CN109740691A - The training device and training system of graph data identification - Google Patents
The training device and training system of graph data identification Download PDFInfo
- Publication number
- CN109740691A CN109740691A CN201910024581.5A CN201910024581A CN109740691A CN 109740691 A CN109740691 A CN 109740691A CN 201910024581 A CN201910024581 A CN 201910024581A CN 109740691 A CN109740691 A CN 109740691A
- Authority
- CN
- China
- Prior art keywords
- module
- training
- event data
- data
- sample
- 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
- 238000013500 data storage Methods 0.000 claims abstract description 62
- 230000005540 biological transmission Effects 0.000 claims description 19
- 238000009434 installation Methods 0.000 claims description 17
- 238000004891 communication Methods 0.000 claims description 13
- 238000000034 method Methods 0.000 claims description 6
- 238000003909 pattern recognition Methods 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 4
- 241001269238 Data Species 0.000 description 3
- 235000013399 edible fruits Nutrition 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000003014 reinforcing effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
Landscapes
- Electrically Operated Instructional Devices (AREA)
- Image Analysis (AREA)
Abstract
A kind of training device and training system of graph data identification, this training device include event data storage and notification module, sample discrimination module and training module.Event data storage module stores the data after tentatively recognizing；Sample discrimination module carries out more accurate identification to this data and obtains accurate identification result, and according to the difference of accurate identification result and identification result and aforementioned data, decides whether the training for notifying training module to carry out graph data identification.The present invention can increase pattern recognition and learn the part for needing manually to participate in the chance of training, reduction automatically, reduce the time spent needed for forming effective image identification data set and cost.
Description
Technical field
The present invention about a kind of data identification training device and training system, in particular to a kind of graph data
The training device and training system of identification.
Background technique
Image identification is one of center of gravity of the prior art, and many researchers are intended to can be by computer vision
It technically makes a breakthrough, and can further be applied in the true world and helpful to the mankind.However, in order to want
It improves computer vision system or training system to be enable to pick out diversified object, be typically necessary billions of images
The data set of scale can just train qualified image identification model.
However, existing image identification model is all by mankind's hand labeled picture come as training dataset, therefore
Even if such learning method can train accurately identification model, but since a large amount of data set is required by artificial
It is marked, so a large amount of cost and time will be expended, in a disguised form causes limitation to promoting image identification.
Summary of the invention
In view of this, figure can be increased the present invention provides the training device and training system of a kind of identification of graph data
The automatic chance for learning and training of shape identification reduces the part for needing manually to participate in, and reduction forms effective image identification data
The time spent needed for collection and cost.
From the point of view of an angle, the present invention provides a kind of training devices of graph data identification.This training device includes
Event data storage and notification module, sample discrimination module and training module.Event data storage and notification module receive simultaneously
At least one event data to be trained is stored, the storage of this event data and notification module calculate the received event data to be trained of institute
Quantity, and when this quantity reaches preset value issue sample notification signal.Sample discrimination module is electrically coupled to event data
Storage and notification module, this sample discrimination module take after receiving sample notification signal from event data storage and notification module
Event data train above-mentioned, recognize acquired by each event data to be trained to obtain corresponding identification result,
Corresponding event data to be trained is handled to obtain corresponding sample data according to identification result, and is sent out after obtaining sample data
Notification signal is trained out.Training module is electrically coupled to event data storage and notification module and sample discrimination module, this training
Module obtains sample data corresponding with training notification signal, and needle from sample discrimination module after receiving trained notification signal
Sample data is trained.
In one embodiment, the training device of graph data identification further includes notification module and input module.Notice
Module is electrically coupled to sample discrimination module, when sample discrimination module obtain identification result expression can not recognize it is corresponding wait instruct
Practice event data, sample discrimination module makes notification module give notice information.Input module is electrically coupled to sample discrimination module,
And input module is suitable for inputting input information corresponding with the event data to be trained that can not be recognized, and this input information is passed
Sample discrimination module is handed to as sample data corresponding with the event data to be trained that can not be recognized.
In one embodiment, each event data to be trained includes a graph data and a preliminary identification number
According to each of sample discrimination module acquired by identification is when trained event data is to obtain corresponding identification result, to figure
Graphic data is recognized to obtain Identification Data before corresponding training, and corresponding training preceding Identification Data expression that can not recognize
When Identification Data is not exactly the same with preliminary Identification Data before graph data or training, identification result above-mentioned is generated.
In one embodiment, training module above-mentioned also generates one after the training that complete paired-sample carries out
Model is recognized after training and is stored to event data storage and notification module.
In one embodiment, event data storage above-mentioned and notification module include event data storage module and thing
Part transmission path module.Wherein, event data storage module is electrically coupled to sample discrimination module and training module, and event number
It is received according to storage module and stores event data to be trained above-mentioned；Event transmission path module is electrically coupled to sample and differentiates mould
Block and training module, and event transmission path module is suitable for calculating the received event data to be trained of event data storage module institute
Quantity, sample notification signal is issued when quantity reaches preset value to sample discrimination module, and by training notice letter above-mentioned
Number turn reach training module.
From another perspective, the present invention also provides a kind of training system of graph data identification, this training systems
Including user's terminal installation and training device above-mentioned.Wherein, user's terminal installation includes processor and communication member
Part.Processor recognizes each graph data and obtains corresponding preliminary Identification Data, and by each graph data and phase
The preliminary Identification Data answered is integrated into event data to be trained；Communication device is transmitted outward from user's terminal installation wait train
Event data.Training device includes event data storage and notification module, sample discrimination module and training module.Event data
Storage and notification module are received and are stored from the received event data to be trained of communication device institute, calculate the received thing to be trained of institute
The quantity of number of packages evidence, and sample notification signal is issued when this quantity reaches preset value.Sample discrimination module is electrically coupled to thing
Part data storage and notification module, this sample discrimination module are stored and are notified from event data after receiving sample notification signal
Module obtains event data to be trained above-mentioned, recognizes each acquired event data to be trained to obtain corresponding identification
As a result, handling corresponding event data to be trained according to identification result to obtain corresponding sample data, and obtaining sample number
According to rear sending training notification signal.Training module is electrically coupled to event data storage and notification module and sample discrimination module,
This training module obtains sample number corresponding with training notification signal from sample discrimination module after receiving trained notification signal
According to, and be trained for sample data.
In one embodiment, communication device above-mentioned also waits training transmitting each outward from user's terminal installation
An event notification signal is transmitted when event data outward, and event data storage and notification module include event data storage mould
Block and event transmission path module.Wherein, event data storage module is electrically coupled to sample discrimination module and training module,
And event data storage module receives and stores event data to be trained above-mentioned；Event transmission path module is electrically coupled to sample
This discrimination module and training module, and event transmission path module receives event notification signal above-mentioned, and is notified according to event
Signal come calculate event data storage module received event data to be trained quantity, the sending when quantity reaches preset value
Sample notification signal turns trained notification signal above-mentioned to reach training module to sample discrimination module.
By above-mentioned technology, the training device and training system of graph data identification provided by the present invention can be persistently right
It is recognized automatically in newly-increased event data to be trained, and automatically into training program when can successfully recognize, because
This can increase the chance that pattern recognition learns automatically with trains, and reduce the part for needing manually to participate in, and reduction formation is effective to be schemed
The time of the cost as needed for Identification Data collection and cost.
Detailed description of the invention
Fig. 1 is the circuit block diagram of the training device recognized according to the graph data of one embodiment of the invention.
Fig. 2 is the framework block schematic diagram of the training system recognized according to the graph data of one embodiment of the invention.
Wherein, symbol is simply described as follows in attached drawing.
10,10a: training device；100: event data storage and notification module；110: sample discrimination module；120: training
Module；130: notification module；140: input module；1002: event data storage module；1004: event transmission path module；
20: training system；200: user's terminal installation；250: network；2010: processor；2020: communication device.
Specific embodiment
Fig. 1 is please referred to, for the circuit block diagram of the training device recognized according to the graph data of one embodiment of the invention.
In the present embodiment, training device 10 mainly include event data storage and notification module 100, sample discrimination module 110 and
Training module 120.Wherein, event data storage and notification module 100 receive and store at a few event data to be trained, meter
Calculate institute received event data train quantity, and when the quantity being calculated reaches preset value sending sample notify
Signal is to sample discrimination module 110.Sample discrimination module 110 after receiving sample notification signal, from event data storage and
Notification module 100 obtains event data to be trained above-mentioned, recognizes each acquired event data to be trained to obtain pair
The identification result answered handles corresponding event data to be trained according to identification result to obtain corresponding sample data, and is obtaining
Training notification signal is issued to training module 120 after sampling notebook data.Training module 120 be electrically coupled to event data storage and
Notification module 100 and sample discrimination module 110, wherein training module 120 is sentenced after receiving trained notification signal from sample
Other module 110 obtains sample data corresponding with received training notification signal, and is directed to received sample data
It is trained.
It is noted that in one embodiment, event data storage above-mentioned and notification module 100 can be used to store
Device is formed with the mode combinations that the logic circuit that energy number of executions calculated and issued signal merges running；Alternatively, at another
In embodiment, event data storage above-mentioned and notification module 100 be can be used with storage device, processor and energy execution module
The mode combinations that the software of required running merges running form.On the other hand, due to sample discrimination module 110 and training module
120 need to be implemented more complicated operation, therefore usually can be by transport needed for storage device, processor and energy execution module
The mode combinations of the merging such as the software of work running form.
In the present embodiment, event data storage and notification module 100 include event data storage module 1002 with
An and event transmission path module 1004.When event data storage and notification module 100 receive event data to be trained
When, received event data to be trained can be stored among event data storage module 1002.Moreover, event data
Storage and notification module 100 when receiving new event data train, also can notification event transmitting simultaneously lead to
Road module 1004, so that event transmission path module 1004 can correctly calculate event data storage module 1002 accordingly and be connect
The quantity of the event data to be trained received.When quantity that event transmission path module 1004 is calculated reach preset value when
It waits, means that and stored enough for trained sample data, event at this time in event data storage module 1002
Transmission path module 1004 will issue sample notification signal to sample discrimination module 110, make sample discrimination module 110 start into
Row running.
After receiving sample notification signal, sample discrimination module 110 can be obtained from event data storage module 1002
With the comparable event data to be trained of preset value quantity above-mentioned.If then sample differentiates for example, preset value above-mentioned is 1000
Module 110 will obtain 1000 event datas to be trained from event data storage module 1002；Similar, if it is above-mentioned pre-
If value is 1, then sample discrimination module 110 will obtain an event data to be trained from event data storage module 1002.One
As for, each pen event data to be trained will include a graph data and carry out to this graph data more simple
Obtained preliminary Identification Data after image identification operation slightly.The more simple image identification operation is by uploading this
Performed by the device of pen event data to be trained, and it then may be included in this graph data and recognize in preliminary Identification Data
The content of the quantity of image out, the position of each image being identified and each image being identified.Therefore,
In the present embodiment, each pen event number to be trained that sample discrimination module 110 is obtained from event data storage module 1002
According to a corresponding graph data and a corresponding preliminary Identification Data can be respectively included.
After obtaining an event data to be trained, sample discrimination module 110 can be based on existing data database (not
Be painted) in image identification model, comparatively detailed figure is carried out for the graph data in this event data to be trained
Shape identification operation.After carrying out comparatively detailed pattern recognition operation to graph data, sample discrimination module 110 can be with
It obtains and this pen event data to be trained (in other words, this graph data) corresponding identification result.Due to preceding
The preliminary Identification Data stated may include: the quantity of the image picked out in this graph data, each is identified
Image position and each image being identified the fields such as content, so as acquired in sample discrimination module 110
Identification result generally also can include same field.Therefore, sample discrimination module 110 can be by identification result and preliminary identification
Data are compared, and determine the mode that event data to be trained is processed into sample data according to resulting conclusion is compared.
In the present embodiment, resulting knot is compared with preliminary Identification Data in identification result by sample discrimination module 110
By being divided into three types:
Type one: the identification result of sample discrimination module 110 is identical with preliminary Identification Data.It is tied when being stored in identification
The quantity of the image picked out in fruit, the position of each image being identified and each image being identified
The information filled in the fields such as content, with the information filled in the corresponding field of preliminary Identification Data it is all identical when, sample
Identification result can be considered as identical as preliminary Identification Data by discrimination module 110.At this moment, sample discrimination module 110, is even instructed
Practice device 10, so that it may directly delete event data to be trained corresponding with this identification result.That is, due to sample
There is no the discovery in graph data and preliminary Identification Datas for the more detailed pattern recognition operation that discrimination module 110 is carried out
Different contents, therefore indicate the part not in need for especially carrying out identification training in this graph data.Then, this
As soon as event data to be trained can be directly deleted without entering in the mechanism for reinforcing identification training, so also not
Corresponding sample data can be generated.
Type two: the identification result of sample discrimination module 110 has differences with preliminary Identification Data.It is tied when being stored in identification
The quantity of the image picked out in fruit, the position of each image being identified or each image being identified
The information filled in the fields such as content, when being had differences with the information filled in the corresponding field of preliminary Identification Data, sample
This discrimination module 110, which just will be considered that, to need to carry out identification training to this graph data, the figure carried out with strengthening subsequent whereby
The accuracy of shape identification.Therefore, for the identification result being compared, generate this identification result graph data and with
The corresponding preliminary Identification Data of this graph data, that is, be used to the identification result that is compared and generate this distinguish
The event data to be trained for knowing result, will be integrated into a sample data.
Type three: the identification result of sample discrimination module 110 shows the part that can not be recognized.When sample discrimination module
110 in graph data when find the image that can not be recognized, no matter whether this image was once recorded in preliminary identification number
In, the quantity for the image that sample discrimination module 110 can not be identified under all noting down in identification result and position.It
Afterwards, sample discrimination module 110 can store identification result and event data to be trained corresponding with this identification result
For a data (subsequent be known as can not Identification Data).
It is aforementioned by sample discrimination module 110 store can not Identification Data, can further by introduce manually sentence
Break to obtain correct image identification result.Wherein, the mode for introducing artificial judgment has very much, and then utilizes in the present embodiment
Notification module 130 completes the task in this stage with input module 140.Please refer to Fig. 1, in the present embodiment, notification module
130 are electrically coupled to sample discrimination module 110 with input module 140 respectively；When the identification result that sample discrimination module 110 obtains
Corresponding event data to be trained can not be recognized by sample discrimination module 110 (that is, sample discrimination module 110 produces nothing by illustrating
Method Identification Data) when, sample discrimination module 110 can make notification module 130 issue a notification information；Input module 140
Suitable for input and the event data to be trained that can not be recognized or can not the corresponding input information of Identification Data, and input module
This input information can be transferred to sample discrimination module 110 by 140, as corresponding with the event data to be trained that can not be recognized
Sample data.
Specifically, in the present embodiment, notification module 130 can be with e-mail terminal program or web page browsing
The display device of device, user then can know sample by the notification information shown on browsing Email or web browser
What this discrimination module 110 generated can not Identification Data.Furthermore the input module 140 in the present embodiment can be keyboard, mouse with
And can show the combination of the display device of selection project and input content, user can pass through operation input module 140
Select it is to be dealt with can not Identification Data, and input with it is selected can not the corresponding content of Identification Data.It is inputting
After completion, the content inputted and corresponding graph data can be integrated into input information above-mentioned by input module 140,
And input information obtained by integration is transferred to sample discrimination module 110.Sample discrimination module 110 is being obtained from input module
After the 140 input information transmitted, input information can be stored as sample data.
According to above-mentioned, sample discrimination module 110 obtains type two or class comparing identification result and preliminary Identification Data
After the result of type three, can while or it obtain generating corresponding sample data later.And after sample data generation, sample
This discrimination module 110 can issue trained notification signal to training module 120.It should be noted that sample discrimination module 110 can be with
Primary training notification signal is issued after the generation of each sample data to training module 120, alternatively, sample discrimination module
110 can also just issue primary training notification signal after the generation of more than two sample datas.
In the present embodiment, training module 120 is electrically coupled to event data storage and notification module 100 and sample differentiates
Module 110.After receiving trained notification signal from sample discrimination module 110, training module 120 can be from sample discrimination module
110 obtain the sample data of predetermined quantity, and the training of image identification is carried out for acquired sample data.In training
The training of the image identification executed in module 120, can be using any training image identification capability used in the prior art
The variation of mechanism or method, this part has no effect on execution of the invention.And whenever training module 120 is by figure one or more times
After training, the image identification model for being used to carry out image identification originally may can be used to trained sample because of being directed to for picture identification
The optimization of notebook data and change.Training module 120 can be (subsequent to be known as by the image identification model for occurring obtaining after changing
Model is recognized after training) storage to data database above-mentioned or event data storage and notification module 100 in, so as to it is subsequent in
It is used when sample 110 recognisable image of discrimination module.
It next referring to figure 2., is the framework of the training system recognized according to the graph data of one embodiment of the invention
Block schematic diagram.In the present embodiment, training system 20 includes user's terminal installation 200 and training device 10a.Training
Device 10a is generally identical as training device 10 above-mentioned, thus in the present embodiment no longer to training device 10a make it is complete and
Detailed description.It, will be in following merging user terminal installation 200 as not existing together for training device 10a and training device 10
It is illustrated together.
As shown in Fig. 2, user's terminal installation 200 includes processor 2010 and communication device 2020.In the present embodiment,
User's terminal installation 200 can be smart phone and portable device that is similar, can simply moving, be also possible to large-scale terminal
The heavy devices such as machine.Anyway, processor 2010 used in user's terminal installation 200 can execute one and be used to carry out
The application program of image identification, and graph data is carried out before resulting result is exactly after image identification by processor 2010
The preliminary Identification Data stated.Each graph data can be integrated into corresponding preliminary Identification Data wait instruct by processor 2010
Practice event data, and is transmitted outward via communication device 2020 from user's terminal installation 200.
In the present embodiment, event data to be trained is transferred to event data by network 250 by communication device 2020
Storage module 1002.And an event data to be trained is transmitted outward from user's terminal installation 200 in communication device 2020
When, communication device 2020, which can also synchronize, externally transmits an event notification signal to event transmission path module 1004.Event
Transmission path module 1004 receives the event notification signal transmitted by communication device 2020 by network 250, and logical according to event
Know signal calculate event data storage module 1002 received event data to be trained quantity, in quantity reach preset value
When issue sample notification signal to sample discrimination module 110, and the training notification signal that sample discrimination module 110 is issued turns
Reach training module 120.
In the present embodiment, in pattern recognition, used identification model can be divided into two kinds, and one is need pole
Big storage space, " heavyweight " suitable for being used in training device 10a recognize model, another then be needed storage space
" lightweight " identification model that is relatively small, being suitable for use on light handheld apparatus.Training module 120 can update simultaneously
Both identification models recognize model after becoming training above-mentioned, and are stored into event data storage for model is recognized after training
Module 1002.While identification model is to event data storage module 1002 after storage training, training module 120 also issues letter
Number to event transmission path module 1004 so that event transmission path module 1004 can notify user's terminal by network 250
Device 200, so that user's terminal installation 200 can recognize model and the image that is improved after suitable time point downloads training
The ability of identification.
It is noted that each training device 10a can service many users simultaneously in training system 20
Terminal installation 200.Provided technology through the invention, training device 10a can be simultaneously from many user's terminal installations
200 obtain event data to be trained and are largely trained, and then accelerate the improvement speed of image identification ability.
In conclusion the training device and training system of graph data identification provided by the present invention can continue for new
The event data to be trained increased is recognized automatically, and automatically into training program when can successfully recognize, therefore can
Increase pattern recognition and learn the part for needing manually to participate in the chance of training, reduction automatically, reduction forms effective image and distinguishes
Know the time spent needed for data set and cost.
Claims (10)
1. a kind of training device of graph data identification characterized by comprising
Event data storage and notification module, receive and store at a few event data to be trained, which stores and lead to
Know module calculate received at least one event data to be trained quantity, and issue sample when the quantity reaches preset value
Notification signal；
Sample discrimination module, is electrically coupled to event data storage and notification module, the sample discrimination module are receiving this
At least one event data to be trained is obtained from event data storage and notification module when sample notification signal, recognizes institute
Each at least one event data to be trained obtained handles this extremely according to the identification result to obtain corresponding identification result
The corresponding event data to be trained of a few event data to be trained is obtaining the sample number to obtain corresponding sample data
According to when issue training notification signal；And
Training module, is electrically coupled to event data storage and notification module and the sample discrimination module, the training module exist
The sample data corresponding with the training notification signal is obtained from the sample discrimination module when receiving the training notification signal, and
It is trained for the sample data.
2. training device according to claim 1, which is characterized in that further include:
Notification module is electrically coupled to the sample discrimination module, when the sample discrimination module obtain the identification result indicate without
Method recognize this at least one when training event data it is corresponding when training event data when, which makes the notice mould
Block is given notice information；And
Input module is electrically coupled to the sample discrimination module, is somebody's turn to do event data phase to be trained with what can not be recognized suitable for inputting
Corresponding input information, and the input information is transferred to the sample discrimination module as with what can not be recognized and is somebody's turn to do event to be trained
The corresponding sample data of data.
3. training device according to claim 1, which is characterized in that each at least one event data to be trained includes figure
Graphic data and preliminary Identification Data, the sample discrimination module each at least one to be trained event data acquired in identification
When obtaining the corresponding identification result, which is recognized to obtain Identification Data before corresponding training, and
Identification Data expression can not recognize Identification Data and preliminary identification number before the corresponding graph data or the training before the training
According to it is not exactly the same when, generate the identification result.
4. training device according to claim 1, which is characterized in that the training module completes the instruction to the sample data
After white silk, model is recognized after also generating training and identification model will store to event data storage after the training and notify mould
Block.
5. training device according to claim 1, which is characterized in that event data storage and notification module include:
Event data storage module is electrically coupled to the sample discrimination module and the training module, the event data storage module
It receives and stores at least one event data to be trained；And
Event transmission path module is electrically coupled to the sample discrimination module and the training module, the event transmission path module
Suitable for calculate the event data storage module received at least one event data to be trained the quantity, reach in the quantity
The sample notification signal is issued when the preset value to the sample discrimination module, and the training notification signal is turned to reach the training mould
Block.
6. a kind of training system of graph data identification characterized by comprising
User's terminal installation, comprising:
Processor, identification at least each of graph data and obtain corresponding preliminary Identification Data, and by each
An at least graph data is integrated at least one event data to be trained with the corresponding preliminary Identification Data；And
Communication device transmits at least one event data to be trained from user's terminal installation outward；And
Training device, comprising:
Event data storage and notification module, receive and store at a few event data to be trained, which stores and lead to
Know module calculate received at least one event data to be trained quantity, and issue sample when the quantity reaches preset value
Notification signal；
Sample discrimination module, is electrically coupled to event data storage and notification module, the sample discrimination module are receiving this
At least one event data to be trained is obtained from event data storage and notification module when sample notification signal, recognizes institute
Each at least one event data to be trained obtained handles this extremely according to the identification result to obtain corresponding identification result
The corresponding event data to be trained of a few event data to be trained is obtaining the sample number to obtain corresponding sample data
According to when issue training notification signal；And
Training module, is electrically coupled to event data storage and notification module and the sample discrimination module, the training module exist
The sample data corresponding with the training notification signal is obtained from the sample discrimination module when receiving the training notification signal, and
It is trained for the sample data.
7. training system according to claim 6, which is characterized in that further include:
Notification module is electrically coupled to the sample discrimination module, when the sample discrimination module obtain the identification result indicate without
Method recognize this at least one when training event data it is corresponding when training event data when, which makes the notice mould
Block is given notice information；And
Input module is electrically coupled to the sample discrimination module, is somebody's turn to do event data phase to be trained with what can not be recognized suitable for inputting
Corresponding input information, and the input information is transferred to the sample discrimination module as with what can not be recognized and is somebody's turn to do event to be trained
The corresponding sample data of data.
8. training system according to claim 6, which is characterized in that the sample discrimination module is each acquired by identification
It is a this at least one when training event data to obtain the corresponding identification result when, to the graph data recognized with obtain pair
Identification Data before the training answered, and Identification Data indicates before can not recognizing the corresponding graph data or the training before the training
When Identification Data and the not exactly the same preliminary Identification Data, the identification result is generated.
9. training system according to claim 6, which is characterized in that the training module is completed to carry out the sample data
After training, model, which is recognized, after also generating training and will recognize model after the training stores to event data storage and notify
Module.
10. training system according to claim 6, which is characterized in that the communication device is also filled from user's terminal
Set transmit outward each this at least one when training event data, also transmitting event notification signal, and the event data outward
Storage and notification module include:
Event data storage module is electrically coupled to the sample discrimination module and the training module, the event data storage module
It receives and stores at least one event data to be trained；And
Event transmission path module is electrically coupled to the sample discrimination module and the training module, the event transmission path module
Receive the event notification signal, and calculated according to the event notification signal event data storage module institute it is received this at least
The quantity of one event data to be trained issues the sample notification signal to the sample when the quantity reaches the preset value and differentiates
Module, and the training notification signal is turned to reach the training module.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| TW107214944U TWM576314U (en) | 2018-11-02 | 2018-11-02 | Training apparatus and training system for graphic data identification |
| TW107214944 | 2018-11-02 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109740691A true CN109740691A (en) | 2019-05-10 |
Family
ID=66364408
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910024581.5A Pending CN109740691A (en) | 2018-11-02 | 2019-01-10 | The training device and training system of graph data identification |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109740691A (en) |
| TW (1) | TWM576314U (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI708190B (en) | 2019-11-15 | 2020-10-21 | 財團法人工業技術研究院 | Image recognition method, training system of object recognition model and training method of object recognition model |
-
2018
- 2018-11-02 TW TW107214944U patent/TWM576314U/en unknown
-
2019
- 2019-01-10 CN CN201910024581.5A patent/CN109740691A/en active Pending
Also Published As
| Publication number | Publication date |
|---|---|
| TWM576314U (en) | 2019-04-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111507104B (en) | Method and device for establishing label labeling model, electronic equipment and readable storage medium | |
| CN108416003A (en) | A kind of picture classification method and device, terminal, storage medium | |
| CN106898173A (en) | The online review system in real time of students' work | |
| CN115658955B (en) | Cross-media retrieval and model training method, device, equipment and menu retrieval system | |
| CN111881908B (en) | Target detection model correction method, detection device, equipment and medium | |
| CN113939791A (en) | Image labeling method, device, equipment and medium | |
| WO2021190470A1 (en) | Electronic device positioning method and apparatus, and server, system and storage medium | |
| CN117671678A (en) | Image labeling method and device | |
| CN111666771B (en) | Semantic tag extraction device, electronic equipment and readable storage medium for document | |
| CN110991279A (en) | Document image analysis and recognition method and system | |
| CN113723926A (en) | Bank pipelining processing method and device combining RPA and AI and electronic equipment | |
| CN117954045B (en) | Automatic drug sorting management system and method based on prescription data analysis | |
| WO2019066231A1 (en) | Representative image generation | |
| CN111726576A (en) | Unmanned aerial vehicle inspection method, device, system and storage medium | |
| CN109740691A (en) | The training device and training system of graph data identification | |
| KR102026475B1 (en) | Processing visual input | |
| CN114241501A (en) | Image document processing method and device and electronic equipment | |
| CN114676705A (en) | Dialogue relation processing method, computer and readable storage medium | |
| CN111768007B (en) | Method and device for mining data | |
| CN113822143A (en) | Text image processing method, device, equipment and storage medium | |
| CN115546824B (en) | Taboo picture identification method, apparatus and storage medium | |
| CN101140561B (en) | Course data processing system and method | |
| CN115631374A (en) | Control operation method, control detection model training method, device and equipment | |
| CN114285114A (en) | Charging control method and device, electronic equipment and storage medium | |
| CN109040774B (en) | Program information extraction method, terminal equipment, server and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |