:Multiple commodity recognition and verification models are generated by the training of multigroup commodity image.
In the present embodiment, the commodity recognition and verification model uses neural network model or Logic Regression Models.
Fig. 3 is the step flow chart of the commodity identification model pre-seted in variation of the present invention, as shown in figure 3, described pre-
Training generates the commodity identification model of setting as follows:
Step M1:Calculate the distribution of color figure of multiple commodity images in pre-stored data library;
Step M2:According between the distribution of color figure of multiple commodity images merchandise classification title corresponding with commodity image
The commodity identification model pre-seted described in contact foundation.
In the present embodiment, the distribution of color figure is each color accounting of commodity image, due to different classes of commodity
Between color there is certain gap, therefore no distribution of color figure may be used to characterize a kind of classification of commodity,
To which the distribution of color figure of each commodity characterizes the commodity, realizes that commodity identification model is established, when identification, can will also input
Commodity image to be identified extract its distribution of color figure after, confirm the trade name corresponding to it.
Fig. 4 is the step flow chart that matching result to be confirmed is confirmed in the present invention, as shown in figure 4, described to be confirmed
Matching result is confirmed, is included the following steps:
Step N1:Other multiple commodity identification models are passed sequentially through to one less than the recognition accuracy threshold value
A commodity image in commodity image or commodity image is identified, until identifying in the commodity image or commodity image
One commodity image;
Step N2:When identifying the commodity image in the commodity image or commodity image, to described to be confirmed
Confirmed with result.
In the present embodiment, the present invention by other multiple commodity identification models to being less than the recognition accuracy threshold value
A commodity image or commodity image on a commodity image region be identified, that is, pass through cross-training between model, realize
The identification of commodity image and mark can not only constantly reinforce commodity identification model ability, and reduce artificial mark picture
Quantity.By in this present embodiment, multiple commodity identification models include more by different multigroup commodity images, every group of commodity image
The commodity image of Zhang Tongyi merchandise classifications, therefore the ability of each commodity identification model identification and differ, the identification of each commodity
Complementary relationship is formed between model, can be realized and alternately be trained.
It is described that matching result to be confirmed is confirmed, specifically, being by matching result to be confirmed described in manual confirmation
It is no correct, and the matching result to be confirmed is fed back;
The feedback, specifically, in the matching result to be confirmed correct labeled button or error identification button into
Row triggering input.
When realizing the iterative cycles recognition methods of commodity image provided by the invention, commodity can be shown on the screen
A commodity image region on image or commodity image and the trade name matched, while showing out correct button and mistake
Button, by achieving that feedback by mouse click button after artificial judgment.
One recognition accuracy of matching of the commodity image in the commodity image of the recognition accuracy threshold value is will be less than to be more than
The merchandise classification title of pre-set threshold, is realized by following either type:
The commodity image extracted region is gone out, and matches the merchandise classification that a recognition accuracy is more than pre-set threshold
Title;
The commodity image region is identified in the commodity image, and matches a recognition accuracy more than default
Determine the merchandise classification title of threshold value.
When the iterative cycles recognition methods for implementing commodity image provided by the invention, include the following steps
Step S1:Multiple commodity images are acquired, the commodity image is divided into the first commodity image group and the second commodity figure
As group, the commodity image quantity of the first commodity image group is less than the commodity image quantity of the second commodity image group;
Step S2:Generation training image collection is labeled to the commodity image in the first commodity image group, passes through institute
It states training image collection and trains commodity identification model;
Step S3:Commodity image in the second commodity image group is inputted the merchandise items identification model to know
The commodity image that do not classify, and the commodity identification model can not identified forms third commodity image group, the merchandise items
The training image collection is added in the commodity image that identification model identifies;
Step S4:A commodity image in the third commodity image group is matched one by merchandise items identification model
Recognition accuracy is more than the trade name of pre-set threshold and corresponding matching result to be confirmed;
Step S5:The matching result to be confirmed is confirmed, when the matching result to be confirmed is correct, by this
Commodity image is added the training image collection and executes step S4 again and then touched when the matching result to be confirmed is mistake
Send out step S6;
Step S6:The highest commodity of another recognition accuracy are matched by merchandise items identification model to the commodity image
Title and corresponding matching result to be confirmed simultaneously execute step S5 again.
Step S7:Until identify commodity image in all third commodity image groups, and by the third commodity
The commodity image in commodity image and the second commodity image group in image group carries out the merchandise items identification model
Training.
Fig. 5 is the module diagram of the iterative cycles identifying system of commodity image in the present invention, as shown in figure 5, of the invention
The iterative cycles identifying system of the commodity image of offer, for realizing the iterative cycles recognition methods of the commodity image, packet
It includes:
Commodity image identification module, for several collected commodity images, being identified by the commodity pre-seted
The commodity image is identified in model, and the commodity identification model has a recognition accuracy threshold value;
Similitude matching module, in the commodity image or commodity image for will be less than the recognition accuracy threshold value
One commodity image Region Matching, one recognition accuracy is more than the merchandise classification title of pre-set threshold and corresponding to be confirmed
Matching result;
Matching result confirmation module, for confirming to the matching result to be confirmed, when the matching to be confirmed is tied
When fruit is correct, then similitude matching module is triggered again, when the matching result to be confirmed is mistake, then by described similar
Property matching module another recognition accuracy is matched more than pre-set threshold by merchandise items identification model to the commodity image
Merchandise classification title and corresponding matching result to be confirmed simultaneously trigger matching result confirmation module again.
A kind of iterative cycles identification equipment of commodity image, including processor are also provided in the embodiment of the present invention.Memory,
Wherein it is stored with the executable instruction of processor.Wherein, processor is configured to be performed commodity via execution executable instruction
The step of iterative cycles recognition methods of image.
As above, the present invention knows the commodity image by the commodity identification model pre-seted in the embodiment
Not, a commodity image of the recognition accuracy threshold value or the commodity image Region Matching in commodity image and be will be less than successively
One recognition accuracy is more than the merchandise classification title of pre-set threshold and corresponding matching result to be confirmed, then by addition
Multiple commodity identification models to the commodity figure in the commodity image or commodity image less than the recognition accuracy threshold value
As region is identified, i.e., by cross-training between model, realizes identification and the mark of commodity image, can not only constantly add
Strong commodity identification model ability, and the quantity of artificial mark picture is reduced, it improves work efficiency.
Person of ordinary skill in the field it is understood that various aspects of the invention can be implemented as system, method or
Program product.Therefore, various aspects of the invention can be embodied in the following forms, i.e.,:It is complete hardware embodiment, complete
The embodiment combined in terms of full Software Implementation (including firmware, microcode etc.) or hardware and software, can unite here
Referred to as " circuit ", " module " or " platform ".
Fig. 6 is the structural schematic diagram of the iterative cycles identification equipment of the commodity image of the present invention.It is described referring to Fig. 6
The electronic equipment 600 of this embodiment according to the present invention.The electronic equipment 600 that Fig. 6 is shown is only an example, is not answered
Any restrictions are brought to the function and use scope of the embodiment of the present invention.
As shown in fig. 6, electronic equipment 600 is showed in the form of universal computing device.The component of electronic equipment 600 can wrap
It includes but is not limited to:At least one processing unit 610, at least one storage unit 620, (including the storage of connection different platform component
Unit 620 and processing unit 610) bus 630, display unit 640 etc..
Wherein, storage unit has program stored therein code, and program code can be executed by processing unit 610 so that processing is single
Member 610 execute described in this specification above-mentioned electronic prescription circulation processing method part according to the various exemplary implementations of the present invention
The step of mode.For example, processing unit 610 can execute step as shown in fig. 1.
Storage unit 620 may include the readable medium of volatile memory cell form, such as Random Access Storage Unit
(RAM) 6201 and/or cache memory unit 6202, it can further include read-only memory unit (ROM) 6203.
Storage unit 620 can also include program/utility with one group of (at least one) program module 6205
6204, such program module 6205 includes but not limited to:Operating system, one or more application program, other program moulds
Block and program data may include the realization of network environment in each or certain combination in these examples.
Bus 630 can be to indicate one or more in a few class bus structures, including storage unit bus or storage
Cell controller, peripheral bus, graphics acceleration port, processing unit use the arbitrary bus structures in a variety of bus structures
Local bus.
Electronic equipment 600 can also be with one or more external equipments 700 (such as keyboard, sensing equipment, bluetooth equipment
Deng) communication, can also be enabled a user to one or more equipment interact with the electronic equipment 600 communicate, and/or with make
Any equipment that the electronic equipment 600 can be communicated with one or more of the other computing device (such as router, modulation /demodulation
Device etc.) communication.This communication can be carried out by input/output (I/O) interface 650.Also, electronic equipment 600 can be with
By network adapter 660 and one or more network (such as LAN (LAN), wide area network (WAN) and/or public network,
Such as internet) communication.Network adapter 660 can be communicated by bus 630 with other modules of electronic equipment 600.It should
Understand, although being not shown in Fig. 6, other hardware and/or software module can be used in conjunction with electronic equipment 600, including unlimited
In:Microcode, device driver, redundant processing unit, external disk drive array, RAID system, tape drive and number
According to backup storage platform etc..
A kind of computer readable storage medium is also provided in the embodiment of the present invention, for storing program, program is performed
The step of iterative cycles recognition methods of the commodity image of realization.In some possible embodiments, each side of the invention
Face is also implemented as a kind of form of program product comprising program code, when program product is run on the terminal device,
Program code be used to make terminal device execute described in this specification above-mentioned electronic prescription circulation processing method part according to this
The step of inventing various illustrative embodiments.
As it appears from the above, the program of the computer readable storage medium of the embodiment is when being executed, the present invention is by pre-seting
A commodity identification model commodity image is identified, and will be less than a commodity of the recognition accuracy threshold value successively
One recognition accuracy of a commodity image Region Matching on image or commodity image is more than the merchandise classification title of pre-set threshold
And corresponding matching result to be confirmed, then by other multiple commodity identification models to being less than the recognition accuracy threshold
A commodity image region in the commodity image or commodity image of value is identified, i.e., real by cross-training between model
The identification of existing commodity image and mark, can not only constantly reinforce commodity identification model ability, and reduce artificial mark figure
The quantity of piece, improves work efficiency.
Fig. 7 is the structural schematic diagram of the computer readable storage medium of the present invention.Refering to what is shown in Fig. 7, describing according to this
The program product 800 for realizing the above method of the embodiment of invention, may be used the read-only storage of portable compact disc
Device (CD-ROM) and include program code, and can be run on terminal device, such as PC.However, the journey of the present invention
Sequence product is without being limited thereto, in this document, readable storage medium storing program for executing can be any include or storage program tangible medium, the journey
Sequence can be commanded the either device use or in connection of execution system, device.
The arbitrary combination of one or more readable mediums may be used in program product.Readable medium can be that readable signal is situated between
Matter or readable storage medium storing program for executing.Readable storage medium storing program for executing for example can be but be not limited to electricity, magnetic, optical, electromagnetic, infrared ray or partly lead
System, device or the device of body, or the arbitrary above combination.More specific example (the non exhaustive row of readable storage medium storing program for executing
Table) include:Electrical connection, portable disc, hard disk, random access memory (RAM) with one or more conducting wires read-only are deposited
Reservoir (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.
Computer readable storage medium may include in a base band or as the data-signal that a carrier wave part is propagated,
In carry readable program code.The data-signal of this propagation may be used diversified forms, including but not limited to electromagnetic signal,
Optical signal or above-mentioned any appropriate combination.Readable storage medium storing program for executing can also be any readable Jie other than readable storage medium storing program for executing
Matter, which can send, propagate either transmission for used by instruction execution system, device or device or and its
The program of combined use.The program code for including on readable storage medium storing program for executing can transmit with any suitable medium, including but not
It is limited to wireless, wired, optical cable, RF etc. or above-mentioned any appropriate combination.
It can be write with any combination of one or more programming languages for executing the program that operates of the present invention
Code, programming language include object oriented program language-Java, C++ etc., further include conventional process
Formula programming language-such as " C " language or similar programming language.Program code can be calculated fully in user
It executes in equipment, partly execute on a user device, being executed, partly in user calculating equipment as an independent software package
Upper part executes or is executed in remote computing device or server completely on a remote computing.It is being related to remotely counting
In the situation for calculating equipment, remote computing device can pass through the network of any kind, including LAN (LAN) or wide area network
(WAN), it is connected to user calculating equipment, or, it may be connected to external computing device (such as utilize ISP
To be connected by internet).
The present invention is identified the commodity image by the commodity identification model pre-seted, and will be less than institute successively
One recognition accuracy of a commodity image Region Matching stated in a commodity image or commodity image for recognition accuracy threshold value is more than
The merchandise classification title of pre-set threshold and corresponding matching result to be confirmed, then identify mould by other multiple commodity
The commodity image region in the commodity image or commodity image less than the recognition accuracy threshold value is identified in type, i.e.,
By cross-training between model, identification and the mark of commodity image are realized, can not only constantly reinforce commodity identification model energy
Power, and the quantity of artificial mark picture is reduced, it improves work efficiency;
The present invention is identified the commodity image by the commodity identification model pre-seted, and will be less than institute successively
One recognition accuracy of a commodity image Region Matching stated in a commodity image or commodity image for recognition accuracy threshold value is more than
The merchandise classification title of pre-set threshold and corresponding matching result to be confirmed, and by manually being tied to the matching to be confirmed
Fruit is fed back, and determines whether matched merchandise classification title is correct, therefore only need by manually to the matching result of machine
Confirmed, manually every pictures need not be labeled successively, reduce the artificial amount of labour.
Specific embodiments of the present invention are described above.It is to be appreciated that the invention is not limited in above-mentioned
Particular implementation, those skilled in the art can make various deformations or amendments within the scope of the claims, this not shadow
Ring the substantive content of the present invention.
Claims (10)
1. a kind of iterative cycles recognition methods of commodity image, which is characterized in that include the following steps:
Step S1:Several commodity images are acquired, the commodity image is known by the commodity identification model pre-seted
Not, the commodity identification model has a recognition accuracy threshold value;
Step S2:It will be less than a commodity image of the recognition accuracy threshold value or the commodity image region in commodity image
It is more than the merchandise classification title of pre-set threshold and corresponding matching result to be confirmed with a recognition accuracy;
Step S3:The matching result to be confirmed is confirmed, when the matching result to be confirmed is correct, is then held again
Row step S2 then triggers step S4 when the matching result to be confirmed is mistake;
Step S4:Another recognition accuracy is matched more than pre-set threshold by merchandise items identification model to the commodity image
Merchandise classification title and corresponding matching result to be confirmed simultaneously execute step S3 again.
2. the iterative cycles recognition methods of commodity image according to claim 1, which is characterized in that further include walking as follows
Suddenly:
When to the commodity image Region Matching in the commodity image or commodity image less than the recognition accuracy threshold value it is correct it
Afterwards, it will be less than the commodity image of the recognition accuracy threshold value or the commodity image region in commodity image be added the commodity and know
The training image collection of other model, and then the commodity identification model is trained again;
Step S1 to step S4 is repeated, realizes the circuit training of the commodity identification model.
3. the iterative cycles recognition methods of commodity image according to claim 1, which is characterized in that the quotient pre-seted
Training generates product identification model as follows:
Step M1:Multigroup commodity image is acquired, every group of commodity image includes the commodity image of multiple same merchandise classifications;
Step M2:Multiple commodity recognition and verification models are generated by the training of multigroup commodity image.
4. the iterative cycles recognition methods of commodity image according to claim 3, which is characterized in that the matching to be confirmed
As a result confirmed, included the following steps:
Step N1:Other multiple commodity identification models are passed sequentially through to the commodity less than the recognition accuracy threshold value
A commodity image on image or commodity image is identified, until identifying the quotient in the commodity image or commodity image
Product image;
Step N2:When identifying the commodity image in the commodity image or commodity image, the matching to be confirmed is tied
Fruit is confirmed.
5. the iterative cycles recognition methods of commodity image according to claim 1, which is characterized in that
Training generates the commodity identification model pre-seted as follows:
Step M1:Calculate the distribution of color figure of multiple commodity images in pre-stored data library;
Step M2:According to contacting between the distribution of color figure of multiple commodity images merchandise classification title corresponding with commodity image
The commodity identification model pre-seted described in foundation.
6. the iterative cycles recognition methods of commodity image according to claim 1, which is characterized in that
It is described that matching result to be confirmed is confirmed, specifically, whether just by matching result to be confirmed described in manual confirmation
Really, and to the matching result to be confirmed it feeds back;
The feedback, specifically, in the matching result to be confirmed correct labeled button or error identification button touch
Hair input.
7. the iterative cycles recognition methods of commodity image according to claim 1, which is characterized in that
It will be less than the commodity image in the commodity image of the recognition accuracy threshold value and match a recognition accuracy more than default
The merchandise classification title for determining threshold value, is realized by following either type:
The commodity image extracted region is gone out, and matches the merchandise classification title that a recognition accuracy is more than pre-set threshold;
The commodity image region is identified in the commodity image, and matches a recognition accuracy and is more than presetting threshold
The merchandise classification title of value.
8. the iterative cycles identifying system of a kind of commodity image, for realizing the commodity figure described in any one of claim 1 to 7
The iterative cycles recognition methods of picture, which is characterized in that including:
Commodity image identification module, for several collected commodity images, passing through the commodity identification model pre-seted
The commodity image is identified, the commodity identification model has a recognition accuracy threshold value;
Similitude matching module, the quotient in a commodity image or commodity image for will be less than the recognition accuracy threshold value
Product image-region matches a recognition accuracy and is more than the merchandise classification title of pre-set threshold and corresponding matching to be confirmed
As a result;
Matching result confirmation module, for confirming to the matching result to be confirmed, when the matching result to be confirmed is
When correct, then similitude matching module is triggered again, when the matching result to be confirmed is mistake, then by the similitude
The commodity that another recognition accuracy is more than pre-set threshold are matched by merchandise items identification model to the commodity image with module
Item name and corresponding matching result to be confirmed simultaneously trigger matching result confirmation module again.
9. a kind of iterative cycles identification equipment of commodity image, which is characterized in that including:
Processor;
Memory, wherein being stored with the executable instruction of the processor;
Wherein, the processor is configured to come any one of perform claim requirement 1 to 7 institute via the execution executable instruction
The step of stating the iterative cycles recognition methods of commodity image.
10. a kind of computer readable storage medium, for storing program, which is characterized in that described program is performed realization power
Profit requires the step of iterative cycles recognition methods of any one of 1 to 7 commodity image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810442387.4A CN108647720A (en) | 2018-05-10 | 2018-05-10 | Iterative cycles recognition methods, system, equipment and the storage medium of commodity image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810442387.4A CN108647720A (en) | 2018-05-10 | 2018-05-10 | Iterative cycles recognition methods, system, equipment and the storage medium of commodity image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108647720A true CN108647720A (en) | 2018-10-12 |
Family
ID=63754174
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810442387.4A Pending CN108647720A (en) | 2018-05-10 | 2018-05-10 | Iterative cycles recognition methods, system, equipment and the storage medium of commodity image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108647720A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948736A (en) * | 2019-04-04 | 2019-06-28 | 上海扩博智能技术有限公司 | Commodity identification model active training method, system, equipment and storage medium |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103778470A (en) * | 2014-02-13 | 2014-05-07 | 上海交通大学 | Distributed generation island detection method with on-line self-learning ability |
CN104978467A (en) * | 2015-07-23 | 2015-10-14 | 华中科技大学 | Multi-disciplinary design optimization subject decoupling method |
CN105868700A (en) * | 2016-03-25 | 2016-08-17 | 哈尔滨工业大学深圳研究生院 | Vehicle type recognition and tracking method and system based on monitoring video |
CN106548190A (en) * | 2015-09-18 | 2017-03-29 | 三星电子株式会社 | Model training method and equipment and data identification method |
CN106874933A (en) * | 2016-12-30 | 2017-06-20 | 平安科技(深圳)有限公司 | Modeling method and device based on data mining |
CN107292248A (en) * | 2017-06-05 | 2017-10-24 | 广州诚予国际市场信息研究有限公司 | A kind of merchandise control method and system based on image recognition technology |
CN107281755A (en) * | 2017-07-14 | 2017-10-24 | 网易(杭州)网络有限公司 | Construction method, device, storage medium, processor and the terminal of detection model |
-
2018
- 2018-05-10 CN CN201810442387.4A patent/CN108647720A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103778470A (en) * | 2014-02-13 | 2014-05-07 | 上海交通大学 | Distributed generation island detection method with on-line self-learning ability |
CN104978467A (en) * | 2015-07-23 | 2015-10-14 | 华中科技大学 | Multi-disciplinary design optimization subject decoupling method |
CN106548190A (en) * | 2015-09-18 | 2017-03-29 | 三星电子株式会社 | Model training method and equipment and data identification method |
CN105868700A (en) * | 2016-03-25 | 2016-08-17 | 哈尔滨工业大学深圳研究生院 | Vehicle type recognition and tracking method and system based on monitoring video |
CN106874933A (en) * | 2016-12-30 | 2017-06-20 | 平安科技(深圳)有限公司 | Modeling method and device based on data mining |
CN107292248A (en) * | 2017-06-05 | 2017-10-24 | 广州诚予国际市场信息研究有限公司 | A kind of merchandise control method and system based on image recognition technology |
CN107281755A (en) * | 2017-07-14 | 2017-10-24 | 网易(杭州)网络有限公司 | Construction method, device, storage medium, processor and the terminal of detection model |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948736A (en) * | 2019-04-04 | 2019-06-28 | 上海扩博智能技术有限公司 | Commodity identification model active training method, system, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108197532B (en) | The method, apparatus and computer installation of recognition of face | |
CN109934249A (en) | Data processing method, device, medium and calculating equipment | |
EP4167130A1 (en) | Neural network training method and related device | |
WO2021120834A1 (en) | Biometrics-based gesture recognition method and apparatus, computer device, and medium | |
Chen et al. | Deep sketch-shape hashing with segmented 3d stochastic viewing | |
US11734954B2 (en) | Face recognition method, device and electronic equipment, and computer non-volatile readable storage medium | |
WO2018196718A1 (en) | Image disambiguation method and device, storage medium, and electronic device | |
CN110472082A (en) | Data processing method, device, storage medium and electronic equipment | |
CN108062377A (en) | The foundation of label picture collection, definite method, apparatus, equipment and the medium of label | |
CN110222780A (en) | Object detecting method, device, equipment and storage medium | |
CN109871311A (en) | A kind of method and apparatus for recommending test case | |
WO2021139316A1 (en) | Method and apparatus for establishing expression recognition model, and computer device and storage medium | |
Govindarajan et al. | Intelligent collaborative patent mining using excessive topic generation | |
CN109740567A (en) | Key point location model training method, localization method, device and equipment | |
CN110443317A (en) | A kind of method, apparatus and electronic equipment of paper shelves electronic data processing | |
CN110489423A (en) | A kind of method, apparatus of information extraction, storage medium and electronic equipment | |
WO2020013956A1 (en) | Systems, methods, and computer-readable media for improved table identification using a neural network | |
CN103530312A (en) | User identification method and system using multifaceted footprints | |
CN111124863B (en) | Intelligent device performance testing method and device and intelligent device | |
CN115439192A (en) | Medical commodity information pushing method and device, storage medium and computer equipment | |
CN109635706B (en) | Gesture recognition method, device, storage medium and device based on neural network | |
Zheng et al. | Submodular attribute selection for visual recognition | |
KR102649947B1 (en) | System and method of understanding deep context using image and text deep learning | |
CN110222144A (en) | Method for extracting content of text, device, electronic equipment and storage medium | |
CN108647720A (en) | Iterative cycles recognition methods, system, equipment and the storage medium of commodity image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181012 |