CN108053031A - A kind of deep learning system and deep learning recognition methods - Google Patents
A kind of deep learning system and deep learning recognition methods Download PDFInfo
- Publication number
- CN108053031A CN108053031A CN201810002406.1A CN201810002406A CN108053031A CN 108053031 A CN108053031 A CN 108053031A CN 201810002406 A CN201810002406 A CN 201810002406A CN 108053031 A CN108053031 A CN 108053031A
- Authority
- CN
- China
- Prior art keywords
- data
- deep learning
- identified
- algorithm model
- learning algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of deep learning systems and deep learning recognition methods, system to include：Data Input Interface, Switching Module, memory module, control module, computing module and data output interface, wherein, the data that Data Input Interface inputs for reception；Switching Module is used to judge whether by institute's data to be identified and the storage of the first deep learning algorithm model into memory module；Memory module is used to, in the case where the judging result of Switching Module is to be, store data to be identified and the first deep learning algorithm model；Control module is used to send the identification instruction for data to be identified to computing module；Computing module is used to that processing to be identified to data to be identified using the first deep learning algorithm model according to identification instruction；Data output interface is used to export the recognition result of computing module and other equipment is directed to the recognition result of data to be identified.Using the embodiment of the present invention, the recognition accuracy to data to be identified can be improved.
Description
Technical field
The present invention relates to depth learning technology fields, and in particular to a kind of deep learning system and deep learning identification side
Method.
Background technology
Deep learning is a branch of machine learning, it is therefore intended that establishes the god that a simulation human brain carries out analytic learning
Through network, data processing is carried out by imitating the working mechanism of human brain, such as image is typically used in video identification, image is known
Other or voice recognition field.
At present, typically deep learning algorithm model burning is entered in computing module, then computing module is integrated in data
In identification device, by the data collecting module collected data to be identified in data identification means, knowledge is then treated by computing module
Other data are identified.One critically important application field of data above identification method is public safety field, for example, can collect
Into in public safety camera, video or image are shot by public safety camera, then by computing module to public safety
The video or image of camera shooting are identified, for example, can identify the personage included in image, vehicle, animal, build
Build object etc..
But due in data identification process, the factor that is set by the model parameter of deep learning algorithm model or
The factor of the model structure of deep learning algorithm model influences, and causes current data identification means, the knowledge to data to be identified
Other accuracy rate is not high.
The content of the invention
The technical problems to be solved by the invention are the provision of a kind of deep learning system and deep learning recognition methods.
The present invention is to solve above-mentioned technical problem by the following technical programs：
An embodiment of the present invention provides a kind of deep learning system, the system comprises：Data Input Interface, interchange mode
Block, memory module, control module, computing module and data output interface, wherein,
The Data Input Interface, for receiving the data of input, the data of the input include：Data to be identified,
One deep learning algorithm model and other equipment are directed to the knowledge of the data to be identified using the second deep learning algorithm model
Other result；
The Switching Module is calculated for the identification information according to the Data Data to be identified and first deep learning
The identification information of method model judges whether to store institute's data to be identified and the first deep learning algorithm model to described
In memory module；
The memory module, in the case of being in the judging result of the Switching Module, storage is described to be identified
Data and the first deep learning algorithm model；
The control module, for sending the identification instruction for the data to be identified to the computing module；
The computing module, for obtaining the first deep learning algorithm model according to the identification instruction and described treating
It identifies data, and processing is identified to the data to be identified using the first deep learning algorithm model；
The data output interface, for exporting the recognition result of the computing module and the other equipment for described
The recognition result of data to be identified.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, the Data Input Interface includes：First number
According to input interface and the second Data Input Interface, wherein, first Data Input Interface, for receiving non-picture data；Institute
The second Data Input Interface is stated, for receiving image data；
The data output interface includes the first data output interface and the second data output interface, wherein, described first
Data output interface, for exporting non-picture data；Second data output interface, for exporting image data.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, first Data Input Interface, described
Two Data Input Interfaces, first data output interface and second data output interface are standard data interface.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, the data output interface, for described
In the case that the judging result of Switching Module is no, the data of the input are exported.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, the control module is specifically used for：To institute
It states computing module and sends training directive；
The computing module, is specifically used for：The first deep learning algorithm model is obtained, instruction is obtained from the memory module
Practice data, and according to the training directive, the first deep learning algorithm model is trained using the training data；
Data to be identified are obtained from the memory module, and using trained first deep learning algorithm model to described to be identified
Processing is identified in data.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, the first deep learning algorithm model is
First deep learning algorithm model of training in advance.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, the computing module is specifically used for：
The first deep learning algorithm model and the first deep learning algorithm model parameter are received, utilizes first depth
Algorithm model parameter is practised to be configured the first deep learning algorithm model.
An embodiment of the present invention provides a kind of deep learning system and deep learning recognition methods, the described method includes：
The data of input are received, the data of the input include：Data to be identified, the first deep learning algorithm model and
Other equipment is directed to the recognition result of the data to be identified using the second deep learning algorithm model；
According to the identification information of the Data Data to be identified and the identification information of the first deep learning algorithm model,
Judge whether storage institute's data to be identified and the first deep learning algorithm model；
In the case where judging whether to store the judging result of the data to be identified to be, first depth is stored
Practise algorithm model and the data to be identified；
The identification obtained for the data to be identified instructs；
It is instructed according to the identification, the data to be identified is identified using the first deep learning algorithm model
Processing, obtains recognition result；
It exports the recognition result and the other equipment is directed to the recognition result of the data to be identified.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, the method further includes：
Obtain the training directive for the first deep learning algorithm model；
Training data is obtained according to the training directive, and using the training data to the first deep learning algorithm
Model is trained, and recycles trained first deep learning algorithm model that processing is identified to data to be identified.
Optionally, in a kind of specific embodiment of the embodiment of the present invention, judge whether to wait to know described in storage described
In the case that the judging result of other data is no, the method further includes：
Export the data to be identified.
The present invention has the following advantages compared with prior art：
Using a kind of deep learning system provided in an embodiment of the present invention and deep learning recognition methods, adopted in miscellaneous equipment
After carrying out data identification with deep learning algorithm model, data identification is carried out using another deep learning algorithm model, is treated more
Identification data are repeatedly identified, can be to avoid the error that bring of single deep learning algorithm model identification, so as to improve pair
The recognition accuracy of data to be identified.
Description of the drawings
Fig. 1 is a kind of structure diagram of deep learning system provided in an embodiment of the present invention；
Fig. 2 is the flow diagram of the first depth deep learning recognition methods provided in an embodiment of the present invention；
Fig. 3 is the flow diagram of second of depth deep learning recognition methods provided in an embodiment of the present invention；
Fig. 4 is the flow diagram of the third depth deep learning recognition methods provided in an embodiment of the present invention.
Specific embodiment
It elaborates below to the embodiment of the present invention, the present embodiment is carried out lower based on the technical solution of the present invention
Implement, give detailed embodiment and specific operating process, but protection scope of the present invention is not limited to following implementation
Example.
To solve prior art problem, an embodiment of the present invention provides a kind of deep learning system and deep learning identification sides
Method, just a kind of deep learning system provided in an embodiment of the present invention is introduced first below.
Fig. 1 is a kind of structure diagram of deep learning system provided in an embodiment of the present invention as shown in Figure 1, the system
Including：Data Input Interface, Switching Module, memory module, control module, computing module and data output interface, wherein,
The Data Input Interface, for receiving the data of input, the data of the input include：Data to be identified,
One deep learning algorithm model and other equipment are directed to the knowledge of the data to be identified using the second deep learning algorithm model
Other result；
The Switching Module is calculated for the identification information according to the Data Data to be identified and first deep learning
The identification information of method model judges whether to store institute's data to be identified and the first deep learning algorithm model to described
In memory module；
The memory module, in the case of being in the judging result of the Switching Module, storage is described to be identified
Data and the first deep learning algorithm model；
The control module, for sending the identification instruction for the data to be identified to the computing module；
The computing module, for obtaining the first deep learning algorithm model according to the identification instruction and described treating
It identifies data, and processing is identified to the data to be identified using the first deep learning algorithm model；
The data output interface, for exporting the recognition result of the computing module and the other equipment for described
The recognition result of data to be identified.
Specifically, Switching Module can by default identification information in the identification information of data to be identified and control module into
Row comparison, if the identification information of data to be identified is identical with default first flag information in control module, Switching Module
Judging result be yes.Default identification information in control module can be the first flag information of user preset, exemplary
, default first flag information is " jpeg image " to user in the control module, and whether Switching Module judges the data received
For jpeg image, if it is, storing that data into memory module.Similar, Switching Module is to the deep learning of reception
Default second identifier information is compared in the identification information and control module of algorithm model, if the deep learning received is calculated
The identification information of method model is consistent with default second identifier information in control module, which is arrived
In memory module.
It is understood that term " identification information " refers to, it can be used in the type of the data of characterization input, the number of input
The flag information of the personage, the vehicle that are included in etc. or refer to the object that is identified of characterization deep learning algorithm model
The information of body and other any data that can uniquely characterize input can uniquely characterize deep learning algorithm model
Information.
Switching Module can all be sent the data that memory module storage is not required by data output interface.
Specifically, the first deep learning algorithm model can be and the second deep learning algorithm model heterogeneous networks framework
The algorithm model of algorithm model or the different identical network framework of model parameter, this is right again for the embodiment of the present invention
It makes restriction.
Specifically, in order to accelerate the arithmetic speed of deep learning algorithm model, computing module can include：Central processing
Device, image processor and deep learning processor, wherein, central processing unit is used for logical process process, and image processor is used for
Image analysis process, deep learning processor are used for deep learning process.
Specifically, in order to be uniformly controlled to Switching Module, computing module, memory module, so that modules can
Co-ordination, control module can also be sent to Switching Module according to the identification information of the Data Data to be identified and described the
The identification information of one deep learning algorithm model judges whether institute's data to be identified and the first deep learning algorithm mould
Type storage is to the instruction in the memory module, so that Switching Module performs above-mentioned steps.Control module can also be to storage mould
Block sends the instruction for storing data to be identified and the first deep learning algorithm model.
In practical applications, deep learning algorithm model includes but are not limited to：Depth belief network model, convolutional Neural
Network model, convolution depth belief network model, large memories and memory neural network model, depth Boltzmann machine mould
Type stacks the models such as autocoder model, deep accumulation network model, tensor depth stacking network model.
It is understood that data to be identified include but are not limited to image data, text data, video data, audio
Data and character string etc..
Using embodiment illustrated in fig. 1 of the present invention, data are carried out using deep learning algorithm model in miscellaneous equipment and identify it
Afterwards, data identification is carried out using another deep learning algorithm model, data how to be identified are repeatedly identified, can be to avoid single
The error that the identification of deep learning algorithm model is brought, so as to improve the recognition accuracy to data to be identified.
In addition, the first deep learning algorithm model in existing pattern recognition device is all to be previously implanted computing module,
The computing module with corresponding first deep learning algorithm model is selected to be installed during use, if to replace depth
Algoritic module is practised, then the former computing module that will be replaced is needed to dismantle, it then will be for replacing the target of former computing module
Computing module installs, and therefore, the prior art there is technical issues that the change of deep learning algorithm model.
In order to facilitate the deep learning algorithm model in change memory module, memory module can be erasable storage medium
Manufactured memory module.
For example, the current depth for being used to carry out person recognition in image that user can will store in memory module learns to calculate
Method model modification be the model parameter in another deep learning algorithm model or change current depth learning algorithm model with
It carries out the identification of vehicle in image or carries out the functions such as identification of plant in video.
Using the above embodiment of the present invention, the deep learning algorithm model in change memory module can be facilitated.
In a kind of specific embodiment of the embodiment of the present invention, the Data Input Interface includes：First data input
Interface and the second Data Input Interface, wherein, first Data Input Interface, for receiving non-picture data；Described second
Data Input Interface, for receiving image data；
The data output interface includes the first data output interface and the second data output interface, wherein, described first
Data output interface, for exporting non-picture data；Second data output interface, for exporting image data.
Illustratively, the first Data Input Interface can be utilized to receive non-picture data, for example, text data, code number
According to, director data etc.；Image data is received using the second Data Input Interface, for example, image data, video data etc..It is similar
, the first data output interface can export corresponding identification information of text data, recognition result etc.；Second data output interface
Image data can be exported, for example, image data, video data etc..As shown in Figure 1, the first Data Input Interface can be claimed
For standard input interface, the second Data Input Interface can be referred to as standard picture input interface；First data output interface can
To be referred to as standard output interface, the second data output interface can be referred to as standard picture output interface.
Using the above embodiment of the present invention, using different Data Input Interface input image datas and non-picture data,
Occupied bandwidth can be transmitted to avoid non-picture data, so as to improve the input speed of image data；Similar, it can also carry
The output speed of hi-vision data.
In a kind of specific embodiment of the embodiment of the present invention, first Data Input Interface, second data
Input interface, first data output interface and second data output interface are standard data interface.
Specifically, standard data interface includes but are not limited to：USB interface, DVI interface, HDMI interface and DP interfaces.
In practical applications, the first Data Input Interface can be referred to as standard input interface, the second Data Input Interface
It can be referred to as standard picture input interface；Similar, the first data output interface can be referred to as standard output interface, and second
Data output interface can be referred to as standard picture output interface.
Using the above embodiment of the present invention, using standard data interface the embodiment of the present invention and other equipment can be made mutual
It is mutually compatible, beneficial to the extensive use of the embodiment of the present invention.In addition, using existing standard data interface, data can also be improved
The reliability of transmission.
In a kind of specific embodiment of the embodiment of the present invention, the data output interface, in the interchange mode
In the case that the judging result of block is no, the data of the input are exported.
Illustratively, when the judging result of Switching Module is no, the data of the received input of output refer to remove and deposit
The deep learning algorithm model that stores can be identified in storage module data to be identified, for being stored in memory module
The model parameter for a little algorithm models of depth that deep learning algorithm model is configured, for storing to the depth in memory module
Spend learning algorithm model and for storing to other outside the model parameter of the deep learning algorithm model in memory module
Data.Data above is useless to the system, therefore need not store in the present system；In addition, data above may pair with
The other equipment of the system connection is useful, and the data of the input that when data above need not store are, it is necessary to by above-mentioned data sending
It goes out.
Using the above embodiment of the present invention, the data sending for the input that need not be stored can be gone out, be set by others
It is standby to be handled, the memory space of memory module can be occupied to avoid these data.
In a kind of specific embodiment of the embodiment of the present invention, the control module is specifically used for：To the calculating mould
Block sends training directive；
The computing module, is specifically used for：The first deep learning algorithm model is obtained, instruction is obtained from the memory module
Practice data, and according to the training directive, the first deep learning algorithm model is trained using the training data；
Data to be identified are obtained from the memory module, and using trained first deep learning algorithm model to described to be identified
Processing is identified in data.
Using the above embodiment of the present invention, deep learning algorithm model can be trained.
In a kind of specific embodiment of the embodiment of the present invention, the first deep learning algorithm model is training in advance
The first deep learning algorithm model.
Using the above embodiment of the present invention, trained deep learning algorithm model can be applied to the present invention and implemented
Example.
In a kind of specific embodiment of the embodiment of the present invention, the computing module is specifically used for：
The first deep learning algorithm model and the first deep learning algorithm model parameter are received, utilizes first depth
Algorithm model parameter is practised to be configured the first deep learning algorithm model.
It, can be by changing deep learning algorithm model parameters revision deep learning algorithm using the above embodiment of the present invention
Model.
Corresponding with embodiment illustrated in fig. 1 of the present invention, the embodiment of the present invention additionally provides a kind of deep learning recognition methods
's.
Fig. 2 is the flow diagram of the first depth deep learning recognition methods provided in an embodiment of the present invention, such as Fig. 2 institutes
Show, the described method includes：
S201：The data of input are received, the data of the input include：Data to be identified, the first deep learning algorithm mould
Type and other equipment are directed to the recognition result of the data to be identified using the second deep learning algorithm model.
S202：According to the mark of the identification information of the Data Data to be identified and the first deep learning algorithm model
Information judges whether storage institute's data to be identified and the first deep learning algorithm model；If so, perform S203 steps.
S203：Store the first deep learning algorithm model and the data to be identified.
S204：The identification obtained for the data to be identified instructs.
S205：According to it is described identification instruct, using the first deep learning algorithm model to the data to be identified into
Row identifying processing, obtains recognition result.
S206：It exports the recognition result and the other equipment is directed to the recognition result of the data to be identified.
Using embodiment illustrated in fig. 2 of the present invention, data are carried out using deep learning algorithm model in miscellaneous equipment and identify it
Afterwards, data identification is carried out using another deep learning algorithm model, data how to be identified are repeatedly identified, can be to avoid single
The error that the identification of deep learning algorithm model is brought, so as to improve the recognition accuracy to data to be identified.
Fig. 3 is the flow diagram of second of depth deep learning recognition methods provided in an embodiment of the present invention, such as Fig. 3 institutes
Show, embodiment illustrated in fig. 3 of the present invention adds following steps on the basis of embodiment illustrated in fig. 2 of the present invention：
S207：Obtain the training directive for the first deep learning algorithm model.
S208：Training data is obtained according to the training directive, and using the training data to first depth
It practises algorithm model to be trained, recycles trained first deep learning algorithm model that place is identified to data to be identified
Reason.
Using embodiment illustrated in fig. 3 of the present invention, deep learning algorithm model can be trained, and utilized trained
Deep learning algorithm model carries out the identification of data to be identified.
Fig. 4 is the flow diagram of the third depth deep learning recognition methods provided in an embodiment of the present invention, such as Fig. 4 institutes
Show, embodiment illustrated in fig. 4 of the present invention is no in the judging result of S202 steps on the basis of embodiment illustrated in fig. 2 of the present invention
In the case of, add S209：Export the data to be identified.
Using embodiment illustrated in fig. 4 of the present invention, the data sending that need not be stored can be gone out.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
All any modification, equivalent and improvement made within refreshing and principle etc., should all be included in the protection scope of the present invention.
Claims (10)
1. a kind of deep learning system, which is characterized in that the system comprises：Data Input Interface, Switching Module, storage mould
Block, control module, computing module and data output interface, wherein,
The Data Input Interface, for receiving the data of input, the data of the input include：It is data to be identified, first deep
Spend the identification knot that learning algorithm model and other equipment are directed to the data to be identified using the second deep learning algorithm model
Fruit；
The Switching Module, for the identification information according to the Data Data to be identified and the first deep learning algorithm mould
The identification information of type judges whether to store institute's data to be identified and the first deep learning algorithm model to the storage
In module；
The memory module in the case of being in the judging result of the Switching Module, stores the data to be identified
With the first deep learning algorithm model；
The control module, for sending the identification instruction for the data to be identified to the computing module；
The computing module, for obtaining the first deep learning algorithm model and described to be identified according to the identification instruction
Data, and processing is identified to the data to be identified using the first deep learning algorithm model；
The data output interface is waited to know for exporting the recognition result of the computing module and the other equipment for described
The recognition result of other data.
2. a kind of deep learning system according to claim 1, which is characterized in that the Data Input Interface includes：The
One Data Input Interface and the second Data Input Interface, wherein, first Data Input Interface, for receiving non-image number
According to；Second Data Input Interface, for receiving image data；
The data output interface includes the first data output interface and the second data output interface, wherein, first data
Output interface, for exporting non-picture data；Second data output interface, for exporting image data.
3. a kind of deep learning system according to claim 2, which is characterized in that first Data Input Interface, institute
It is that normal data connects to state the second Data Input Interface, first data output interface and second data output interface
Mouthful.
4. a kind of deep learning system according to claim 1, which is characterized in that the data output interface, for
In the case that the judging result of the Switching Module is no, the data of the input are exported.
5. a kind of deep learning system according to claim 1, which is characterized in that the control module is specifically used for：To
The computing module sends training directive；
The computing module, is specifically used for：The first deep learning algorithm model is obtained, training number is obtained from the memory module
According to, and according to the training directive, the first deep learning algorithm model is trained using the training data；From institute
It states and data to be identified is obtained in memory module, and using trained first deep learning algorithm model to the data to be identified
Processing is identified.
A kind of 6. deep learning system according to claim 1, which is characterized in that the first deep learning algorithm model
For the first deep learning algorithm model of training in advance.
7. a kind of deep learning system according to claim 1, which is characterized in that the computing module is specifically used for：
The first deep learning algorithm model and the first deep learning algorithm model parameter are received, is calculated using first deep learning
Method model parameter is configured the first deep learning algorithm model.
8. a kind of use a kind of deep learning recognition methods of claim 1-7 any one of them, which is characterized in that the method
Including：
The data of input are received, the data of the input include：Data to be identified, the first deep learning algorithm model and other
Equipment is directed to the recognition result of the data to be identified using the second deep learning algorithm model；
According to the identification information of the Data Data to be identified and the identification information of the first deep learning algorithm model, judge
Whether institute to be identified data and the first deep learning algorithm model are stored；
In the case where judging whether to store the judging result of the data to be identified to be, store first deep learning and calculate
Method model and the data to be identified；
The identification obtained for the data to be identified instructs；
It is instructed according to the identification, place is identified to the data to be identified using the first deep learning algorithm model
Reason, obtains recognition result；
It exports the recognition result and the other equipment is directed to the recognition result of the data to be identified.
9. a kind of deep learning recognition methods according to claim 8, which is characterized in that the method further includes：
Obtain the training directive for the first deep learning algorithm model；
Training data is obtained according to the training directive, and using the training data to the first deep learning algorithm model
It is trained, recycles trained first deep learning algorithm model that processing is identified to data to be identified.
10. a kind of deep learning recognition methods according to claim 8, which is characterized in that judge whether to store described
In the case that the judging result of the data to be identified is no, the method further includes：
Export the data to be identified.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810002406.1A CN108053031A (en) | 2018-01-02 | 2018-01-02 | A kind of deep learning system and deep learning recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810002406.1A CN108053031A (en) | 2018-01-02 | 2018-01-02 | A kind of deep learning system and deep learning recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108053031A true CN108053031A (en) | 2018-05-18 |
Family
ID=62126302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810002406.1A Pending CN108053031A (en) | 2018-01-02 | 2018-01-02 | A kind of deep learning system and deep learning recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108053031A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109492241A (en) * | 2018-08-10 | 2019-03-19 | 北京中科寒武纪科技有限公司 | Conversion method, device, computer equipment and storage medium |
| US11836616B2 (en) | 2018-12-04 | 2023-12-05 | Jinan University | Auditable privacy protection deep learning platform construction method based on block chain incentive mechanism |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106934364A (en) * | 2017-03-09 | 2017-07-07 | 腾讯科技（上海）有限公司 | The recognition methods of face picture and device |
| CN107305636A (en) * | 2016-04-22 | 2017-10-31 | 株式会社日立制作所 | Target identification method, Target Identification Unit, terminal device and target identification system |
-
2018
- 2018-01-02 CN CN201810002406.1A patent/CN108053031A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107305636A (en) * | 2016-04-22 | 2017-10-31 | 株式会社日立制作所 | Target identification method, Target Identification Unit, terminal device and target identification system |
| CN106934364A (en) * | 2017-03-09 | 2017-07-07 | 腾讯科技（上海）有限公司 | The recognition methods of face picture and device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109492241A (en) * | 2018-08-10 | 2019-03-19 | 北京中科寒武纪科技有限公司 | Conversion method, device, computer equipment and storage medium |
| US11836616B2 (en) | 2018-12-04 | 2023-12-05 | Jinan University | Auditable privacy protection deep learning platform construction method based on block chain incentive mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Goodfellow | On distinguishability criteria for estimating generative models | |
| US11983245B2 (en) | Unmanned driving behavior decision-making and model training | |
| CN108038459A (en) | A kind of detection recognition method of aquatic organism, terminal device and storage medium | |
| CN109241903A (en) | Sample data cleaning method, device, computer equipment and storage medium | |
| CN107566781A (en) | Video frequency monitoring method and video monitoring equipment | |
| CN108205708A (en) | A kind of novel expansible deep learning system and data identification method | |
| CN112183166B (en) | Method and device for determining training samples and electronic equipment | |
| CN106454108B (en) | Track up method, apparatus and electronic equipment based on artificial intelligence | |
| CN110472494A (en) | Face feature extracts model training method, facial feature extraction method, device, equipment and storage medium | |
| CN109697458A (en) | Control equipment mobile method, apparatus, storage medium and electronic equipment | |
| CN111467074B (en) | Method and device for detecting livestock status | |
| CN108053031A (en) | A kind of deep learning system and deep learning recognition methods | |
| CN110175505A (en) | Determination method, apparatus, storage medium and the electronic device of micro- expression type | |
| CN112686075A (en) | Behavior detection method, device and system | |
| CN109408649A (en) | Slag piece intelligent analysis method and system based on textural characteristics and machine learning | |
| CN113240430A (en) | Mobile payment verification method and device | |
| CN115213890A (en) | Grabbing control method and device, server, electronic equipment and storage medium | |
| CN113569627A (en) | Human body posture prediction model training method, human body posture prediction method and device | |
| CN114385846A (en) | Image classification method, electronic device, storage medium and program product | |
| CN114120089A (en) | Fire fighting equipment auxiliary installation method and system, electronic equipment and storage medium | |
| CN109241893A (en) | Road selection method, device and readable storage medium storing program for executing based on artificial intelligence technology | |
| CN117437467A (en) | Model training method and device, electronic equipment and storage medium | |
| CN110414322A (en) | Extract method, apparatus, equipment and the storage medium of picture | |
| CN108848117A (en) | A kind of cultivation surroundings monitoring system and its working method | |
| US12058441B2 (en) | Information processing apparatus with automatic adjustment of focal point |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180518 |