CN113849314A - Data processing model deployment method and device - Google Patents
Data processing model deployment method and device Download PDFInfo
- Publication number
- CN113849314A CN113849314A CN202111163897.6A CN202111163897A CN113849314A CN 113849314 A CN113849314 A CN 113849314A CN 202111163897 A CN202111163897 A CN 202111163897A CN 113849314 A CN113849314 A CN 113849314A
- Authority
- CN
- China
- Prior art keywords
- data processing
- internet
- network structure
- model
- things equipment
- 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
- 238000012545 processing Methods 0.000 title claims abstract description 395
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000012360 testing method Methods 0.000 claims description 34
- 230000006870 function Effects 0.000 claims description 30
- 230000008569 process Effects 0.000 claims description 18
- 230000000007 visual effect Effects 0.000 claims description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 5
- 238000003672 processing method Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 238000000586 desensitisation Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5027—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals
- G06F9/505—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resource being a machine, e.g. CPUs, Servers, Terminals considering the load
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
- G06F21/31—User authentication
- G06F21/32—User authentication using biometric data, e.g. fingerprints, iris scans or voiceprints
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
- G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- 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)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Security & Cryptography (AREA)
- Computer Hardware Design (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Bioethics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
One or more embodiments of the present specification provide a data processing model deployment method and apparatus, where the method is applied to a server, and may include: searching a target network structure matched with processing resources contained in the Internet of things equipment in a network structure search space based on the processing resource limiting conditions of the Internet of things equipment; and training a data processing model based on the target network structure, deploying the trained data processing model to the Internet of things equipment, and processing the detected data by the Internet of things equipment based on the data processing model.
Description
Technical Field
One or more embodiments of the present disclosure relate to the field of terminal technologies, and in particular, to a data processing model deployment method and apparatus.
Background
With the continuous development of the neural network technology, the internet of things equipment also introduces a data processing model obtained through neural network training for processing the obtained data. For example, in a biometric scene, the internet of things device may deploy a biometric model obtained through neural network training locally, and process biometric information based on the biometric model to obtain a corresponding recognition result when acquiring biometric information of a recognition object.
In the related art, a data processing model matched with the functions of the internet of things equipment is generally deployed based on the functions of the internet of things equipment so as to meet the requirement of processing the acquired data.
Disclosure of Invention
In view of this, one or more embodiments of the present disclosure provide a method and an apparatus for deploying a data processing model.
To achieve the above object, one or more embodiments of the present disclosure provide the following technical solutions:
according to a first aspect of one or more embodiments of the present specification, there is provided a data processing model deployment method, applied to a server, including:
searching a target network structure matched with processing resources contained in the Internet of things equipment in a network structure search space based on the processing resource limiting conditions of the Internet of things equipment;
and training a data processing model based on the target network structure, deploying the trained data processing model to the Internet of things equipment, and processing the detected data by the Internet of things equipment based on the data processing model.
According to a second aspect of one or more embodiments of the present specification, there is provided a data processing method applied to an internet of things device, including:
receiving and deploying a data processing model sent by a server; the data processing model is obtained by the server through training based on a target network structure, and the target network structure is obtained by the server through searching from a network structure searching space according to the processing resource limiting condition of the Internet of things equipment and is matched with the processing resource contained in the Internet of things equipment;
and inputting the detected data into the data processing model for processing, and outputting a corresponding processing result.
According to a third aspect of one or more embodiments of the present specification, there is provided a data processing model deployment apparatus, applied to a server, including:
the searching unit is used for searching a target network structure matched with the processing resource contained in the Internet of things equipment in a network structure searching space based on the processing resource limiting condition of the Internet of things equipment;
and the deployment unit is used for training a data processing model based on the target network structure and deploying the trained data processing model to the Internet of things equipment so as to process the detected data by the Internet of things equipment based on the data processing model.
According to a fourth aspect of one or more embodiments of the present specification, there is provided a data processing apparatus applied to an internet of things device, including:
the deployment unit is used for receiving and deploying the data processing model sent by the server; the data processing model is obtained by the server through training based on a target network structure, and the target network structure is obtained by the server through searching from a network structure searching space according to the processing resource limiting condition of the Internet of things equipment and is matched with the processing resource contained in the Internet of things equipment;
and the processing unit is used for inputting the detected data into the data processing model for processing and outputting a corresponding processing result.
According to a fifth aspect of one or more embodiments herein, there is provided an electronic device, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor implements the method according to the first aspect or the second aspect by executing the executable instructions.
According to a sixth aspect of one or more embodiments of the present specification, there is provided a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the method according to the first or second aspect.
Drawings
FIG. 1 is a flowchart of a data processing model deployment method provided by an exemplary embodiment.
Fig. 2 is a flowchart of a data processing method according to an exemplary embodiment.
Fig. 3 is a flowchart of a privacy preserving model deployment method according to an exemplary embodiment.
Fig. 4 is a flowchart of a privacy protecting method according to an exemplary embodiment.
Fig. 5 is a schematic structural diagram of an apparatus according to an exemplary embodiment.
FIG. 6 is a block diagram of a data processing model deployment apparatus provided in an exemplary embodiment.
Fig. 7 is a block diagram of a data processing apparatus according to an example embodiment.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with one or more embodiments of the present specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of one or more embodiments of the specification, as detailed in the claims which follow.
It should be noted that: in other embodiments, the steps of the corresponding methods are not necessarily performed in the order shown and described herein. In some other embodiments, the method may include more or fewer steps than those described herein. Moreover, a single step described in this specification may be broken down into multiple steps for description in other embodiments; multiple steps described in this specification may be combined into a single step in other embodiments.
The internet of things equipment can process the acquired data through the deployed data processing model. In practical application, a server usually performs training of a data processing model based on a neural network, and after the training is completed, the data processing model is sent to the internet of things device so as to be deployed by the internet of things device.
In the related art, when a data processing model is deployed in the internet of things equipment, the data processing model is deployed only according to the type, the function and the like of the internet of things equipment. For example, when a certain internet of things device is used for biometric identification, a biometric identification model is deployed in the internet of things device; when a certain internet of things device is used for data statistics, a data statistics model is deployed in the internet of things device.
However, in practical applications, the load capacity or the processing resources included in different internet of things devices are different. If the method in the related art is adopted, the deployed data processing model is probably not adapted to the processing resources contained in the internet of things equipment, and the problems of low data processing efficiency or high load are caused. For example, when processing resources required by the deployed data processing model during running exceed processing resources included in the internet of things equipment, the situation that the data processing model cannot run is likely to occur; when the processing resources required by the deployed data processing model during operation are much smaller than the processing resources included in the physical network device, the data processing efficiency may be low due to the poor performance of the data processing model.
In view of this, the present specification provides a method for deploying a data processing model, which can deploy an adaptive data processing model according to processing resources included in an internet of things device, and further fully utilize the processing resources included in the internet of things device, so as to improve data processing efficiency.
Fig. 1 is a flowchart of a data processing model deployment method according to an exemplary embodiment of the present specification. The method is applied to a server, and as shown in fig. 1, the method may include:
As can be seen from the above, in the related art, there are situations that the data processing efficiency is low or the data processing model cannot operate, and the data processing model is deployed only according to the type and function of the internet of things device without taking processing resources included in the internet of things device into consideration when the data processing model is deployed for the internet of things device.
In view of this, the data processing model is no longer deployed according to the type and function of the internet of things device, but a data processing model adapted to the processing resource included in the internet of things device is deployed for the internet of things device by taking the processing resource included in the internet of things device into consideration, so as to fully utilize the processing resource included in the internet of things device and improve the data processing efficiency as much as possible.
Specifically, the server may obtain a processing resource restriction condition corresponding to the internet of things device, and search a target network structure adapted to the processing resource included in the internet of things device in a network structure search space according to the processing resource restriction condition, on the basis, the data processing model may be trained based on the searched target network structure. After the training of the data processing model is completed, the obtained data processing model can be sent to the Internet of things equipment, the received data processing model is deployed locally by the Internet of things equipment, and after the deployment is completed, the Internet of things equipment can process data through the data processing model when detecting the data.
It should be appreciated that the processing resources occupied by the data model at runtime are generally determined by what network architecture it is trained on. Therefore, on the basis of acquiring the processing resource limiting condition of the internet of things device, the server in the specification searches for a target network structure matched with the processing resource contained in the internet of things device, and trains a data processing model based on the searched network structure, so that the specification can be obviously deployed to the data processing model in the internet of things device, and the server can be applied to the internet of things device, and the problem that the processing resource contained in the internet of things device is not fully utilized to cause low data processing efficiency is solved.
The processing resource limitation condition in this specification may be regarded as information of a processing resource included in the corresponding internet of things device, and may be acquired in various ways. For example, the internet of things device may obtain hardware information of a hardware structure included in the internet of things device, calculate to obtain processing resources included in the internet of things device based on the obtained hardware information, and further upload information of the processing resources obtained by calculation to the server; for another example, the server may preferentially obtain a model of the internet of things device, and determine information of processing resources included in the internet of things device based on the model. Of course, the above manner of acquiring the processing resource included in the internet of things device is only illustrative, and how to acquire the processing resource limitation condition may be determined by those skilled in the art according to actual needs, which is not limited in this specification.
And 104, training a data processing model based on the target network structure, deploying the trained data processing model to the Internet of things equipment, and processing the detected data by the Internet of things equipment based on the data processing model.
In practical application, processing resources contained in the internet of things equipment cannot be used for running the data processing model in some cases. Therefore, when the server in the specification trains the data processing model for the internet of things device, the server can train a plurality of data processing models based on the target network structure, and the processing resources occupied by the plurality of data processing models during operation are different and do not exceed the processing resources contained in the internet of things device, so that the internet of things device can perform data processing by calling different data processing models in different operation states. Specifically, after the plurality of data processing models are deployed to the internet of things device, if data to be processed is detected, the internet of things device can preferentially acquire the occupation status of the processing resources of the internet of things device, and the data processing models adapted to the occupation status are selected for data processing.
In the present specification, when training a data processing model for processing an image including biometric information, a server may input a sample image including biometric information and a noise image into a target network structure obtained by a search, and may converge a visual difference loss function corresponding to the target network structure by adjusting the target network structure. The visual difference loss function is used to constrain the visual difference between the output processed image and the input noisy image, whereas during the training process, when the visual difference loss function achieves convergence, the visual difference between the processed image output by the target network structure and the input noisy image is within a preset range.
It should be appreciated that the manner in which the noisy image is introduced during the training process and the visual disparity loss function described above is constructed so that it converges is essentially by hiding the characteristic information of the sample image into the noisy image, which achieves desensitization of the sample image.
Furthermore, a characteristic information loss function corresponding to the target network structure can be constructed, so that the characteristic information loss function is converged in a mode of adjusting the target network structure; when the characteristic information loss function is converged, the output processed image and the input characteristic information of the sample image are in a preset range, and therefore the characteristic information of the sample image is still kept after desensitization.
When the data processing model is actually trained, whether the adjusted target network structure can be used as the trained data processing model can be judged.
In one case, after the visual difference loss function corresponding to the target network structure converges, the output processed image and the input sample image may be respectively subjected to biometric identification based on a biometric identification algorithm, and the identification results obtained by the two biometric identification operations are compared, so that when the two images are consistent, the adjusted target network structure is determined as the trained data processing model.
In another case, after the visual difference loss function corresponding to the target network structure converges, the obtained processed image and the sample image corresponding to the processed image may be input into a preset countermeasure network to obtain a corresponding countermeasure result, and on this basis, it may be further determined whether the adjusted target network structure meets a model standard of a preset data processing model according to the countermeasure result, where when the countermeasure result indicates that the adjusted target network structure meets the preset data processing model, the adjusted target network structure may be determined as the trained data processing model.
In this specification, whether the trained data processing model meets a preset model standard or not may also be determined by a real-machine test, and repeated training may be performed if the trained data processing model does not meet the model standard. Specifically, after the data processing model is obtained based on the training of the target network structure, the data processing model may be transmitted to the sample internet of things device, so as to test the received data processing model through the sample internet of things device. In practical application, the sample internet of things equipment can input a test image into a received data processing model, and compare the output processed test image with the input test image to obtain a test result. And returning a test result obtained in the sample Internet of things equipment to the server, and judging whether the trained data processing model meets a preset model standard or not by the server based on the test result, wherein when the test result shows that the trained data processing model does not meet the preset model standard, the server can train the processing model based on the target network structure again, perform model test on the trained data processing model again, judge whether the trained data processing model meets the preset model standard or not based on the test result, and so on, and iteratively perform operation of training the data processing model based on the target network structure and operation of performing model test through the sample Internet of things equipment until the trained data processing model meets the preset model standard.
By the method, the precision of the data processing model obtained based on the target network structure training can be ensured, and the problem of poor data processing effect caused by insufficient precision of the data processing model is avoided.
It should be noted that, the sample internet of things device for model testing may be: and the Internet of things equipment has the same model as the Internet of things equipment needing to be deployed with the data processing model. In addition, the technical solution of the present specification can be applied to various application scenarios. For example, the method can be applied to a privacy protection scene, the data processing model obtained through the training in the above manner is the privacy protection model, and on the basis, after the privacy protection model is deployed to the internet of things equipment, the internet of things equipment can input the data to be protected into the privacy protection model under the condition that the data to be protected is obtained, so that the data to be protected is converted into corresponding privacy protection data through the privacy protection model, and further the leakage of the data to be protected is avoided.
According to the technical scheme, the server in the specification can preferentially determine the processing resources contained in the equipment of the internet of things, search the target network structure matched with the processing resources contained in the network structure search space, and train a data processing model for the equipment of the internet of things on the basis of the target network structure obtained through searching.
It should be understood that the server in the present specification is equivalent to training a proprietary data processing model adapted to the processing resources contained in each internet of things device. Compared with the mode of training the data processing model only based on the type and the function of the internet of things equipment in the related art, the data processing model is taken into consideration in the specification, and the problem that the data processing efficiency cannot be improved by fully utilizing the processing resources contained in the internet of things equipment on the basis of ensuring the normal operation of the data processing model due to the fact that the data processing model deployed in the internet of things equipment is not matched with the processing resources contained in the data processing model is solved.
Further, the server in this specification may train a plurality of data processing models based on a target network structure, and ensure that processing resources occupied by the plurality of data processing models are different when the plurality of data processing models operate. Based on this, after the plurality of data processing models are deployed to the internet of things device, the internet of things device may select the data processing model adapted to the current operating condition for data processing after receiving the data to be processed. In other words, by deploying a plurality of data processing models, the internet of things device in the present specification can select different data processing models for data processing at different times according to its own operating conditions, so that under different operating scenarios, processing resources that can be called at present can be fully utilized to improve data processing efficiency on the basis of ensuring normal operation of the data processing models.
The specification further provides a data processing method applied to the internet of things equipment, and in the method, the internet of things equipment can receive the data processing model obtained by the training of the server and deploy the data processing model locally for processing the detected data. In this method, most operation modes, for example, how to train the data processing model, how to process the detected data, and the like, can refer to the description of the previous embodiment, and are not described in detail below.
Fig. 2 is a flowchart of a data processing method according to an exemplary embodiment of the present disclosure. The method is applied to the internet of things equipment, and as shown in fig. 2, the method may include:
As described above, the internet of things device may obtain hardware information of a hardware structure included in the internet of things device, calculate processing resources included in the internet of things device based on the obtained hardware information, and then send information of the processing resources included in the internet of things device to the server as a processing resource restriction condition, so that the server searches for a target network structure adapted to the processing resources included in the internet of things device in a network structure space, and trains a data processing model based on the target network structure. Of course, the processing resources included in the internet of things device may also be obtained by the server based on the device model of the internet of things device.
As described above, in practical applications, the processing resources included in the internet of things device cannot be used for the operation of the data processing model in some cases. Therefore, when the server in the specification trains the data processing model for the internet of things device, the server can train a plurality of data processing models based on the target network structure, and the processing resources occupied by the plurality of data processing models during operation are different and do not exceed the processing resources contained in the internet of things device, so that the internet of things device can perform data processing by calling different data processing models in different operation states. Specifically, after the plurality of data processing models are deployed to the internet of things device, if data to be processed is detected, the internet of things device can preferentially acquire the occupation status of the processing resources of the internet of things device, and the data processing models adapted to the occupation status are selected for data processing.
As described above, when the server trains the data processing model for processing the image including the biometric information, the server may input the sample image including the biometric information and the noise image into the target network structure obtained by the search, converge the visual difference loss function corresponding to the target network structure by adjusting the target network structure, and use the adjusted target network structure as the trained data processing model.
As described above, when the data processing model is actually trained, it can be further determined whether the adjusted target network structure can be used as the trained data processing model. In one case, after the visual difference loss function corresponding to the target network structure converges, performing biometric identification on the output processed image and the input sample image respectively based on a biometric identification algorithm, and comparing the identification results obtained by the two biometric identification operations, so that when the two are consistent, the adjusted target network structure is determined as the trained data processing model; in another case, after the visual difference loss function corresponding to the target network structure converges, the obtained processed image and the sample image corresponding to the processed image may be input into a preset countermeasure network to obtain a corresponding countermeasure result, and on this basis, it may be further determined whether the adjusted target network structure meets a model standard of a preset data processing model according to the countermeasure result, where when the countermeasure result indicates that the adjusted target network structure meets the preset data processing model, the adjusted target network structure may be determined as the trained data processing model.
As described above, whether the trained data processing model meets the preset model standard or not can be determined by a real-machine test, and repeated training is performed if the trained data processing model does not meet the model standard. Specifically, after the data processing model is obtained based on the training of the target network structure, the data processing model may be transmitted to the sample internet of things device, so as to test the received data processing model through the sample internet of things device. In practical application, the sample internet of things equipment can input a test image into a received data processing model, and compare the output processed test image with the input test image to obtain a test result. And returning a test result obtained in the sample Internet of things equipment to the server, and judging whether the trained data processing model meets a preset model standard or not by the server based on the test result, wherein when the test result shows that the trained data processing model does not meet the preset model standard, the server can train the processing model based on the target network structure again, perform model test on the trained data processing model again, judge whether the trained data processing model meets the preset model standard or not based on the test result, and so on, and iteratively perform operation of training the data processing model based on the target network structure and operation of performing model test through the sample Internet of things equipment until the trained data processing model meets the preset model standard.
And 204, inputting the detected data into the data processing model for processing, and outputting a corresponding processing result.
As described above, the sample internet of things device for model testing may be: and the type of the Internet of things equipment is the same as that of the sample Internet of things equipment needing to be deployed with the data processing model. In addition, the technical solution of the present specification can be applied to various application scenarios. For example, the method can be applied to a privacy protection scene, the data processing model obtained through the training in the above manner is the privacy protection model, and on the basis, after the privacy protection model is deployed to the internet of things equipment, the internet of things equipment can input the data to be protected into the privacy protection model under the condition that the data to be protected is obtained, so that the data to be protected is converted into corresponding privacy protection data through the privacy protection model, and further the leakage of the data to be protected is avoided.
According to the technical scheme, the server can search the target network structure matched with the processing resources contained in the internet of things equipment according to the processing resources contained in the internet of things equipment, and train the data processing model based on the searched target network structure, so that the data processing model deployed by the internet of things equipment is matched with the processing resources contained in the internet of things equipment. On the basis, the problem that in the related technology, due to the fact that only the type and the function of the internet of things equipment are considered, the deployed data processing model is not matched with the processing resources contained in the data processing model, and therefore the data processing model cannot run normally or the processing resources contained in the data processing model cannot be fully utilized to improve the data processing efficiency can be solved.
In the following, the technical solution of the present specification is described by taking a privacy protection scenario as an example.
Fig. 3 is a flowchart illustrating a privacy protection model deployment method according to an exemplary embodiment of the present disclosure, applied to a server, and as shown in fig. 3, the method may include the following steps:
In this embodiment, the target internet of things device may obtain hardware information of its hardware structure, and calculate processing resources contained in itself based on the hardware information, and then may send the calculated information of the processing resources to the server, so that the server searches for the target network structure based on the information.
In the field of neural networks, the process of searching for a target network structure is to change the hyper-parameters in the basic network structure practically continuously until the hyper-parameters are adjusted to match the search conditions. In this embodiment, the information of the processing resources may be used as a condition for adjusting the hyper-parameter, so that the adjusted target network structure is adapted to the processing resources included in the target internet of things device.
In this embodiment, after the target network structure is obtained, the privacy preserving model may be trained based on the target network structure. In the actual training process, the purpose of privacy protection can be achieved by adopting a mode of hiding image features in a noise image. Therefore, in the actual training process, the sample biometric image and the noise image can be used as model input for training.
It should be understood that, in the training process, a loss function corresponding to the target network structure may be constructed, and the requirements of the finally trained privacy protection model are constrained by the loss function. Specifically, the target network structure may include three networks, one of which is: an encoding network G, the input of which is a sample biometric image I and a noise image I', and the output of which is a privacy-preserving image O; the second is as follows: a feature extraction network F for inputting the privacy-preserving image O and the sample biological feature image I and outputting the biological features F of the privacy-preserving image O and the sample biological feature image IOAnd fI(ii) a The third is: the countermeasure network D, the inputs of which are the noise image I' and the privacy-preserving image O.
Accordingly, the global loss function corresponding to the target network structure may be:
Ltotal＝Lmse(I’,O)+Lmse(fO,fI)+lce((I’,O),(yI′,yO))
wherein L ismse(I', O) corresponds to the coding network G, constrained by: the output privacy protection image O and the noise image I' are close to each other in visual effect, and the privacy information of the sample biological characteristic image I cannot be revealed;
Lmse(fO,fI) Corresponding to feature extractionNetwork F, constrained by: ensuring that the output privacy protection image O and the sample biological characteristic image I can be successfully identified by a biological identification algorithm and the identification results are consistent;
Lce((I’,O),(yI′,yO) Corresponding to the countermeasure network D, constrained by: the output privacy-preserving image O is ensured to have no high-frequency modification trace, and the intensity and irreversibility of privacy preservation are enhanced. Wherein, yI′Class label (also called class parameter), y for the noise image IOA category label for the privacy preserving image O.
It should be understood that, when the global loss function converges, it is ensured that all the parts of the constraint are implemented, and therefore, when the global loss function converges, the adjusted target network structure is used as the trained privacy protection model.
And 305, taking the adjusted target network structure as a trained privacy protection model.
According to the technical scheme, when model training is carried out on the target Internet of things equipment for privacy protection, information of processing resources contained in the target Internet of things equipment can be preferentially acquired, and a target network structure matched with the processing resources contained in the target Internet of things equipment is searched based on the information. On the basis, the privacy protection model can be trained based on the searched target network structure, so that the trained privacy protection model is adapted to the target Internet of things equipment.
Still taking a privacy protection scenario as an example, on the basis of the previous embodiment, how to perform privacy protection on the biometric image by the internet of things device based on the deployed privacy protection model is introduced.
Fig. 4 is a flowchart illustrating a privacy protection method according to an exemplary embodiment of the present disclosure, applied to an internet of things device, and as shown in fig. 4, the method may include the following steps:
In this embodiment, a plurality of privacy protection models may be trained based on a target network structure in the manner of the previous embodiment, and all the privacy protection models are deployed to target internet of things devices. The occupied processing resources are different when the privacy protection models operate, so that the target Internet of things equipment processes the biological characteristic images through different privacy protection models in different operation states, and privacy protection is further achieved.
On the basis, after the target internet of things equipment acquires the biological characteristic image, the target internet of things equipment can preferentially acquire the operation data of the target internet of things equipment so as to determine the processing resources which can be called currently according to the operation data.
In step 403, the processing resources that can be invoked are determined based on the operational data.
After the processing resources which can be called are determined, the target internet of things device can select a target privacy protection model which is matched with the determined processing resources from the multiple deployed privacy protection models so as to convert the biological feature images into privacy protection images.
It should be noted that "adapted" in this embodiment means: the processing resources occupied by the target privacy protection model during operation do not exceed the processing resources which can be called currently, but are as close as possible to the processing resources which can be called, so that the processing resources which can be called can be utilized to the maximum extent to improve the efficiency of data processing.
For example, if the processing resources included in the target internet of things device are divided into 10 equal parts, and the current operation data indicates that 2 parts of the processing resources are occupied by other services, the processing resources that can be called currently are only 8 parts. If 3 privacy protection models are deployed in the target Internet of things equipment, 9 processing resources are occupied by the model A in running, 7 processing resources are occupied by the model B in running, and 5 processing resources are occupied by the model C in running. In this case, the model B may be determined as the target privacy protection model, and the biometric image may be input into the model B to output the privacy protection image through the model B.
After the privacy-preserving image is obtained based on the target privacy-preserving model, the privacy-preserving image can be stored and the like.
According to the technical scheme, the server can train a plurality of privacy protection models for the target Internet of things equipment based on the target network structure, and the processing resources occupied by the plurality of privacy protection models during operation are different. Based on this, after the plurality of privacy protection models are deployed to the target internet of things device, the target internet of things device can select the target privacy protection model matched with the self running state to perform data processing according to the self running state when acquiring the data to be protected, so that the target internet of things device can fully utilize the processing resources which can be called under different running states, and the data processing efficiency is improved to the maximum extent.
FIG. 5 is a schematic block diagram of an apparatus provided in an exemplary embodiment. Referring to fig. 5, at the hardware level, the apparatus includes a processor 502, an internal bus 504, a network interface 506, a memory 508 and a non-volatile memory 510, but may also include hardware required for other services. One or more embodiments of the present description may be implemented in software, such as by processor 502 reading corresponding computer programs from non-volatile storage 510 into memory 508 and then running. Of course, besides software implementation, the one or more embodiments in this specification do not exclude other implementations, such as logic devices or combinations of software and hardware, and so on, that is, the execution subject of the following processing flow is not limited to each logic unit, and may also be hardware or logic devices.
Referring to fig. 6, the data processing model deployment apparatus may be applied to the device shown in fig. 5 to implement the technical solution of the present specification. Wherein, the data processing model deployment device may include:
a searching unit 601, configured to search, in a network structure search space, a target network structure adapted to a processing resource included in an internet of things device based on a processing resource restriction condition of the internet of things device;
a deployment unit 602, configured to train a data processing model based on the target network structure, and deploy the trained data processing model to the internet of things device, so that the internet of things device processes the detected data based on the data processing model.
Optionally, the deployment unit 602 is further configured to:
training a plurality of data processing models based on the target network structure; processing resources occupied by the data processing models in running are different and do not exceed the processing resources contained in the Internet of things equipment;
and deploying the plurality of data processing models to the Internet of things equipment, and selecting the data processing models adaptive to the occupation states for data processing by the Internet of things equipment according to the occupation states of the processing resources of the Internet of things equipment.
Optionally, the deployment unit 602 is further configured to:
inputting a sample image containing biological characteristic information and a noise image into the target network structure, and adjusting the target network structure to make a visual difference loss function corresponding to the target network structure converge; wherein the visual difference of the processed image output by the target network structure and the noise image is within a certain range when the visual difference loss function achieves convergence;
and determining the adjusted target network structure as a data processing model obtained by training.
Optionally, the deployment unit 602 is further configured to:
respectively carrying out biological recognition on the processed image and the sample image based on a biological recognition algorithm;
and under the condition that the recognition results of the two biological recognition operations are consistent, determining the adjusted target network structure as a trained data processing model.
Optionally, the deployment unit 602 is further configured to:
inputting the sample image and the processed image into a preset countermeasure network to obtain a countermeasure result;
and determining the adjusted target network structure as the trained data processing model under the condition that the countermeasure result shows that the adjusted target network structure meets the preset model standard of the data processing model.
Optionally, the deployment unit 602 is further configured to:
training a data processing model based on the target network structure, and transmitting the trained data processing model to sample Internet of things equipment so as to test the data processing model through the sample Internet of things equipment;
and receiving a test result returned by the sample Internet of things equipment, and iteratively executing the operation of training the data processing model based on the target network structure and the operation of performing model test through the sample physical network equipment under the condition that the test result shows that the trained data processing model does not accord with the preset model standard until the trained data processing model accords with the model standard.
Referring to fig. 7, the data processing apparatus may be applied to the device shown in fig. 5 to implement the technical solution of the present specification. Wherein the data processing apparatus may include:
a deployment unit 701, which receives and deploys the data processing model sent by the server; the data processing model is obtained by the server through training based on a target network structure, and the target network structure is obtained by the server through searching from a network structure searching space according to the processing resource limiting condition of the Internet of things equipment and is matched with the processing resource contained in the Internet of things equipment;
the processing unit 702 inputs the detected data into the data processing model for processing, and outputs a corresponding processing result.
Optionally, the method further includes:
an obtaining unit 703 that obtains hardware information of a hardware structure included in the internet of things device, and calculates a processing resource included in the internet of things device based on the hardware information; and sending information of processing resources contained in the internet of things equipment to the server as the processing resource limiting condition.
Alternatively to this, the first and second parts may,
the deployment unit 701 is further configured to: receiving a plurality of data processing models sent by the server, and deploying the data processing models locally, wherein processing resources occupied by the data processing models during operation are different;
the processing unit 702 is further configured to: and under the condition that the data to be processed is obtained, determining the operating condition of the Internet of things equipment, selecting a target data processing model matched with the operating condition from the plurality of data processing models, inputting the data to be processed into the target data processing model for data processing, and outputting a corresponding processing result.
Optionally, the data processing model is a privacy protection model, and the processing unit 702 is further configured to:
and under the condition of acquiring the data to be protected, inputting the data to be protected into the privacy protection model so as to convert the data to be protected into corresponding privacy protection data through the privacy protection model.
The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. A typical implementation device is a computer, which may take the form of a personal computer, laptop computer, cellular telephone, camera phone, smart phone, personal digital assistant, media player, navigation device, email messaging device, game console, tablet computer, wearable device, or a combination of any of these devices.
In a typical configuration, a computer includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The terminology used in the description of the one or more embodiments is for the purpose of describing the particular embodiments only and is not intended to be limiting of the description of the one or more embodiments. As used in one or more embodiments of the present specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.
It should be understood that although the terms first, second, third, etc. may be used in one or more embodiments of the present description to describe various information, such information should not be limited to these terms. These terms are only used to distinguish one type of information from another. For example, first information may also be referred to as second information, and similarly, second information may also be referred to as first information, without departing from the scope of one or more embodiments herein. The word "if" as used herein may be interpreted as "at … …" or "when … …" or "in response to a determination", depending on the context.
The above description is only for the purpose of illustrating the preferred embodiments of the one or more embodiments of the present disclosure, and is not intended to limit the scope of the one or more embodiments of the present disclosure, and any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the one or more embodiments of the present disclosure should be included in the scope of the one or more embodiments of the present disclosure.
Claims (14)
1. A data processing model deployment method is applied to a server and comprises the following steps:
searching a target network structure matched with processing resources contained in the Internet of things equipment in a network structure search space based on the processing resource limiting conditions of the Internet of things equipment;
and training a data processing model based on the target network structure, deploying the trained data processing model to the Internet of things equipment, and processing the detected data by the Internet of things equipment based on the data processing model.
2. The method of claim 1, wherein the first and second light sources are selected from the group consisting of,
the training of the data processing model based on the target network structure includes: training a plurality of data processing models based on the target network structure; processing resources occupied by the data processing models in running are different and do not exceed the processing resources contained in the Internet of things equipment;
the deploying the trained data processing model to the internet of things equipment comprises: and deploying the plurality of data processing models to the Internet of things equipment, and selecting the data processing models adaptive to the occupation states for data processing by the Internet of things equipment according to the occupation states of the processing resources of the Internet of things equipment.
3. The method of claim 1, the training a data processing model based on the target network structure, comprising:
inputting a sample image containing biological characteristic information and a noise image into the target network structure, and adjusting the target network structure to make a visual difference loss function corresponding to the target network structure converge; wherein the visual difference of the processed image output by the target network structure and the noise image is within a certain range when the visual difference loss function achieves convergence;
and determining the adjusted target network structure as a data processing model obtained by training.
4. The method of claim 3, the determining the adjusted target network structure as a trained data processing model, comprising:
respectively carrying out biological recognition on the processed image and the sample image based on a biological recognition algorithm;
and under the condition that the recognition results of the two biological recognition operations are consistent, determining the adjusted target network structure as a trained data processing model.
5. The method of claim 3, the determining the adjusted target network structure as a trained data processing model, comprising:
inputting the sample image and the processed image into a preset countermeasure network to obtain a countermeasure result;
and determining the adjusted target network structure as the trained data processing model under the condition that the countermeasure result shows that the adjusted target network structure meets the preset model standard of the data processing model.
6. The method of claim 1, the training a data processing model based on the target network structure, comprising:
training a data processing model based on the target network structure, and transmitting the trained data processing model to sample Internet of things equipment so as to test the data processing model through the sample Internet of things equipment;
and receiving a test result returned by the sample Internet of things equipment, and iteratively executing the operation of training the data processing model based on the target network structure and the operation of performing model test through the sample physical network equipment under the condition that the test result shows that the trained data processing model does not accord with the preset model standard until the trained data processing model accords with the model standard.
7. A data processing method is applied to Internet of things equipment and comprises the following steps:
receiving and deploying a data processing model sent by a server; the data processing model is obtained by the server through training based on a target network structure, and the target network structure is obtained by the server through searching from a network structure searching space according to the processing resource limiting condition of the Internet of things equipment and is matched with the processing resource contained in the Internet of things equipment;
and inputting the detected data into the data processing model for processing, and outputting a corresponding processing result.
8. The method of claim 7, further comprising:
acquiring hardware information of a hardware structure contained in the Internet of things equipment, and calculating processing resources contained in the Internet of things equipment based on the hardware information;
and sending information of processing resources contained in the Internet of things equipment to the server as the processing resource limiting condition.
9. The method of claim 7, wherein the first and second light sources are selected from the group consisting of,
the receiving and deploying of the data processing model sent by the server comprises the following steps: receiving a plurality of data processing models sent by the server, and deploying the data processing models locally, wherein processing resources occupied by the data processing models during operation are different;
the inputting the detected data into the data processing model for processing and outputting the corresponding processing result includes: and under the condition that the data to be processed is obtained, determining the operating condition of the Internet of things equipment, selecting a target data processing model matched with the operating condition from the plurality of data processing models, inputting the data to be processed into the target data processing model for data processing, and outputting a corresponding processing result.
10. The method of claim 7, wherein the data processing model is a privacy protection model, and the inputting the detected data into the data processing model for processing and outputting the corresponding processing result comprises:
and under the condition of acquiring the data to be protected, inputting the data to be protected into the privacy protection model so as to convert the data to be protected into corresponding privacy protection data through the privacy protection model.
11. A data processing model deployment device is applied to a server and comprises:
the searching unit is used for searching a target network structure matched with the processing resource contained in the Internet of things equipment in a network structure searching space based on the processing resource limiting condition of the Internet of things equipment;
and the deployment unit is used for training a data processing model based on the target network structure and deploying the trained data processing model to the Internet of things equipment so as to process the detected data by the Internet of things equipment based on the data processing model.
12. A data processing device is applied to Internet of things equipment and comprises:
the deployment unit is used for receiving and deploying the data processing model sent by the server; the data processing model is obtained by the server through training based on a target network structure, and the target network structure is obtained by the server through searching from a network structure searching space according to the processing resource limiting condition of the Internet of things equipment and is matched with the processing resource contained in the Internet of things equipment;
and the processing unit is used for inputting the detected data into the data processing model for processing and outputting a corresponding processing result.
13. An electronic device, comprising:
a processor;
a memory for storing processor-executable instructions;
wherein the processor implements the method of any one of claims 1-10 by executing the executable instructions.
14. A computer readable storage medium having stored thereon computer instructions which, when executed by a processor, carry out the steps of the method according to any one of claims 1 to 10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111163897.6A CN113849314A (en) | 2021-09-30 | 2021-09-30 | Data processing model deployment method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111163897.6A CN113849314A (en) | 2021-09-30 | 2021-09-30 | Data processing model deployment method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113849314A true CN113849314A (en) | 2021-12-28 |
Family
ID=78977441
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111163897.6A Pending CN113849314A (en) | 2021-09-30 | 2021-09-30 | Data processing model deployment method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113849314A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115474000A (en) * | 2022-08-16 | 2022-12-13 | 支付宝(杭州)信息技术有限公司 | Data processing method and device |
| WO2023206097A1 (en) * | 2022-04-26 | 2023-11-02 | 西门子股份公司 | Ai model deployment method and apparatus, and electronic device and computer-readable medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130031035A1 (en) * | 2011-07-31 | 2013-01-31 | International Business Machines Corporation | Learning admission policy for optimizing quality of service of computing resources networks |
| US20160171379A1 (en) * | 2014-12-15 | 2016-06-16 | Intuit Inc. | System and method for deploying predictive models |
| CN110795722A (en) * | 2019-10-25 | 2020-02-14 | 支付宝(杭州)信息技术有限公司 | Deployment method and device of security authentication model and electronic equipment |
| CN111222637A (en) * | 2020-01-17 | 2020-06-02 | 上海商汤智能科技有限公司 | Neural network model deployment method and device, electronic equipment and storage medium |
| CN111310575A (en) * | 2020-01-17 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Face living body detection method, related device, equipment and storage medium |
| CN111783997A (en) * | 2020-06-29 | 2020-10-16 | 杭州海康威视数字技术股份有限公司 | Data processing method, device and equipment |
| WO2021114625A1 (en) * | 2020-05-28 | 2021-06-17 | 平安科技（深圳）有限公司 | Network structure construction method and apparatus for use in multi-task scenario |
| CN113435585A (en) * | 2021-07-15 | 2021-09-24 | 支付宝(杭州)信息技术有限公司 | Service processing method, device and equipment |
-
2021
- 2021-09-30 CN CN202111163897.6A patent/CN113849314A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130031035A1 (en) * | 2011-07-31 | 2013-01-31 | International Business Machines Corporation | Learning admission policy for optimizing quality of service of computing resources networks |
| US20160171379A1 (en) * | 2014-12-15 | 2016-06-16 | Intuit Inc. | System and method for deploying predictive models |
| CN110795722A (en) * | 2019-10-25 | 2020-02-14 | 支付宝(杭州)信息技术有限公司 | Deployment method and device of security authentication model and electronic equipment |
| CN111222637A (en) * | 2020-01-17 | 2020-06-02 | 上海商汤智能科技有限公司 | Neural network model deployment method and device, electronic equipment and storage medium |
| CN111310575A (en) * | 2020-01-17 | 2020-06-19 | 腾讯科技（深圳）有限公司 | Face living body detection method, related device, equipment and storage medium |
| WO2021114625A1 (en) * | 2020-05-28 | 2021-06-17 | 平安科技（深圳）有限公司 | Network structure construction method and apparatus for use in multi-task scenario |
| CN111783997A (en) * | 2020-06-29 | 2020-10-16 | 杭州海康威视数字技术股份有限公司 | Data processing method, device and equipment |
| CN113435585A (en) * | 2021-07-15 | 2021-09-24 | 支付宝(杭州)信息技术有限公司 | Service processing method, device and equipment |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023206097A1 (en) * | 2022-04-26 | 2023-11-02 | 西门子股份公司 | Ai model deployment method and apparatus, and electronic device and computer-readable medium |
| CN115474000A (en) * | 2022-08-16 | 2022-12-13 | 支付宝(杭州)信息技术有限公司 | Data processing method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109544166B (en) | Risk identification method and risk identification device | |
| CN109347787B (en) | Identity information identification method and device | |
| CN113849314A (en) | Data processing model deployment method and device | |
| CN109214421B (en) | Model training method and device and computer equipment | |
| US9761221B2 (en) | Order statistic techniques for neural networks | |
| CN108734304B (en) | Training method and device of data model and computer equipment | |
| US11593587B2 (en) | Object detection system and an object detection method | |
| CN108876847B (en) | Image positioning method, device, system and storage medium | |
| US20140279734A1 (en) | Performing Cross-Validation Using Non-Randomly Selected Cases | |
| US20180129914A1 (en) | Image recognition device and image recognition method | |
| CN111415015A (en) | Business model training method, device and system and electronic equipment | |
| CN111199157B (en) | Text data processing method and device | |
| CN111008620A (en) | Target user identification method and device, storage medium and electronic equipment | |
| CN112347512A (en) | Image processing method, device, equipment and storage medium | |
| CN114821823A (en) | Image processing, training of human face anti-counterfeiting model and living body detection method and device | |
| CN112116018A (en) | Sample classification method, apparatus, computer device, medium, and program product | |
| US11593686B2 (en) | Methods, systems and apparatus to improve deep learning resource efficiency | |
| CN116720214A (en) | Model training method and device for privacy protection | |
| CN114254588B (en) | Data tag processing method and device | |
| CN114818458A (en) | System parameter optimization method, device, computing equipment and medium | |
| CN112926678B (en) | Model similarity determination method and device | |
| CN110443746B (en) | Picture processing method and device based on generation countermeasure network and electronic equipment | |
| CN113963236A (en) | Target detection method and device | |
| CN113657317A (en) | Cargo position identification method and system, electronic equipment and storage medium | |
| CN112200711A (en) | Training method and system of watermark classification model |
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 |