CN110443241A - Car license recognition model training method, licence plate recognition method and device - Google Patents
Car license recognition model training method, licence plate recognition method and device Download PDFInfo
- Publication number
- CN110443241A CN110443241A CN201910691469.7A CN201910691469A CN110443241A CN 110443241 A CN110443241 A CN 110443241A CN 201910691469 A CN201910691469 A CN 201910691469A CN 110443241 A CN110443241 A CN 110443241A
- Authority
- CN
- China
- Prior art keywords
- character
- difficult
- model
- license plate
- loss value
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Character Discrimination (AREA)
Abstract
Present disclose provides a kind of Car license recognition model training method, licence plate recognition method and devices.Wherein Car license recognition model training method includes: input step, by one or more license plate image input models with multiple characters;Step is exported, by model, output obtains the Classification Loss value and/or recognition result of each character of each license plate image;Difficult character determines step, and Classification Loss value and/or recognition result based on character determine the difficult character in character;Step is returned, the Classification Loss value of difficult character is returned, the Classification Loss value of remaining character is ignored;Model parameter step is updated, according to the Classification Loss value of the difficult character of passback, the model parameter of more new model.By determining that difficult character does targetedly a greater amount of training to difficult sample so that training is more focused on the bigger sample of difficulty, so that it be made to improve the recognition capability to difficult sample, training effect is improved, reduces training cost.
Description
Technical field
Present disclose relates generally to field of image recognition, and in particular to a kind of Car license recognition model training method, license plate are known
Other method and device.
Background technique
With the rapid development of Modern Transportation Technology, Car label recognition technology already becomes a very important research
Project, and be used widely in real life.Such as expressway tol lcollection management, overspeed violation automatic camera, parking lot
The occasions such as management, the management of cell vehicles passing in and out, traffic administration, vehicle location all be unable to do without license plate recognition technology.
In license plate recognition technology, since the digit of license plate is fixed, using model, such as depth convolutional network, to each
It is a kind of common method that position carries out the classification of character respectively.But due to characters on license plate classification have represent each province and city Chinese,
English alphabet and number, the classification difficulty of each character is different in sample.Firstly because in training set each character training
Sample is likely to unbalanced, and secondly their classification difficulty itself is different, for example the character of Chinese will than English and number
Complexity, therefore compare and be difficult to identify.Using conventional method, the recognition accuracy for improving model cannot be trained well, and instruct
Experienced higher cost.
Summary of the invention
In order to solve the above-mentioned problems in the prior art, the first aspect of the disclosure provides a kind of Car license recognition model
Training method, wherein method includes: input step, by one or more license plate image input models with multiple characters;It is defeated
Step out, by model, output obtains the Classification Loss value and/or recognition result of each character of each license plate image;It is difficult
Character determines step, and Classification Loss value and/or recognition result based on character determine the difficult character in character;Step is returned,
The Classification Loss value of difficult character is returned, the Classification Loss value of remaining character is ignored;Model parameter step is updated, according to
The Classification Loss value of the difficult character of passback, the model parameter of more new model.
In one example, difficult character determines that step includes: the Classification Loss value based on character, by the whole in license plate image
The maximum N number of character of the Classification Loss value of character, as difficult character, wherein N is positive integer.
In one example, difficult character determines that step includes: the Classification Loss value based on character, by the whole in license plate image
The Classification Loss value of character is more than the character of loss threshold value, as difficult character.
In one example, difficult character determines that step includes: based on recognition result, using the character of identification mistake as difficult words
Symbol.
In one example, after difficult character determines step, method further include: secondary identification step will include difficult words
The license plate image input model of symbol, by model, output obtains the Classification Loss value of difficult character.
The second aspect of the disclosure provides a kind of licence plate recognition method, wherein method includes: obtaining step, obtains license plate
Image;Identification step identifies license plate image input model to license plate image, obtains license plate recognition result, wherein mould
Type passes through the Car license recognition model training method training of first aspect such as and obtains.
The third aspect of the disclosure provides a kind of Car license recognition model training apparatus, wherein and device includes: input module,
For one or more to be had to the license plate image input models of multiple characters;Output module, for by model, output to be obtained
The Classification Loss value and/or recognition result of each character of each license plate image;Difficult character determining module, for being based on character
Classification Loss value and/or recognition result, determine the difficult character in character;Module is returned, for by the classification of difficult character
Penalty values are returned, and the Classification Loss value of remaining character is ignored;Model parameter module is updated, for the difficult words according to passback
The Classification Loss value of symbol, the model parameter of more new model.
The fourth aspect of the disclosure provides a kind of license plate recognition device, wherein device includes: acquisition module, for obtaining
License plate image;Identification module, for being identified license plate image input model to license plate image, obtaining license plate recognition result,
Wherein, model passes through the Car license recognition model training method training of first aspect such as and obtains.
5th aspect of the disclosure provides a kind of electronic equipment, comprising: memory, for storing instruction;And processor,
The Car license recognition model training method of instruction execution first aspect for calling memory to store or the license plate of second aspect are known
Other method.
6th aspect of the disclosure provides a kind of computer readable storage medium, wherein being stored with instruction, instructs processed
When device executes, the Car license recognition model training method of first aspect or the licence plate recognition method of second aspect are executed.
Car license recognition model training method, licence plate recognition method and the device that the disclosure provides, which pass through, determines difficult character,
So that training is more focused on the bigger sample of difficulty, different training side is taken to difficult sample and remaining easy sample
Method does targetedly a greater amount of training to difficult sample, so that it be made to improve the recognition capability to difficult sample, keeps to appearance
The recognition capability of easy sample reduces training cost to improve training effect.
Detailed description of the invention
The following detailed description is read with reference to the accompanying drawings, above-mentioned and other purposes, the feature of disclosure embodiment
It will become prone to understand with advantage.In the accompanying drawings, several implementations of the disclosure are shown by way of example rather than limitation
Mode, in which:
Fig. 1 shows the flow diagram according to one embodiment Car license recognition model training method of the disclosure;
Fig. 2 shows the flow diagrams according to the disclosure another embodiment Car license recognition model training method;
Fig. 3 shows the flow diagram according to one embodiment licence plate recognition method of the disclosure;
Fig. 4 shows the schematic diagram according to one embodiment Car license recognition model training apparatus of the disclosure;
Fig. 5 shows the schematic diagram according to one embodiment license plate recognition device of the disclosure;
Fig. 6 is a kind of electronic equipment schematic diagram that the embodiment of the present disclosure provides.
In the accompanying drawings, identical or corresponding label indicates identical or corresponding part.
Specific embodiment
The principle and spirit of the disclosure are described below with reference to several illustrative embodiments.It should be appreciated that providing this
A little embodiments are used for the purpose of making those skilled in the art can better understand that realizing the disclosure in turn, and be not with any
Mode limits the scope of the present disclosure.
Although being noted that the statements such as " first " used herein, " second " to describe implementation of the disclosure mode not
Same module, step and data etc., still the statement such as " first ", " second " is merely in different modules, step and data etc.
Between distinguish, and be not offered as specific sequence or significance level.In fact, the statements such as " first ", " second " are complete
It may be used interchangeably.
In order to reduce trained cost, while the training effect of Car license recognition model is improved, the model that training is obtained
Enough more accurately identification characters on license plate, the recognition accuracy of especially highly difficult character, Fig. 1 show the embodiment of the present disclosure and mention
A kind of Car license recognition model training method 100 supplied, comprising: input step 110 exports step 120, and difficult character determines step
130, step 140 is returned, model parameter step 150 is updated.Above-mentioned steps are described in detail below.
Input step 110, by one or more license plate image input models with multiple characters.Wherein, model can be with
It is convolutional neural networks, Classification and Identification is carried out by each character of the sorting algorithm to license plate, each character is as a sample
Unit.
Export step 120, by model, output obtain each character of each license plate image Classification Loss value and/or
Recognition result.Characteristic value, and the identification of available each character are extracted by one or more layers convolution of convolutional neural networks
As a result, can simultaneously obtain Classification Loss value according to recognition result, wherein Classification Loss value can be obtained by Classification Loss function,
Such as cross entropy loss function (Cross Entropy Error Function).
Difficult character determines step 130, Classification Loss value and/or recognition result based on character, determines tired in character
Unfamiliar word symbol.Wherein, difficult character be difficult to identification character, the relatively high character of the error rate identified in other words, need emphasis,
Targetedly model is trained.
In one example, difficult character determines that step 130 includes: the Classification Loss value based on character, will be in license plate image
The maximum N number of character of the Classification Loss value of alphabet, as difficult character, wherein N is positive integer.It can be according to output step
120 obtained Classification Loss values are ranked up by size, and using maximum N number of character as difficult character, Classification Loss value greatly may be used
To reflect the identification degree of difficulty of the character to a certain extent.
In one example, difficult character determines that step 130 includes: the Classification Loss value based on character, will be in license plate image
The Classification Loss value of alphabet is more than the character of loss threshold value, as difficult character.Pass through default loss threshold value, Classification Loss
Value can consider the character recognition relative difficulty more than the character of the loss threshold value, therefore as difficult character.Lose threshold
Value can be set according to the actual situation, and loss threshold value setting is high, and the difficult character of definition is few, and training is at low cost;Lose threshold
Value setting the, the difficult character of definition is more, and training effect is more preferable.
In one example, difficult character determines that step 130 includes: based on recognition result, using the character of identification mistake as tired
Unfamiliar word symbol.According to the recognition result that output step 120 obtains, identify that the character of mistake can consider the character for being difficult to identification,
Targeted training pattern is needed, therefore can be directly using the character of identification mistake as difficult character.
In another example, Fig. 2 shows the embodiment of the present disclosure provide another Car license recognition model training method 100,
As shown in Fig. 2, after difficult character determines step 130, Car license recognition model training method 100 further include: secondary identification step
Rapid 160, by the license plate image input model comprising difficult character, by model, output obtains the Classification Loss value of difficult character.
When exporting step 120 does not have output category penalty values, difficult character has been determined based on recognition result, difficult character will be contained
License plate picture input model obtains the Classification Loss value of difficult character by model.
Step 140 is returned, the Classification Loss value of difficult character is returned, the Classification Loss value of remaining character is ignored.
Here only the Classification Loss value for the difficult character being previously obtained is returned, and the Classification Loss value of remaining character is neglected
Slightly, so that in each license plate, those of hardly possible training character is trained, and model those of can train word very well
Symbol, just without having trained.
Model parameter step 150 is updated, according to the Classification Loss value of the difficult character of passback, the weight of more new model is joined
Number.Model is trained according only to the Classification Loss value of the difficult character of passback, model parameter is updated, to reduce training
Cost improves training effectiveness.Model is done by the Classification Loss value of difficult character sample and is targetedly trained, thus efficiently
The performance of the raising model of rate.
Through the foregoing embodiment, can be in order to reduce the model training cost for Car license recognition, while improving license plate knowledge
The training effect of other model is targetedly trained for indiscernible character, so that accuracy of identification greatly improves.
Below the disclosure by the specific embodiment of online and offline two states Car license recognition model training method 100 into
Row explanation, but just for the sake of making those skilled in the art can better understand that realizing the disclosure in turn, and be not with any
Mode limits the scope of the present disclosure.
In presence, to mode input training sample, i.e. license plate image, epicycle training sample is completed during propagated forward
The difficult negative sample screening of this (as unit of each character of each sample), and carry out part passback.From Project Realization and
Whole characters is sorted for each license plate sample when propagated forward by speech according to penalty values later by propagated forward;Then
The character sample that the maximum top n of penalty values is only chosen in back-propagation process carries out model parameter update, ignores to low damage
The penalty values of mistake value sample.
In off-line state, pre-training first is carried out to identification model using pre-training data.The identification that pre-training is good is enabled again
Model is run in training sample original image respectively, obtains the character largely misidentified as difficult sample.All due to these characters
It, may be comprising having any problem and being easy character sample in a license plate in license plate, therefore what is obtained is that these include difficulty in fact
The license plate figure of character and difficult character marking therein.Then these difficult samples are mixed into part original number by large scale
In, obtaining one more has the new data set of characterization power for difficult case, and finely tunes model again.Wherein comprising difficult character
License plate figure, according to only passback wherein marks the penalty values of difficult character in label training.
The disclosure also provides a kind of licence plate recognition method 200, and Fig. 3 is another Car license recognition that the embodiment of the present disclosure provides
200 flow diagram of method obtains license plate image as shown in figure 3, licence plate recognition method 200 includes: obtaining step 210;Identification
Step 220, by license plate image input model, license plate image is identified, obtains license plate recognition result, wherein model passes through
The training of Car license recognition model training method 100 as described in aforementioned any embodiment obtains.Pass through Car license recognition model training side
The model that the training of method 100 obtains, carries out Car license recognition, can obtain more accurate result.
Fig. 4 shows a kind of Car license recognition model training apparatus 300 of embodiment of the present disclosure offer, as shown in figure 4, license plate
Identification model training device 300 includes: input module 310, for one or more license plate images with multiple characters are defeated
Enter model;Output module 320, for by model, output obtain each character of each license plate image Classification Loss value and/
Or recognition result;Difficult character determining module 330 determines character for Classification Loss value and/or recognition result based on character
In difficult character;Module 340 is returned, for returning the Classification Loss value of difficult character, ignores point of remaining character
Class penalty values;Model parameter module 350 is updated, for the Classification Loss value according to the difficult character of passback, the mould of more new model
Shape parameter.
In one example, difficult character determining module 330 is also used to: the Classification Loss value based on character, will be in license plate image
Alphabet the maximum N number of character of Classification Loss value, as difficult character, wherein N is positive integer.
In one example, difficult character determining module 330 is also used to: the Classification Loss value based on character, will be in license plate image
Alphabet Classification Loss value be more than loss threshold value character, as difficult character.
In one example, difficult character determining module 330 is also used to: be based on recognition result, will identification mistake character as
Difficult character.
In one example, Car license recognition model training apparatus 300 further include: secondary identification module, for that will include difficult words
The license plate image input model of symbol, by model, output obtains the Classification Loss value of difficult character.
In one example, model parameter module 350 is updated, is also used to the Classification Loss value of the difficult character according to passback, more
The model parameter corresponding with difficult character of new model.
A kind of license plate recognition device 400 that the embodiment of the present disclosure also provides, as shown in Figure 5, comprising: obtain module 410, use
In acquisition license plate image;Identification module 420, for being identified license plate image input model to license plate image, obtaining license plate
Recognition result, wherein model is obtained by aforementioned any Car license recognition model training method training.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
As shown in fig. 6, an embodiment of the disclosure provides a kind of electronic equipment 500.Wherein, the electronic equipment
500 include memory 501, processor 502, input/output (Input/Output, I/O) interface 503.Wherein, memory 501,
For storing instruction.Processor 502, the image data of the instruction execution embodiment of the present disclosure for calling memory 501 to store
Cleaning method or image processing method.Wherein, processor 502 is connect with memory 501, I/O interface 503 respectively, such as can
It is attached by bindiny mechanism's (not shown) of bus system and/or other forms.Memory 501 can be used for storing program and
Data, including image data cleaning method involved in the embodiment of the present disclosure or image processing method program, processor 502
By running the program for being stored in memory 501 thereby executing the various function application and data processing of electronic equipment 500.
Processor 502 can use digital signal processor (Digital Signal in the embodiment of the present disclosure
Processing, DSP), field programmable gate array (Field-Programmable Gate Array, FPGA), programmable patrol
At least one of volume array (Programmable Logic Array, PLA) example, in hardware realizes, the processor 502
It can be central processing unit (Central Processing Unit, CPU) or there is data-handling capacity and/or instruction
The combination of one or more of the processing unit of other forms of executive capability.
Memory 501 in the embodiment of the present disclosure may include one or more computer program products, the computer
Program product may include various forms of computer readable storage mediums, such as volatile memory and/or non-volatile deposit
Reservoir.The volatile memory for example may include random access memory (Random Access Memory, RAM) and/
Or cache memory (cache) etc..The nonvolatile memory for example may include read-only memory (Read-Only
Memory, ROM), flash memory (Flash Memory), hard disk (Hard Disk Drive, HDD) or solid state hard disk
(Solid-State Drive, SSD) etc..
In the embodiment of the present disclosure, I/O interface 503 can be used for receiving input instruction (such as number or character information, and
Generate key signals input related with the user setting of electronic equipment 500 and function control etc.), it can also be output to the outside various
Information (for example, image or sound etc.).In the embodiment of the present disclosure I/O interface 503 may include physical keyboard, function button (such as
Volume control button, switch key etc.), mouse, operating stick, trace ball, microphone, one in loudspeaker and touch panel etc.
It is a or multiple.
It is understood that although description operation in a particular order in the accompanying drawings in the embodiment of the present disclosure, is not answered
It is understood as requiring particular order or serial order shown in execute these operations, or requires to execute whole institutes
The operation shown is to obtain desired result.In specific environment, multitask and parallel processing may be advantageous.
The method and apparatus that the embodiment of the present disclosure is related to can be completed using standard programming technology, and utilization is rule-based
Logic or other logics realize various method and steps.It should also be noted that herein and used in claims
Word " device " and " module " are intended to include using the realization of a line or multirow software code and/or hardware realization and/or use
In the equipment for receiving input.
One or more combined individually or with other equipment can be used in any step, operation or program described herein
A hardware or software module are executed or are realized.In one embodiment, software module use includes comprising computer program
The computer program product of the computer-readable medium of code is realized, can be executed by computer processor any for executing
Or whole described step, operation or programs.
For the purpose of example and description, the preceding description of disclosure implementation is had been presented for.Preceding description is not poor
The disclosure is restricted to exact form disclosed by also not the really wanting of act property, according to the above instruction there is likely to be various modifications and
Modification, or various changes and modifications may be obtained from the practice of the disclosure.Select and describe these embodiments and be in order to
Illustrate the principle and its practical application of the disclosure, so that those skilled in the art can be to be suitable for the special-purpose conceived
Come in a variety of embodiments with various modifications and using the disclosure.
Claims (10)
1. a kind of Car license recognition model training method, wherein the described method includes:
Input step, by one or more license plate image input models with multiple characters;
Step is exported, by the model, output obtains the Classification Loss value of each of each license plate image character
And/or recognition result;
Difficult character determines step, and the Classification Loss value and/or the recognition result based on the character determine the word
Difficult character in symbol;
Step is returned, the Classification Loss value of the difficult character is returned, described point for ignoring remaining character
Class penalty values;
Model parameter step is updated, according to the Classification Loss value of the difficult character of passback, updates the mould of the model
Shape parameter.
2. according to the method described in claim 1, wherein, the difficulty character determines that step includes: the institute based on the character
Classification Loss value is stated, by the maximum N number of character of the Classification Loss value of the whole character in the license plate image,
As the difficult character, wherein N is positive integer.
3. method according to claim 1 or 2, wherein the difficulty character determines that step includes: based on the character
The Classification Loss value of the whole character in the license plate image is more than the institute of loss threshold value by the Classification Loss value
Character is stated, as the difficult character.
4. according to the method described in claim 1, wherein, the difficulty character determine step include: based on the recognition result,
Using the character of identification mistake as the difficult character.
5. according to the method described in claim 4, wherein, after the difficult character determines step, the method also includes:
The license plate image comprising the difficult character is inputted the model and is exported by the model by secondary identification step
To the Classification Loss value of the difficult character.
6. a kind of licence plate recognition method, wherein the described method includes:
Obtaining step obtains license plate image;
The license plate image input model identifies the license plate image, obtains license plate recognition result by identification step,
Wherein, the model is obtained by Car license recognition model training method as claimed in claims 1-5 training.
7. a kind of Car license recognition model training apparatus, wherein described device includes:
Input module, for one or more to be had to the license plate image input models of multiple characters;
Output module, for by the model, output to obtain the classification damage of each of each license plate image character
Mistake value and/or recognition result;
Difficult character determining module, for based on the character the Classification Loss value and/or the recognition result, determine institute
State the difficult character in character;
Module is returned, for returning the Classification Loss value of the difficult character, ignores the institute of remaining character
State Classification Loss value;
Model parameter module is updated, for the Classification Loss value according to the difficult character of passback, updates the model
Model parameter.
8. a kind of license plate recognition device, wherein described device includes:
Module is obtained, for obtaining license plate image;
Identification module, for being identified the license plate image input model to the license plate image, obtaining Car license recognition knot
Fruit, wherein the model is obtained by Car license recognition model training method as claimed in claims 1-5 training.
9. a kind of electronic equipment, wherein the electronic equipment includes:
Memory, for storing instruction;And
Processor, for calling the instruction execution license plate according to any one of claims 1 to 5 of the memory storage to know
Other model training method or licence plate recognition method as claimed in claim 6.
10. a kind of computer readable storage medium when described instruction is executed by processor, is executed as weighed wherein being stored with instruction
Benefit requires Car license recognition model training method or licence plate recognition method as claimed in claim 6 described in any one of 1-5.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910691469.7A CN110443241A (en) | 2019-07-29 | 2019-07-29 | Car license recognition model training method, licence plate recognition method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910691469.7A CN110443241A (en) | 2019-07-29 | 2019-07-29 | Car license recognition model training method, licence plate recognition method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110443241A true CN110443241A (en) | 2019-11-12 |
Family
ID=68432070
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910691469.7A Pending CN110443241A (en) | 2019-07-29 | 2019-07-29 | Car license recognition model training method, licence plate recognition method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110443241A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111310850A (en) * | 2020-03-02 | 2020-06-19 | 杭州雄迈集成电路技术股份有限公司 | License plate detection model construction method and system and license plate detection method and system |
CN113496227A (en) * | 2020-04-08 | 2021-10-12 | 顺丰科技有限公司 | Training method and device of character recognition model, server and storage medium |
CN114140802A (en) * | 2022-01-29 | 2022-03-04 | 北京易真学思教育科技有限公司 | Text recognition method and device, electronic equipment and storage medium |
WO2022174805A1 (en) * | 2021-02-22 | 2022-08-25 | 上海商汤智能科技有限公司 | Model training method and apparatus, image processing method and apparatus, electronic device and storage medium |
CN116935373A (en) * | 2023-09-18 | 2023-10-24 | 珠海亿智电子科技有限公司 | Training method, device, equipment and storage medium for multi-type license plate recognition model |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106529532A (en) * | 2016-11-07 | 2017-03-22 | 湖南源信光电科技有限公司 | License plate identification system based on integral feature channels and gray projection |
WO2017074759A1 (en) * | 2015-10-29 | 2017-05-04 | Roger Tracy | System and method for reading a tire code and obtaining tire-related information |
CN106874894A (en) * | 2017-03-28 | 2017-06-20 | 电子科技大学 | A kind of human body target detection method based on the full convolutional neural networks in region |
CN107133616A (en) * | 2017-04-02 | 2017-09-05 | 南京汇川图像视觉技术有限公司 | A kind of non-division character locating and recognition methods based on deep learning |
WO2018184195A1 (en) * | 2017-04-07 | 2018-10-11 | Intel Corporation | Joint training of neural networks using multi-scale hard example mining |
-
2019
- 2019-07-29 CN CN201910691469.7A patent/CN110443241A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2017074759A1 (en) * | 2015-10-29 | 2017-05-04 | Roger Tracy | System and method for reading a tire code and obtaining tire-related information |
CN106529532A (en) * | 2016-11-07 | 2017-03-22 | 湖南源信光电科技有限公司 | License plate identification system based on integral feature channels and gray projection |
CN106874894A (en) * | 2017-03-28 | 2017-06-20 | 电子科技大学 | A kind of human body target detection method based on the full convolutional neural networks in region |
CN107133616A (en) * | 2017-04-02 | 2017-09-05 | 南京汇川图像视觉技术有限公司 | A kind of non-division character locating and recognition methods based on deep learning |
WO2018184195A1 (en) * | 2017-04-07 | 2018-10-11 | Intel Corporation | Joint training of neural networks using multi-scale hard example mining |
Non-Patent Citations (1)
Title |
---|
ABHINAV SHRIVASTAVA 等: "Training Region-based Object Detectors with Online Hard Example Mining", 《CVPR 2016》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111310850A (en) * | 2020-03-02 | 2020-06-19 | 杭州雄迈集成电路技术股份有限公司 | License plate detection model construction method and system and license plate detection method and system |
CN111310850B (en) * | 2020-03-02 | 2023-06-16 | 杭州雄迈集成电路技术股份有限公司 | License plate detection model construction method and system, license plate detection method and system |
CN113496227A (en) * | 2020-04-08 | 2021-10-12 | 顺丰科技有限公司 | Training method and device of character recognition model, server and storage medium |
WO2022174805A1 (en) * | 2021-02-22 | 2022-08-25 | 上海商汤智能科技有限公司 | Model training method and apparatus, image processing method and apparatus, electronic device and storage medium |
CN114140802A (en) * | 2022-01-29 | 2022-03-04 | 北京易真学思教育科技有限公司 | Text recognition method and device, electronic equipment and storage medium |
CN114140802B (en) * | 2022-01-29 | 2022-04-29 | 北京易真学思教育科技有限公司 | Text recognition method and device, electronic equipment and storage medium |
CN116935373A (en) * | 2023-09-18 | 2023-10-24 | 珠海亿智电子科技有限公司 | Training method, device, equipment and storage medium for multi-type license plate recognition model |
CN116935373B (en) * | 2023-09-18 | 2024-03-12 | 珠海亿智电子科技有限公司 | Training method, device, equipment and storage medium for multi-type license plate recognition model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110443241A (en) | Car license recognition model training method, licence plate recognition method and device | |
CN109582793B (en) | Model training method, customer service system, data labeling system and readable storage medium | |
CN109376615B (en) | Method, device and storage medium for improving prediction performance of deep learning network | |
WO2021232229A1 (en) | Virtual scene generation method and apparatus, computer device and storage medium | |
WO2021238262A1 (en) | Vehicle recognition method and apparatus, device, and storage medium | |
CN110929577A (en) | Improved target identification method based on YOLOv3 lightweight framework | |
CN110766038B (en) | Unsupervised landform classification model training and landform image construction method | |
CN108229588B (en) | Machine learning identification method based on deep learning | |
KR20180044295A (en) | How to improve the performance of a trained machine learning model | |
CN110442718A (en) | Sentence processing method, device and server and storage medium | |
KR20170140214A (en) | Filter specificity as training criterion for neural networks | |
EP3295382A1 (en) | Bit width selection for fixed point neural networks | |
CN111044045B (en) | Navigation method and device based on neural network and terminal equipment | |
CN110399487A (en) | A kind of file classification method, device, electronic equipment and storage medium | |
CN109582774A (en) | Natural language classification method, device, equipment and storage medium | |
CN103971136A (en) | Large-scale data-oriented parallel structured support vector machine classification method | |
CN112287656A (en) | Text comparison method, device, equipment and storage medium | |
CN110334716A (en) | Characteristic pattern processing method, image processing method and device | |
US12026592B2 (en) | Machine learning model training method and machine learning model training device | |
CN111949530A (en) | Test result prediction method and device, computer equipment and storage medium | |
CN110334576A (en) | A kind of hand method for tracing and device | |
CN113807407B (en) | Target detection model training method, model performance detection method and device | |
KR102413588B1 (en) | Object recognition model recommendation method, system and computer program according to training data | |
CN110503205A (en) | A kind of multi-modal data processing method and system | |
CN113239034A (en) | Big data resource integration method and system based on artificial intelligence and cloud platform |
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: 20191112 |