CN106503617A - Model training method and device - Google Patents
Model training method and device Download PDFInfo
- Publication number
- CN106503617A CN106503617A CN201610840101.9A CN201610840101A CN106503617A CN 106503617 A CN106503617 A CN 106503617A CN 201610840101 A CN201610840101 A CN 201610840101A CN 106503617 A CN106503617 A CN 106503617A
- Authority
- CN
- China
- Prior art keywords
- sample
- training
- negative
- identification
- test
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The disclosure discloses a kind of model training method and device, belongs to image processing field.The method includes:Concentrate from test sample and choose training sample set, the sample training for including is concentrated to be identified model according to training sample, training sample set is the subset of test sample collection, training sample is concentrated includes positive sample and negative sample, the number of the negative sample that the number of the negative sample that training sample is concentrated is concentrated less than test sample, the sample for including is concentrated to be identified test sample according to identification model, according to the Sample Refreshment identification model of identification mistake;Solve due to limited sample size and the not accurate enough problem of human face recognition model that caused training is obtained;After model being identified by a limited number of sample trainings, again according to the Sample Refreshment identification model of identification mistake, identification model constantly can be optimized, improve accuracy of the identification model when being identified.
Description
Technical field
It relates to image processing field, more particularly to a kind of model training method and device.
Background technology
Face recognition technology is widely used in the different fields such as auto-focusing, gate control system and the identity identification of camera,
Current face recognition technology is typically identified to facial image according to the human face recognition model of training in advance.
When human face recognition model is trained, need training sample set to be extracted from one or more pre-set images, preset figure
Facial image and inhuman face image are generally included as in, the training sample for therefore extracting is concentrated and generally includes several positive samples
Sheet and several negative samples, positive sample is the facial image included in pre-set image, and negative sample is included in pre-set image
Inhuman face image, obtains human face recognition model by concentrating each sample for including to be trained training sample.Due to calculating
Motor speed and the restriction of internal memory, choose training sample out and concentrate and be generally only capable of the sample comprising finite number, therefore, from
When extracting training sample set in pre-set image, typically from pre-set image, the random parts of images that extracts is used as sample, it is impossible to contain
Lid all people's face image and inhuman face image, ultimately resulting in training, to obtain human face recognition model not accurate enough.
Content of the invention
Cause to train the human face recognition model for obtaining not accurate enough to solve the problems, such as due to limited sample size,
The disclosure provides a kind of model training method and device.The technical scheme is as follows:
According to the first aspect of the embodiment of the present disclosure, there is provided a kind of model training method, the method include:
Concentrate from test sample and choose training sample set, concentrate the sample training for including to be identified mould according to training sample
Type, training sample set are the subsets of test sample collection, and training sample is concentrated includes positive sample and negative sample, and training sample is concentrated
The number of the negative sample that the number of negative sample is concentrated less than test sample;
The sample for including is concentrated to be identified test sample according to identification model;
Sample Refreshment identification model according to identification mistake;
Wherein, positive sample is the sample of the image for including predefined type, and negative sample is the image for not including predefined type
Sample.
Optionally, according to the Sample Refreshment identification model of identification mistake, including:
Error rate when concentrating the sample for including to be identified test sample according to identification model is calculated, error rate is to know
The shared ratio in the sample that test sample collection includes of not wrong sample;
When error rate reaches predetermined threshold value, according to the Sample Refreshment identification model of identification mistake.
Optionally, test sample is concentrated includes that positive sample and negative sample, the method also include:
For test sample concentrates each sample for including, determine that the sample attribute of sample, sample attribute are used for indicating sample
Originally it is positive sample or negative sample;
When recognize sample include predefined type image and sample attribute be used for indicate that sample is negative sample when, or
Person, when the image and sample attribute that do not include predefined type in sample is recognized for indicating that sample is positive sample, determines sample
Originally it is the sample of identification mistake.
Optionally, training sample concentrates the whole positive samples and part negative sample for including that test sample is concentrated;According to knowledge
Not wrong Sample Refreshment identification model, including:
Identification model is updated according to the target negative sample that test sample is concentrated, target negative sample is recognized including predetermined
The negative sample of the image of type.
Optionally, the sample for including is concentrated to be identified test sample according to identification model, including:
Other samples in addition to the sample that training sample is concentrated are concentrated to be identified test sample according to identification model.
According to the second aspect of the embodiment of the present disclosure, there is provided a kind of model training apparatus, the device include:
Module is chosen, is configured to concentrate from test sample and is chosen training sample set, included according to training sample concentration
Sample training is identified model, and training sample set is the subset of test sample collection, and training sample is concentrated to be included positive sample and bear
Sample, the number of the negative sample that training sample is concentrated are less than the number of the negative sample that test sample is concentrated;
Identification module, is configured to concentrate the sample for including to be identified test sample according to identification model;
Update module, is configured to the Sample Refreshment identification model according to identification mistake;
Wherein, positive sample is the sample of the image for including predefined type, and negative sample is the image for not including predefined type
Sample.
Optionally, update module includes:
Calculating sub module, is configured to calculate when concentrating the sample that includes to be identified test sample according to identification model
Error rate, error rate is the ratio shared in the sample that test sample collection includes of sample of identification mistake;
Submodule is updated, when being configured as error rate and reaching predetermined threshold value, according to the Sample Refreshment identification of identification mistake
Model.
Optionally, test sample is concentrated includes positive sample and negative sample, and the device also includes:
First determining module, is configured to concentrate each sample for including for test sample, determines the sample category of sample
Property, sample attribute is used for indicating that sample is positive sample or negative sample;
Second determining module, being configured as recognizing sample includes that the image and sample attribute of predefined type are used for referring to
When sample is originally negative sample, or, when recognize do not include predefined type in sample image and sample attribute be used for indicate sample
When being originally positive sample, determine that sample is the sample of identification mistake.
Optionally, training sample concentrates the whole positive samples and part negative sample for including that test sample is concentrated;
Update module, is additionally configured to update identification model according to the target negative sample that test sample is concentrated, and target bears sample
Originally it is the negative sample of the image including predefined type for recognizing.
Optionally, identification module, is additionally configured to concentrate test sample except training sample concentration according to identification model
Other samples outside sample are identified.
According to the third aspect of the embodiment of the present disclosure, there is provided a kind of model training apparatus, the device include:
Processor;
For storing the memory of processor executable;
Wherein, processor is configured to:
Concentrate from test sample and choose training sample set, concentrate the sample training for including to be identified mould according to training sample
Type, training sample set are the subsets of test sample collection, and training sample is concentrated includes positive sample and negative sample, and training sample is concentrated
The number of the negative sample that the number of negative sample is concentrated less than test sample;
The sample for including is concentrated to be identified test sample according to identification model;
Sample Refreshment identification model according to identification mistake;
Wherein, positive sample is the sample of the image for including predefined type, and negative sample is the image for not including predefined type
Sample.
The technical scheme that embodiment of the disclosure is provided can include following beneficial effect:
Training sample set is chosen by concentrating from test sample, concentrates the sample training for including to be known according to training sample
Other model, concentrates the sample for including to be identified test sample according to identification model, and the Sample Refreshment according to identification mistake is known
Other model;Solve due to limited sample size and cause to train the not accurate enough problem of the human face recognition model for obtaining;Passing through
After a limited number of sample trainings are identified model, again according to the Sample Refreshment identification model of identification mistake, can be to knowing
Other model is constantly optimized, and improves accuracy of the identification model when being identified.
It should be appreciated that above general description and detailed description hereinafter are only exemplary, this can not be limited
Open.
Description of the drawings
Accompanying drawing herein is merged in specification and constitutes the part of this specification, shows the enforcement for meeting the disclosure
Example, and the principle for being used for together explaining the disclosure in specification.
Fig. 1 is a kind of flow chart of the model training method according to an exemplary embodiment;
Fig. 2 is a kind of flow chart of the model training method that implements to exemplify according to another exemplary;
Fig. 3 is a kind of flow chart of the model training method that implements to exemplify according to another exemplary;
Fig. 4 is a kind of block diagram of the model training apparatus according to an exemplary embodiment;
Fig. 5 is a kind of block diagram of the model training apparatus that implements to exemplify according to another exemplary;
Fig. 6 is a kind of block diagram of the terminal device according to an exemplary embodiment.
Specific embodiment
Here in detail exemplary embodiment will be illustrated, its example is illustrated in the accompanying drawings.Explained below is related to
During accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represent same or analogous key element.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the disclosure.Conversely, they be only with as appended by
The example of consistent apparatus and method in terms of some that described in detail in claims, the disclosure.
The model training method that the disclosure each embodiment is provided, can be used for such as portable computer, desk-top calculating
In the terminal devices such as machine, smart mobile phone and panel computer, the model training method by terminal device in processor realizing.Can
Choosing, processor is GPU (Graphics Processing Unit, graphic process unit) or CPU in terminal device
(Central Processing Unit, central processing unit).
Fig. 1 is a kind of flow chart of the model training method according to an exemplary embodiment, and the method can be used for
In above-mentioned terminal device, the method includes several steps as follows:
In a step 101, concentrating from test sample and choosing training sample set, the sample instruction for including is concentrated according to training sample
Get identification model.
Wherein, training sample set is the subset of test sample collection, and training sample is concentrated includes positive sample and negative sample, trains
The number of the negative sample that the number of the negative sample in sample set is concentrated less than test sample.Positive sample is the figure for including predefined type
The sample of picture, negative sample are the samples of the image for not including predefined type.
In a step 102, the sample for including is concentrated to be identified test sample according to identification model.
In step 103, according to the Sample Refreshment identification model of identification mistake.
In sum, the model training method that the embodiment of the present disclosure is provided, chooses training sample by concentrating from test sample
This collection, concentrates the sample training for including to be identified model according to training sample, concentrates bag according to identification model to test sample
The sample for including is identified, according to the Sample Refreshment identification model of identification mistake;Solve due to limited sample size and cause
The not accurate enough problem of the human face recognition model that obtains of training;Model is being identified by a limited number of sample trainings
Afterwards, again according to the Sample Refreshment identification model of identification mistake, identification model constantly can be optimized, improves identification mould
Accuracy of the type when being identified.
Fig. 2 is a kind of flow chart of the model training method according to an exemplary embodiment, and the method can be used for
In above-mentioned terminal device, the method comprises the steps:
In step 201, concentrating from test sample and choosing training sample set, the sample instruction for including is concentrated according to training sample
Get identification model.
Test sample is concentrated includes positive sample and negative sample, and positive sample is the sample of the image for including predefined type, bears sample
Originally it is the sample of the image for not including predefined type.The sample that test sample is concentrated is that user pre-enters or terminal device
The picture for getting, and/or, test sample concentrate sample be that user pre-enters or terminal device get one
Subregion in individual picture.Optionally, terminal device is in the picture of receiving user's input or after getting picture, also to figure
Piece carries out the pretreatment such as image gray processing, mean filter and normalization.
Optionally, each positive sample corresponds to the first default mark for being used for indicating that sample is positive sample, each negative sample
Correspond to the second default mark for being used for indicating that sample is negative sample;Or, each positive sample is corresponded to be used for indicating that sample is
The default mark of positive sample, negative sample is not comprising default mark.The present embodiment is corresponded to positive sample to be used for indicating that sample is just
The default mark of sample, negative sample are illustrated as a example by not including default mark.
When actually realizing, when user is by one or more picture input terminal equipment, first is arranged to positive sample and is preset
Mark, arranges the second default mark to negative sample;Or, default mark is set to positive sample, and pre- bidding is not provided with to negative sample
Know.
Optionally, predefined type is any types such as face, numeral, vehicle and building, and predefined type is identification model
The type of the image for recognizing, such as, when identification model is used for recognizing face, predefined type is face.The present embodiment is with pre-
Determining type is illustrated as a example by face, then positive sample is the sample for including face, and negative sample is the sample for not including face.Can
Choosing, negative sample is the sample of the image for including part predefined type, and such as, negative sample is the sample for including half face.
Training sample set is the subset of test sample collection, and training sample is concentrated includes positive sample and negative sample, terminal device
Concentrate from test sample and randomly select the positive sample that several positive samples are concentrated as training sample, concentrate from test sample random
The negative sample that several negative samples are concentrated is chosen as training sample.The number of the negative sample that training sample is concentrated is less than test specimens
The number of the negative sample of this concentration, the number of the positive sample that training sample is concentrated is less than or equal to the positive sample that test sample is concentrated
This number.Under normal circumstances, training sample concentrates the number of the negative sample for including more than the number of positive sample.
Optionally, terminal device extracts sample characteristics to positive sample and negative sample, and special to sample according to pre-defined algorithm
Levy training and be identified model.Optionally, pre-defined algorithm includes artificial neural network, Adaboost, SVM (Support
Vector Machine, SVMs), at least one in genetic algorithm and naive Bayesian.Optionally, ANN
Network can be that (Region-based Convolutional Neural Networks, the depth based on region candidate are rolled up R-CNN
Product network), Fast R-CNN algorithms or Faster R-CNN algorithms.
In step 202., other in addition to the sample that training sample is concentrated are concentrated according to identification model to test sample
Sample is identified.
For test sample concentrates each sample in addition to the sample that training sample is concentrated, terminal device to extract sample
Sample characteristics, by these sample characteristics input identification models, when the image that sample includes predefined type is recognized, represent
According to identification model by the specimen discerning be positive sample;When the image for not including predefined type in sample is recognized, root is represented
According to identification model by the specimen discerning be negative sample.
In step 203, for test sample concentrates each sample for including, determine that the sample attribute of sample, sample belong to
Property for indicating that sample is positive sample or negative sample.
Whether terminal device determines the sample attribute of sample by detecting sample comprising default mark, when detecting sample bag
During containing default mark, determine that sample attribute is used for indicating that sample is positive sample;When sample is detected not comprising default mark, really
Random sample sheet is used for indicating that sample is negative sample.Or, when sample being detected comprising the first default mark, determine that sample attribute is used
In indicate sample be positive sample;When sample being detected comprising the second default mark, determine that sample is negative sample for indicating sample
This.
In step 204, when recognize sample include predefined type image and sample attribute be used for indicate that sample is
During negative sample, or, when recognize do not include predefined type in sample image and sample attribute be used for indicate that sample is positive sample
This when, determine that sample is the sample of identification mistake.
When training sample concentrates the number of the positive sample for including to concentrate the number of the positive sample for including again smaller than test sample
When, can there is the situation of the unidentified image to predefined type in positive sample in terminal device, can also exist and know in negative sample
The situation of the image of predefined type is clipped to, in these cases, terminal device all thinks that these samples are all the samples of identification mistake
This.
In step 205, mistake when concentrating the sample for including to be identified test sample according to identification model is calculated
Rate.
Wherein, error rate is ratio of the sample of identification mistake shared by the sample that test sample collection includes.
Optionally, the positive sample that the number of the positive sample for including when training sample concentration includes less than test sample concentration
Number, and training sample is when concentrating the number that the number of negative sample for including concentrates the negative sample for including less than test sample, wrong
Rate includes the first error rate and the second error rate by mistake.The step includes:The image for not including predefined type that calculating is recognized
Positive sample concentrates shared ratio in all positive samples as the first error rate in test sample, and not including that calculating is recognized is pre-
The positive sample for determining the image of type concentrates shared ratio in all positive samples as the second error rate in test sample.Such as,
When the image of predefined type is facial image, test sample is concentrated includes 2000 positive samples, and 4000 negative samples are unidentified
Positive sample to face includes 50, and the negative sample for recognizing face includes 400, then the first error rate is 2.5%, and second is wrong
Rate is 10% by mistake.
In step 206, when error rate reaches predetermined threshold value, according to the Sample Refreshment identification model of identification mistake.
Optionally, when the first error rate reaches predetermined threshold value, training sample is added to concentrate the positive sample of identification mistake,
Positive sample in supplementary training sample set;When the second error rate reaches predetermined threshold value, the negative sample of identification mistake is added instruction
Practice sample set, the negative sample in supplementary training sample set.Using identification mistake sample again identification model is iterated with
Update identification model.Wherein, predetermined threshold value is systemic presupposition value or User Defined value, value of the present embodiment to predetermined threshold value
It is not construed as limiting.
In step 207, when error rate is less than predetermined threshold value, terminate flow process.
In sum, the model training method for providing in the embodiment of the present disclosure, chooses training by concentrating from test sample
Sample set, concentrates the sample training for including to be identified model according to training sample, test sample is concentrated according to identification model
Including sample be identified, according to identification mistake Sample Refreshment identification model;Solve due to limited sample size and lead
Cause to train the not accurate enough problem of the human face recognition model for obtaining;After model is identified by a limited number of sample trainings,
Again according to the Sample Refreshment identification model of identification mistake, identification model constantly can be optimized, improve identification model
Accuracy when being identified.
Optionally, in other alternative embodiments based on above-described embodiment, as test sample concentrates the positive sample for including
This number is generally less, and in order to improve recall rate, therefore, all positive samples that can generally choose test sample concentration are made
For the sample that training sample is concentrated, then training sample is concentrated includes the negative sample of whole positive samples of test sample concentration and part
This, then above-mentioned steps 204- step 206 can be exemplified as step by replacement, as shown in Figure 3:
In step 301, when recognize sample include predefined type image and sample attribute be used for indicate that sample is
During negative sample, determine that sample is the sample of identification mistake.
When training sample concentrates the whole positive samples for including test sample concentration, terminal device is according to identification model pair
When the sample that test sample is concentrated is identified, the sample of all images comprising predefined type can all be identified, but
Still can there is the situation of the image that predefined type is recognized in negative sample, therefore, recognize that the sample of mistake only includes recognizing
Negative sample to the image of predefined type.
In step 302, the negative sample for calculating the image including predefined type for recognizing is all in test sample concentration
In negative sample, shared ratio is error rate.
In step 303, identification model is updated according to the target negative sample that test sample is concentrated, target negative sample is identification
The negative sample of the image including predefined type for arriving.
Following for disclosure device embodiment, can be used for executing method of disclosure embodiment.For disclosure device reality
The details not disclosed in example is applied, method of disclosure embodiment is refer to.
Fig. 4 is a kind of block diagram of the model training apparatus according to an exemplary embodiment, as shown in figure 4, the device
Can by hardware, software or both be implemented in combination with become above-mentioned terminal device, the device is included but is not limited to:
Module 410 is chosen, is configured to concentrate from test sample and is chosen training sample set, being concentrated according to training sample includes
Sample training be identified model, training sample set is the subset of test sample collection, training sample concentrate include positive sample with
Negative sample, the number of the negative sample that training sample is concentrated are less than the number of the negative sample that test sample is concentrated.Wherein, positive sample is
Including the sample of the image of predefined type, negative sample is the sample of the image for not including predefined type.
Identification module 420, is configured to concentrate the sample for including to be identified test sample according to identification model.
Update module 430, is configured to the Sample Refreshment identification model according to identification mistake.
In sum, the model training apparatus that the embodiment of the present disclosure is provided, choose training sample by concentrating from test sample
This collection, concentrates the sample training for including to be identified model according to training sample, concentrates bag according to identification model to test sample
The sample for including is identified, according to the Sample Refreshment identification model of identification mistake;Solve due to limited sample size and cause
Train the not accurate enough problem of the human face recognition model for obtaining;After model is identified by a limited number of sample trainings, then
The secondary Sample Refreshment identification model according to identification mistake, can constantly be optimized to identification model, improve identification model and exist
Accuracy when being identified.
Fig. 5 is a kind of block diagram of the model training apparatus according to an exemplary embodiment, as shown in figure 5, the device
Can by hardware, software or both be implemented in combination with become above-mentioned terminal device, the device is included but is not limited to:
Module 510 is chosen, is configured to concentrate from test sample and is chosen training sample set, being concentrated according to training sample includes
Sample training be identified model, training sample set is the subset of test sample collection, training sample concentrate include positive sample with
Negative sample, the number of the negative sample that training sample is concentrated are less than the number of the negative sample that test sample is concentrated.
Identification module 520, is configured to concentrate the sample for including to be identified test sample according to identification model.
Identification module 520, is additionally configured to concentrate the sample that concentrates except training sample according to identification model to test sample
Outside other samples be identified.
First determining module 530, is configured to concentrate each sample for including for test sample, determines the sample of sample
Attribute, sample attribute are used for indicating that sample is positive sample or negative sample.
Second determining module 540, being configured as recognizing sample includes that the image and sample attribute of predefined type are used
In indicate sample be negative sample when, or, when recognize do not include predefined type in sample image and sample attribute be used for refer to
When sample is originally positive sample, determine that sample is the sample of identification mistake.
Update module 550, is configured to the Sample Refreshment identification model according to identification mistake.
Optionally, update module 550 includes:
Calculating sub module 551, is configured to calculate and concentrates the sample for including to know test sample according to identification model
Error rate when other, error rate are ratio of the sample of identification mistake shared by the sample that test sample collection includes.
Submodule 552 is updated, when being configured as error rate and reaching predetermined threshold value, the Sample Refreshment according to identification mistake is known
Other model.
Optionally, training sample concentrates the whole positive samples and part negative sample for including that test sample is concentrated;Update mould
Block 550, is additionally configured to update identification model according to the target negative sample that test sample is concentrated, and target negative sample is recognized
Negative sample including the image of predefined type.
In sum, the model training apparatus that the embodiment of the present disclosure is provided, choose training sample by concentrating from test sample
This collection, concentrates the sample training for including to be identified model according to training sample, concentrates bag according to identification model to test sample
The sample for including is identified, according to the Sample Refreshment identification model of identification mistake;Solve due to limited sample size and cause
Train the not accurate enough problem of the human face recognition model for obtaining;After model is identified by a limited number of sample trainings, then
The secondary Sample Refreshment identification model according to identification mistake, can constantly be optimized to identification model, improve identification model and exist
Accuracy when being identified.
Device in regard to above-described embodiment, wherein modules execute the concrete mode of operation in relevant the method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
One exemplary embodiment of the disclosure provides a kind of model training apparatus, can realize the model instruction that the disclosure is provided
Practice method, the device includes:Processor, for storing the memory of processor executable;
Wherein, processor is configured to:
Concentrate from test sample and choose training sample set, concentrate the sample training for including to be identified mould according to training sample
Type, training sample set are the subsets of test sample collection, and training sample is concentrated includes positive sample and negative sample, and training sample is concentrated
The number of the negative sample that the number of negative sample is concentrated less than test sample;
The sample for including is concentrated to be identified test sample according to identification model;
Sample Refreshment identification model according to identification mistake;
Wherein, positive sample is the sample of the image for including predefined type, and negative sample is the image for not including predefined type
Sample.
Fig. 6 is a kind of block diagram of the model training apparatus according to an exemplary embodiment.For example, device 600 can be with
It is mobile phone, computer, digital broadcast terminal, messaging devices, game console, tablet device, personal digital assistant
Deng.
With reference to Fig. 6, device 600 can include following one or more assemblies:Process assembly 602, memory 604, power supply
Component 606, multimedia groupware 608, audio-frequency assembly 610, input/output (I/O) interface 612, sensor cluster 614, Yi Jitong
Letter component 616.
The integrated operation of 602 usual control device 600 of process assembly, such as with display, call, data communication, phase
The associated operation of machine operation and record operation.Process assembly 602 can refer to execute including one or more processors 618
Order, to complete all or part of step of above-mentioned method.Additionally, process assembly 602 can include one or more modules, just
Interaction between process assembly 602 and other assemblies.For example, process assembly 602 can include multi-media module, many to facilitate
Interaction between media component 608 and process assembly 602.
Memory 604 is configured to store various types of data to support the operation in device 600.These data are shown
Example includes the instruction of any application program or method for operating on device 600, and contact data, telephone book data disappear
Breath, picture, video etc..Memory 604 can be by any kind of volatibility or non-volatile memory device or their group
Close and realize, such as static RAM (SRAM), Electrically Erasable Read Only Memory (EEPROM) erasable are compiled
Journey read-only storage (EPROM), programmable read only memory (PROM), read-only storage (ROM), magnetic memory, flash
Device, disk or CD.
Power supply module 606 provides electric power for the various assemblies of device 600.Power supply module 606 can include power management system
System, one or more power supplys, and other generate, manage and distribute the component that electric power is associated with for device 600.
Multimedia groupware 608 is included in the screen of one output interface of offer between device 600 and user.In some realities
Apply in example, screen can include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel, screen can
To be implemented as touch-screen, to receive the input signal from user.Touch panel include one or more touch sensors with
Gesture on sensing touch, slip and touch panel.Touch sensor can not only sensing touch or sliding action border, and
And also detect the duration related to touch or slide and pressure.In certain embodiments, multimedia groupware 608 includes
One front-facing camera and/or post-positioned pick-up head.When device 600 is in operator scheme, such as screening-mode or during video mode is front
Put the multi-medium data that camera and/or post-positioned pick-up head can receive outside.Each front-facing camera and post-positioned pick-up head can
To be a fixed optical lens system or there is focusing and optical zoom capabilities.
Audio-frequency assembly 610 is configured to output and/or input audio signal.For example, audio-frequency assembly 610 includes a Mike
Wind (MIC), when device 600 is in operator scheme, such as call model, logging mode and speech recognition mode, microphone is matched somebody with somebody
It is set to reception external audio signal.The audio signal for being received can be further stored in memory 604 or via communication set
Part 616 sends.In certain embodiments, audio-frequency assembly 610 also includes a loudspeaker, for exports audio signal.
I/O interfaces 612 are to provide interface between process assembly 602 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include but be not limited to:Home button, volume button, start button and lock
Determine button.
Sensor cluster 614 includes one or more sensors, comments for providing the state of various aspects for device 600
Estimate.For example, sensor cluster 614 can detect the opening/closed mode of device 600, the such as relative positioning of component, component
For the display and keypad of device 600, sensor cluster 614 can be with the position of 600 1 components of detection means 600 or device
Change is put, user is presence or absence of with what device 600 was contacted, the temperature of 600 orientation of device or acceleration/deceleration and device 600
Change.Sensor cluster 614 can include proximity transducer, be configured to when without any physical contact near detection
The presence of object.Sensor cluster 614 can also include optical sensor, such as CMOS or ccd image sensor, for answering in imaging
With used in.In certain embodiments, the sensor cluster 614 can also include acceleration transducer, gyro sensor, magnetic
Sensor, pressure sensor or temperature sensor.
Communication component 616 is configured to facilitate the communication of wired or wireless way between device 600 and other equipment.Device
600 can access the wireless network based on communication standard, such as Wi-Fi, 2G or 3G, or combinations thereof.In an exemplary reality
Apply in example, communication component 616 receives the broadcast singal or the related letter of broadcast from external broadcasting management system via broadcast channel
Breath.In one exemplary embodiment, communication component 616 also includes near-field communication (NFC) module, to promote junction service.Example
Such as, NFC module can be based on RF identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra broadband (UWB) technology,
Bluetooth (BT) technology and other technologies are realizing.
In the exemplary embodiment, device 600 can be by one or more application specific integrated circuits (ASIC), numeral letter
Number processor (DSP), digital signal processing appts (DSPD), PLD (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, for executing above-mentioned model training method.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instructing, example are additionally provided
Such as include that the memory 604 for instructing, above-mentioned instruction can be executed to complete above-mentioned model training side by the processor 618 of device 600
Method.For example, non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape,
Floppy disk and optical data storage devices etc..
Those skilled in the art will readily occur to its of the disclosure after considering specification and putting into practice invention disclosed herein
Its embodiment.The application is intended to any modification, purposes or the adaptations of the disclosure, these modifications, purposes or
Person's adaptations follow the general principle of the disclosure and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.Description and embodiments be considered only as exemplary, the true scope of the disclosure and spirit by following
Claim is pointed out.
It should be appreciated that the disclosure is not limited to the precision architecture for being described above and being shown in the drawings, and
And various modifications and changes can carried out without departing from the scope.The scope of the present disclosure is only limited by appended claim.
Claims (11)
1. a kind of model training method, it is characterised in that methods described includes:
Concentrate from test sample and choose training sample set, concentrate the sample training for including to be identified mould according to the training sample
Type, the training sample set are the subsets of the test sample collection, and the training sample is concentrated includes positive sample and negative sample, institute
The number for stating the negative sample of training sample concentration is less than the number of the negative sample that the test sample is concentrated;
The sample for including is concentrated to be identified the test sample according to the identification model;
Identification model according to the Sample Refreshment of identification mistake;
Wherein, the positive sample is the sample of the image for including predefined type, and the negative sample is not include the predefined type
Image sample.
2. method according to claim 1, it is characterised in that described recognize mould according to the Sample Refreshment of identification mistake
Type, including:
Calculate error rate when concentrating the sample for including to be identified the test sample according to the identification model, the mistake
Rate is ratio of the sample of the identification mistake shared by the sample that the test sample collection includes by mistake;
When the error rate reaches predetermined threshold value, identification model according to the Sample Refreshment of the identification mistake.
3. method according to claim 1 and 2, it is characterised in that the test sample is concentrated includes positive sample and negative sample
This, methods described also includes:
For the test sample concentrates each sample for including, determine that the sample attribute of the sample, the sample attribute are used
In indicating that the sample is positive sample or negative sample;
Include that the image and the sample attribute of the predefined type are used for indicating that the sample is when the sample is recognized
During negative sample, or, when recognize do not include the predefined type in the sample image and the sample attribute be used for refer to
When to show the sample be positive sample, determine that the sample is the sample of the identification mistake.
4. method according to claim 1, it is characterised in that the training sample is concentrated includes the test sample concentration
Whole positive samples and part negative sample;
The identification model according to the Sample Refreshment of identification mistake, including:
The identification model is updated according to the target negative sample that the test sample is concentrated, the target negative sample is recognized
Negative sample including the image of the predefined type.
5. method according to claim 1 and 2, it is characterised in that described according to the identification model to the test specimens
The sample that this concentration includes is identified, including:
Other samples in addition to the sample that the training sample is concentrated are concentrated according to the identification model to the test sample
It is identified.
6. a kind of model training apparatus, it is characterised in that described device includes:
Module is chosen, is configured to concentrate from test sample and is chosen training sample set, included according to training sample concentration
Sample training is identified model, and the training sample set is the subset of the test sample collection, and the training sample concentrates bag
Positive sample and negative sample is included, the number of the negative sample that the training sample is concentrated is less than the negative sample of test sample concentration
Number;
Identification module, is configured to concentrate the sample for including to be identified the test sample according to the identification model;
Update module, is configured to identification model according to the Sample Refreshment of identification mistake;
Wherein, the positive sample is the sample of the image for including predefined type, and the negative sample is not include the predefined type
Image sample.
7. device according to claim 6, it is characterised in that the update module includes:
Calculating sub module, is configured to calculate and concentrates the sample for including to know the test sample according to the identification model
Error rate when other, the error rate are that the sample of the identification mistake is shared in the sample that the test sample collection includes
Ratio;
Submodule is updated, when being configured as the error rate and reaching predetermined threshold value, according to the Sample Refreshment of the identification mistake
The identification model.
8. the device according to claim 6 or 7, it is characterised in that the test sample is concentrated includes positive sample and negative sample
This, described device also includes:
First determining module, is configured to concentrate each sample for including for the test sample, determines the sample of the sample
This attribute, the sample attribute are used for indicating that the sample is positive sample or negative sample;
Second determining module, being configured as recognizing the sample includes image and the sample category of the predefined type
Property for when to indicate the sample be negative sample, or, when recognizing the image that do not include the predefined type in the sample
And the sample attribute is when being used for that to indicate the sample to be positive sample, determine that the sample is the sample of the identification mistake.
9. device according to claim 6, it is characterised in that the training sample is concentrated includes the test sample concentration
Whole positive samples and part negative sample;
The update module, is additionally configured to update the identification model according to the target negative sample that the test sample is concentrated,
The target negative sample is the negative sample of the image including the predefined type for recognizing.
10. the device according to claim 6 or 7, it is characterised in that
The identification module, is additionally configured to concentrate except the training sample set test sample according to the identification model
In sample outside other samples be identified.
11. a kind of model training apparatus, it is characterised in that described device includes:
Processor;
For storing the memory of the processor executable;
Wherein, the processor is configured to:
Concentrate from test sample and choose training sample set, concentrate the sample training for including to be identified mould according to the training sample
Type, the training sample set are the subsets of the test sample collection, and the training sample is concentrated includes positive sample and negative sample, institute
The number for stating the negative sample of training sample concentration is less than the number of the negative sample that the test sample is concentrated;
The sample for including is concentrated to be identified the test sample according to the identification model;
Identification model according to the Sample Refreshment of identification mistake;
Wherein, the positive sample is the sample of the image for including predefined type, and the negative sample is not include the predefined type
Image sample.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610840101.9A CN106503617A (en) | 2016-09-21 | 2016-09-21 | Model training method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610840101.9A CN106503617A (en) | 2016-09-21 | 2016-09-21 | Model training method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106503617A true CN106503617A (en) | 2017-03-15 |
Family
ID=58290737
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610840101.9A Pending CN106503617A (en) | 2016-09-21 | 2016-09-21 | Model training method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106503617A (en) |
Cited By (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107194464A (en) * | 2017-04-25 | 2017-09-22 | 北京小米移动软件有限公司 | The training method and device of convolutional neural networks model |
CN107292154A (en) * | 2017-06-09 | 2017-10-24 | 北京奇安信科技有限公司 | A kind of terminal feature recognition methods and system |
CN107341457A (en) * | 2017-06-21 | 2017-11-10 | 北京小米移动软件有限公司 | Method for detecting human face and device |
CN107426147A (en) * | 2016-03-28 | 2017-12-01 | 阿里巴巴集团控股有限公司 | For the method and apparatus for the anti-spam performance for determining application |
CN107463906A (en) * | 2017-08-08 | 2017-12-12 | 深图(厦门)科技有限公司 | The method and device of Face datection |
CN107491790A (en) * | 2017-08-25 | 2017-12-19 | 北京图森未来科技有限公司 | A kind of neural network training method and device |
CN108108821A (en) * | 2017-12-29 | 2018-06-01 | 广东欧珀移动通信有限公司 | Model training method and device |
CN108197777A (en) * | 2017-12-14 | 2018-06-22 | 阿里巴巴集团控股有限公司 | A kind of method, apparatus and equipment for adjusting air control rule |
CN108304868A (en) * | 2018-01-25 | 2018-07-20 | 阿里巴巴集团控股有限公司 | Model training method, data type recognition methods and computer equipment |
CN109214175A (en) * | 2018-07-23 | 2019-01-15 | 中国科学院计算机网络信息中心 | Method, apparatus and storage medium based on sample characteristics training classifier |
CN109326286A (en) * | 2018-10-23 | 2019-02-12 | 出门问问信息科技有限公司 | Voice information processing method, device and electronic equipment |
CN109344862A (en) * | 2018-08-21 | 2019-02-15 | 中国平安人寿保险股份有限公司 | Acquisition methods, device, computer equipment and the storage medium of positive sample |
CN109359056A (en) * | 2018-12-21 | 2019-02-19 | 北京搜狗科技发展有限公司 | A kind of applied program testing method and device |
WO2019179189A1 (en) * | 2018-03-23 | 2019-09-26 | 北京达佳互联信息技术有限公司 | Image classification model optimization method and device and terminal |
CN110460488A (en) * | 2019-07-01 | 2019-11-15 | 华为技术有限公司 | Business stream recognition method and device, model generating method and device |
CN110533057A (en) * | 2019-04-29 | 2019-12-03 | 浙江科技学院 | A kind of Chinese character method for recognizing verification code under list sample and few sample scene |
WO2020024737A1 (en) * | 2018-08-02 | 2020-02-06 | 腾讯科技(深圳)有限公司 | Method and apparatus for generating negative sample of face recognition, and computer device |
CN110969210A (en) * | 2019-12-02 | 2020-04-07 | 中电科特种飞机系统工程有限公司 | Small and slow target identification and classification method, device, equipment and storage medium |
CN110998648A (en) * | 2018-08-09 | 2020-04-10 | 北京嘀嘀无限科技发展有限公司 | System and method for distributing orders |
CN111062342A (en) * | 2019-12-20 | 2020-04-24 | 中国银行股份有限公司 | Debugging method and device of face recognition system |
CN111199175A (en) * | 2018-11-20 | 2020-05-26 | 株式会社日立制作所 | Training method and device for target detection network model |
CN112259085A (en) * | 2020-09-28 | 2021-01-22 | 上海声瀚信息科技有限公司 | Two-stage voice awakening algorithm based on model fusion framework |
CN112257670A (en) * | 2020-11-16 | 2021-01-22 | 北京爱笔科技有限公司 | Image processing model and machine learning model training method and device |
CN112800403A (en) * | 2021-01-05 | 2021-05-14 | 北京小米松果电子有限公司 | Method, apparatus and medium for generating prediction model and predicting fingerprint recognition abnormality |
CN113343051A (en) * | 2021-06-04 | 2021-09-03 | 全球能源互联网研究院有限公司 | Abnormal SQL detection model construction method and detection method |
CN113591782A (en) * | 2021-08-12 | 2021-11-02 | 北京惠朗时代科技有限公司 | Training-based face recognition intelligent safety box application method and system |
CN113780485A (en) * | 2021-11-12 | 2021-12-10 | 浙江大华技术股份有限公司 | Image acquisition, target recognition and model training method and equipment |
CN113887283A (en) * | 2021-08-30 | 2022-01-04 | 兴圣创(苏州)科技有限公司 | Strange face recognition score threshold determination method, face recognition method and system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
US20120321128A1 (en) * | 2008-04-01 | 2012-12-20 | University Of Southern California | Video feed target tracking |
CN103166830A (en) * | 2011-12-14 | 2013-06-19 | 中国电信股份有限公司 | Spam email filtering system and method capable of intelligently selecting training samples |
CN103680495A (en) * | 2012-09-26 | 2014-03-26 | 中国移动通信集团公司 | Speech recognition model training method, speech recognition model training device and terminal |
CN105320957A (en) * | 2014-07-10 | 2016-02-10 | 腾讯科技(深圳)有限公司 | Classifier training method and device |
CN105426857A (en) * | 2015-11-25 | 2016-03-23 | 小米科技有限责任公司 | Training method and device of face recognition model |
-
2016
- 2016-09-21 CN CN201610840101.9A patent/CN106503617A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20120321128A1 (en) * | 2008-04-01 | 2012-12-20 | University Of Southern California | Video feed target tracking |
CN101853400A (en) * | 2010-05-20 | 2010-10-06 | 武汉大学 | Multiclass image classification method based on active learning and semi-supervised learning |
CN103166830A (en) * | 2011-12-14 | 2013-06-19 | 中国电信股份有限公司 | Spam email filtering system and method capable of intelligently selecting training samples |
CN103680495A (en) * | 2012-09-26 | 2014-03-26 | 中国移动通信集团公司 | Speech recognition model training method, speech recognition model training device and terminal |
CN105320957A (en) * | 2014-07-10 | 2016-02-10 | 腾讯科技(深圳)有限公司 | Classifier training method and device |
CN105426857A (en) * | 2015-11-25 | 2016-03-23 | 小米科技有限责任公司 | Training method and device of face recognition model |
Non-Patent Citations (1)
Title |
---|
何星等: ""具有选择与补偿机制的加权集合序贯极端学习机及其应用"", 《系统工程理论与实践》 * |
Cited By (38)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107426147A (en) * | 2016-03-28 | 2017-12-01 | 阿里巴巴集团控股有限公司 | For the method and apparatus for the anti-spam performance for determining application |
CN107194464A (en) * | 2017-04-25 | 2017-09-22 | 北京小米移动软件有限公司 | The training method and device of convolutional neural networks model |
CN107194464B (en) * | 2017-04-25 | 2021-06-01 | 北京小米移动软件有限公司 | Training method and device of convolutional neural network model |
CN107292154A (en) * | 2017-06-09 | 2017-10-24 | 北京奇安信科技有限公司 | A kind of terminal feature recognition methods and system |
CN107341457A (en) * | 2017-06-21 | 2017-11-10 | 北京小米移动软件有限公司 | Method for detecting human face and device |
CN107463906A (en) * | 2017-08-08 | 2017-12-12 | 深图(厦门)科技有限公司 | The method and device of Face datection |
CN107491790A (en) * | 2017-08-25 | 2017-12-19 | 北京图森未来科技有限公司 | A kind of neural network training method and device |
CN108197777A (en) * | 2017-12-14 | 2018-06-22 | 阿里巴巴集团控股有限公司 | A kind of method, apparatus and equipment for adjusting air control rule |
CN108108821A (en) * | 2017-12-29 | 2018-06-01 | 广东欧珀移动通信有限公司 | Model training method and device |
US11475244B2 (en) | 2017-12-29 | 2022-10-18 | Guangdong Oppo Mobile Telecommunications Corp., Ltd. | Method for training model and information recommendation system |
CN108108821B (en) * | 2017-12-29 | 2022-04-22 | Oppo广东移动通信有限公司 | Model training method and device |
CN108304868A (en) * | 2018-01-25 | 2018-07-20 | 阿里巴巴集团控股有限公司 | Model training method, data type recognition methods and computer equipment |
US11544496B2 (en) | 2018-03-23 | 2023-01-03 | Beijing Dajia Internet Information Technology Co., Ltd. | Method for optimizing image classification model, and terminal and storage medium thereof |
WO2019179189A1 (en) * | 2018-03-23 | 2019-09-26 | 北京达佳互联信息技术有限公司 | Image classification model optimization method and device and terminal |
CN109214175B (en) * | 2018-07-23 | 2021-11-16 | 中国科学院计算机网络信息中心 | Method, device and storage medium for training classifier based on sample characteristics |
CN109214175A (en) * | 2018-07-23 | 2019-01-15 | 中国科学院计算机网络信息中心 | Method, apparatus and storage medium based on sample characteristics training classifier |
US11302118B2 (en) | 2018-08-02 | 2022-04-12 | Tencent Technology (Shenzhen) Company Limited | Method and apparatus for generating negative sample of face recognition, and computer device |
WO2020024737A1 (en) * | 2018-08-02 | 2020-02-06 | 腾讯科技(深圳)有限公司 | Method and apparatus for generating negative sample of face recognition, and computer device |
CN110998648A (en) * | 2018-08-09 | 2020-04-10 | 北京嘀嘀无限科技发展有限公司 | System and method for distributing orders |
CN109344862B (en) * | 2018-08-21 | 2023-11-28 | 中国平安人寿保险股份有限公司 | Positive sample acquisition method, device, computer equipment and storage medium |
CN109344862A (en) * | 2018-08-21 | 2019-02-15 | 中国平安人寿保险股份有限公司 | Acquisition methods, device, computer equipment and the storage medium of positive sample |
CN109326286A (en) * | 2018-10-23 | 2019-02-12 | 出门问问信息科技有限公司 | Voice information processing method, device and electronic equipment |
CN111199175A (en) * | 2018-11-20 | 2020-05-26 | 株式会社日立制作所 | Training method and device for target detection network model |
CN109359056A (en) * | 2018-12-21 | 2019-02-19 | 北京搜狗科技发展有限公司 | A kind of applied program testing method and device |
CN110533057B (en) * | 2019-04-29 | 2022-08-12 | 浙江科技学院 | Chinese character verification code identification method under single-sample and few-sample scene |
CN110533057A (en) * | 2019-04-29 | 2019-12-03 | 浙江科技学院 | A kind of Chinese character method for recognizing verification code under list sample and few sample scene |
CN110460488A (en) * | 2019-07-01 | 2019-11-15 | 华为技术有限公司 | Business stream recognition method and device, model generating method and device |
CN110969210A (en) * | 2019-12-02 | 2020-04-07 | 中电科特种飞机系统工程有限公司 | Small and slow target identification and classification method, device, equipment and storage medium |
CN111062342B (en) * | 2019-12-20 | 2023-04-07 | 中国银行股份有限公司 | Debugging method and device of face recognition system |
CN111062342A (en) * | 2019-12-20 | 2020-04-24 | 中国银行股份有限公司 | Debugging method and device of face recognition system |
CN112259085A (en) * | 2020-09-28 | 2021-01-22 | 上海声瀚信息科技有限公司 | Two-stage voice awakening algorithm based on model fusion framework |
CN112257670A (en) * | 2020-11-16 | 2021-01-22 | 北京爱笔科技有限公司 | Image processing model and machine learning model training method and device |
CN112800403A (en) * | 2021-01-05 | 2021-05-14 | 北京小米松果电子有限公司 | Method, apparatus and medium for generating prediction model and predicting fingerprint recognition abnormality |
CN113343051A (en) * | 2021-06-04 | 2021-09-03 | 全球能源互联网研究院有限公司 | Abnormal SQL detection model construction method and detection method |
CN113343051B (en) * | 2021-06-04 | 2024-04-16 | 全球能源互联网研究院有限公司 | Abnormal SQL detection model construction method and detection method |
CN113591782A (en) * | 2021-08-12 | 2021-11-02 | 北京惠朗时代科技有限公司 | Training-based face recognition intelligent safety box application method and system |
CN113887283A (en) * | 2021-08-30 | 2022-01-04 | 兴圣创(苏州)科技有限公司 | Strange face recognition score threshold determination method, face recognition method and system |
CN113780485A (en) * | 2021-11-12 | 2021-12-10 | 浙江大华技术股份有限公司 | Image acquisition, target recognition and model training method and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106503617A (en) | Model training method and device | |
CN104850828B (en) | Character recognition method and device | |
CN106295511B (en) | Face tracking method and device | |
CN107454465A (en) | Video playback progress display method and device, electronic equipment | |
CN105488527A (en) | Image classification method and apparatus | |
CN106650575A (en) | Face detection method and device | |
CN106651955A (en) | Method and device for positioning object in picture | |
CN106951884A (en) | Gather method, device and the electronic equipment of fingerprint | |
CN105095881A (en) | Method, apparatus and terminal for face identification | |
CN105631408A (en) | Video-based face album processing method and processing device | |
CN105224924A (en) | Living body faces recognition methods and device | |
CN103886284B (en) | Character attribute information identifying method, device and electronic equipment | |
CN107832036A (en) | Sound control method, device and computer-readable recording medium | |
CN106454336A (en) | Method and device for detecting whether camera of terminal is covered or not, and terminal | |
CN104077563B (en) | Face identification method and device | |
CN106250921A (en) | Image processing method and device | |
CN106682736A (en) | Image identification method and apparatus | |
CN106228556A (en) | Image quality analysis method and device | |
CN104238912A (en) | Application control method and application control device | |
CN106652113A (en) | Access control method and device | |
CN109360197A (en) | Processing method, device, electronic equipment and the storage medium of image | |
CN104461014A (en) | Screen unlocking method and device | |
CN105139033A (en) | Classifier construction method and device and image processing method and device | |
CN106557759A (en) | A kind of sign board information getting method and device | |
CN106503628A (en) | method and device for fingerprint matching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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: 20170315 |