CN113010435A - Method and device for screening algorithm model and test platform - Google Patents
Method and device for screening algorithm model and test platform Download PDFInfo
- Publication number
- CN113010435A CN113010435A CN202110379695.9A CN202110379695A CN113010435A CN 113010435 A CN113010435 A CN 113010435A CN 202110379695 A CN202110379695 A CN 202110379695A CN 113010435 A CN113010435 A CN 113010435A
- Authority
- CN
- China
- Prior art keywords
- test
- algorithm model
- model
- screening
- data set
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3684—Test management for test design, e.g. generating new test cases
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/36—Preventing errors by testing or debugging software
- G06F11/3668—Software testing
- G06F11/3672—Test management
- G06F11/3692—Test management for test results analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Quality & Reliability (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Testing Or Calibration Of Command Recording Devices (AREA)
Abstract
The invention discloses a screening method, a screening device and a testing platform of an algorithm model, and particularly relates to the technical field of detection. The method comprises the steps of obtaining a scene detail data set; setting test parameters of the algorithm model; executing a preset test flow based on the scene detail data set and the test parameters to obtain a test result of the algorithm model; and evaluating and screening the algorithm model according to the test result of the algorithm model. The method comprises the steps of obtaining scene subdivision data and setting parameters of an algorithm model, detecting the algorithm model according to a preset test flow, and evaluating and screening the tested algorithm model according to a detection result, so that the algorithm model with the optimal comprehensive performance is quickly screened out.
Description
Technical Field
The invention relates to the technical field of detection, in particular to a method and a device for screening an algorithm model and a test platform.
Background
In the prior art, after data are trained by using a deep learning or classification algorithm, a large number of data models are generated, and in order to be able to screen out the optimal model for production. The method used in the prior art is to execute a test script in a server environment, perform a recharge test on a test data set based on a GPU or CPU environment of the server, calculate relevant indexes of a model, comprehensively compare and select the model with the best index, perform format conversion and apply the model to a product, or execute the test script in a test equipment terminal, introduce a model file to be tested into equipment for conversion, simultaneously put test data into the equipment, perform the recharge test by adopting an actual running state close to the equipment, collect test results, manually collect the results and screen the model.
However, for the above method, because the operating environment of the data training is GPU or CPU, when in actual application, the model may be transplanted to a corresponding AI chip for operation, so that there is a quantization error between the chip and the host, and the effect of the quantization error is different from the effect of the operation on the host, which may cause the screened model to have a deviation; or when the test result is more consistent with the real running state of the equipment, the performance index reliability is higher, but the test flow and the environment are complex, manual energy observation needs to be kept, the test state cannot be mastered in time, the problem is found and repaired in time, and the next model test cannot be skipped to enter in time when the test index is obviously poor, so that time and computing resources are wasted.
Disclosure of Invention
In view of this, embodiments of the present invention provide a screening method, an apparatus, and a test platform for an algorithm model, so as to solve the problem that the existing screening method cannot quickly screen out an algorithm model with optimal comprehensive performance.
According to a first aspect, an embodiment of the present invention provides a method for screening an algorithm model, including:
acquiring a scene detail data set; setting test parameters of the algorithm model; executing a preset test flow based on the scene detail data set and the test parameters to obtain a test result of the algorithm model; and evaluating and screening the algorithm model according to the test result of the algorithm model.
According to the screening method of the algorithm model, provided by the embodiment of the invention, the scene subdivision data is obtained, the parameter setting is carried out on the algorithm model, the detection of the algorithm model is carried out according to the preset test flow, and then the tested algorithm model is evaluated and screened according to the detection result, so that the algorithm model with the optimal comprehensive performance is rapidly screened out.
With reference to the first aspect, in a first implementation manner of the first aspect, the acquiring the scene detail data set includes: acquiring a test data set; and carrying out scene subdivision on the basis of the test data set to obtain a scene subdivision data set.
With reference to the first implementation manner of the first aspect, in a second implementation manner of the first aspect, the performing scene segmentation based on the test data set to obtain a scene segmentation data set includes: carrying out weight distribution according to dimension information based on the test data set to obtain a dimension data set; and carrying out weight distribution on the data sets under the same dimensionality according to the dimensionality data sets to obtain subdata sets.
According to the screening method of the algorithm model, provided by the embodiment of the invention, the scene segmentation is carried out on the test data set, so that the subdivided and sorted data can be extracted quickly and accurately, the screening capability is improved, and the algorithm model with the optimal comprehensive performance is further screened out quickly.
With reference to the first aspect, in a third implementation manner of the first aspect, the setting test parameters of the algorithm model includes: setting test execution conditions, configuring running equipment, a model to be tested and data to be tested.
According to the method for screening the algorithm model, provided by the embodiment of the invention, the algorithm model to be tested and the equipment for operating the algorithm model are determined by setting the parameters of the algorithm model, so that the algorithm model with the optimal comprehensive performance can be screened out quickly.
With reference to the first aspect, in a fourth implementation manner of the first aspect, the executing a preset test flow based on the test data set and the test parameters further includes: performing data synchronization based on the test parameters to obtain synchronous data; extracting model data in the test parameters, and converting the model data to obtain an execution model; and performing data recharging test according to the synchronous data and the execution model, and outputting the test result of the algorithm model.
According to the method for screening the algorithm model, provided by the embodiment of the invention, the preset test flow is executed, so that the screening capacity of the screening method is improved, and the algorithm model with the optimal comprehensive performance is further rapidly screened.
With reference to the first aspect, in a fifth implementation manner of the first aspect, the performing evaluation screening on the algorithm model according to the test result of the algorithm model includes: obtaining at least 2 test results; and carrying out comparative analysis based on the test result to determine an optimal algorithm model.
According to the method for screening the algorithm model, provided by the embodiment of the invention, the algorithm model with the optimal comprehensive performance is screened out by comparing and analyzing the plurality of test results, so that the algorithm model with the optimal comprehensive performance is rapidly screened out.
According to a second aspect, an embodiment of the present invention provides an apparatus for screening algorithm models, including: an acquisition module for acquiring a scene detail data set; the setting module is used for setting the test parameters of the algorithm model; the execution module is used for executing a preset test flow based on the scene detail data set and the test parameters to obtain a test result of the algorithm model; and the screening module is used for evaluating and screening the algorithm model according to the test result of the algorithm model.
According to the screening device for the algorithm model, provided by the embodiment of the invention, the scene detail data set is obtained through the obtaining module, the setting of the test parameters of the algorithm model is completed through the setting module, then the execution module is used for testing the scene detail data set according to the preset flow to obtain the test result, and finally the test result is sent to the screening module to determine the optimal screening result, so that the algorithm model with the optimal comprehensiveness is rapidly screened out.
According to a third aspect, an embodiment of the present invention provides a test platform, including: a screening device, configured to perform the screening method of the algorithm model described in the first aspect or any one of the implementation manners of the first aspect.
According to a third aspect, an embodiment of the present invention provides an electronic device, including: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing therein computer instructions, and the processor executing the computer instructions to perform the method for screening an algorithm model according to the first aspect or any one of the embodiments of the first aspect.
According to a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the method for screening an algorithm model according to the first aspect or any one of the implementation manners of the first aspect.
Drawings
The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:
FIG. 1 is a flow chart of a method for screening algorithm models according to an embodiment of the present invention;
FIG. 2 is a block flow diagram of an alternative algorithm model screening method step S10 according to an embodiment of the present invention;
fig. 3 is a block diagram of a passenger car overload detection apparatus according to an embodiment of the present invention;
fig. 4 is a block diagram of an electronic device according to an embodiment of the present invention.
Reference numerals
10-an acquisition module; 11-setting a module; 12-an execution module; 13-a screening module; 20-a processor; 21-memory.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In addition, it should be noted that the screening method of the algorithm model provided by the present invention can be used in cooperation with a test platform, wherein the test platform can include a test management platform module: the platform comprises four basic modules which respectively realize a unified management function on a test flow, a model, data and a test report and provide an interactive interface of a user; jenkins: the method is characterized in that key nodes between a connecting platform and test equipment are connected to carry out automatic test scheduling, and functions of automatic task configuration, triggering, control, state monitoring and the like are realized; example of test equipment: considering that the model needs to operate in different environments, such as GPU, CPU, NPU chip equipment, ARM chip equipment and the like, so as to realize the test result closest to the real scene, all the test equipment resources are uniformly brought into management, and the Jenkins completes the test on the corresponding equipment according to the parameter configuration during the test. Specifically, an embodiment of the present invention provides a method for screening an algorithm model, as shown in fig. 1, the method for screening an algorithm model includes:
s10, a scene detail data set is obtained.
In this embodiment, the acquired scene detail data set is determined according to an algorithm model screened by the data currently required to be tested, for example: an optimal vehicle detection algorithm needs to be obtained, so the required scene detail data set is a video data frame or image data of the road running vehicle.
And S11, setting the test parameters of the algorithm model.
In this embodiment, the test platform needs to be used to set the test parameters of the algorithm model.
And S12, executing a preset test flow based on the scene detail data set and the test parameters to obtain a test result of the algorithm model.
And S13, evaluating and screening the algorithm model according to the test result of the algorithm model.
According to the screening method of the algorithm model, the scene subdivision data set is obtained, the parameter setting is carried out on the algorithm model, the algorithm model is detected according to the preset test flow, and then the tested algorithm model is evaluated and screened according to the detection result, so that the algorithm model with the optimal comprehensive performance is rapidly screened out.
Alternatively, as shown in fig. 2, step S10 includes:
s101, a test data set is obtained.
In this embodiment, the test data set is a set of picture data or a set of video frames, and the detection object in the test data set is determined by a user.
And S102, carrying out scene subdivision based on the test data set to obtain a scene subdivision data set.
In this embodiment, the scene detail data set is subjected to weight distribution according to dimension information based on the test data set to obtain a dimension data set; and carrying out weight distribution on the data sets under the same dimensionality according to the dimensionality data sets to obtain the subdata sets.
Optionally, the algorithm model obtained through deep learning training is finally applied to a real production environment, if only a single scene data set is used for evaluation, a test result is deviated, and performances and test results of the screened model in the real application environment may not be ideal, so that construction of a subdivided scene data set is required, and the model is evaluated in detail from multiple dimensions.
For an application scene of a certain model, the construction of a subdivided scene data set can be performed from n dimensions, for example, for a vehicle detection model, the subdivision can be performed from the perspectives of weather, road conditions, time periods and the like. For the ith dimension, the data set j can be divided into a plurality of subdata sets j, for example, the weather can be divided into sunny days, cloudy days, rainy days, snowy days and foggy days, and the dimension number m at this timei(ii) 5; the road conditions can be divided into urban, high-speed and suburban areas, and the dimension number is mi3; and so on. The total data set constructed is then:
D＝{di,j|i∈[1...n],j∈[1...mi]}
to the sameThe number of subdata sets in the dimension is represented by j, di,jThe data is a subdata set corresponding to the specified dimensionality; because of the difference of factors such as data difficulty, data quantity, scene occurrence frequency and the like, the degree of dependence on indexes of the data is different during testing, and different weights given to the subdata sets are as follows:
W＝{wi,j|i∈[1...n],j∈[1...mi]}
and is provided with
Wherein wi,jThe weight value of the subdata set under the corresponding dimension is obtained;
similarly, there are different weights for dataset dimensions.
V＝{vi|i∈[1...n]}
And is provided with
Wherein v isiIs the weight value of the corresponding dimension.
When the model test is performed based on the subdivided scene data set, the final index of the test result is a weighted result.
In the above formula pi,jThe method is a calculation index of a model test result on a subdivision data set, and p is a comprehensive evaluation index of the model on the test set. The evaluation index may be precision, recall, AP, or the like.
For example, taking vehicle detection as an example, three sub data sets can be constructed by a vehicle detection data set according to three dimensions of road, weather and time, the weights of the three sub data sets are different from each other, and the three sub data sets can be divided according to actual conditions and construct specific data in detail. It can be found that, actually, the data set is divided into sub data sets according to the dimensions, a tree structure is formed, and different weights are given to the results, so that the importance degrees of different sub data sets are reflected in the final index. If necessary, a tree relationship of three layers and four layers of subdata sets can be completely constructed, but generally, the division of two layers of subdata sets is enough, and the complex relationship is not beneficial to data set management and index weighting calculation.
In this embodiment, a hierarchical tree structure may be used to establish a structural awareness of a data set and purposefully guide the construction of a subdivision scenario of the data set. In the automatic test, the data set index at the bottom layer is always directly calculated, and then the weighting calculation is carried out upwards. And finally, when the model is screened, comprehensively screening through the weighting index and the subdata set index to obtain the model with better comprehensive performance.
In the embodiment, the construction of the subdivision scene data set endows the automation test with stronger flexibility, and the optimal model meeting the actual requirement is screened out in different iteration stages according to the product requirement with emphasis and automatic calculation and evaluation.
Optionally, step S11 may be to set test execution conditions and configure the running device, the model to be tested, and the data to be tested. When the specific setting is performed, each item of the test execution condition, the configuration running device, the model to be tested, and the data to be tested may be configured, and in the actual operation, the default value or the last parameter value may be set as required, and the set parameters may specifically include:
triggering conditions are as follows: setting a trigger condition of the automatic test, which may be a code, a model, a data set, equipment and the like, and starting to execute the automatic test step when Jenkins detects that the trigger condition is met;
example of the apparatus: configuring an equipment instance of which the current task needs to be operated;
the model to be tested: configuring a model to be tested for the current task;
data to be tested: configuring a data set to be tested by a current task;
other parameters: other detailed parameters that need to be used are set, including but not limited to confidence thresholds, iou thresholds, nms thresholds, etc.
Optionally, the triggering method provided by Jenkins itself includes: triggering modes such as manual triggering, periodic triggering, precondition triggering, code updating triggering and the like. The manual triggering mode can also be realized by sending a triggering instruction to a Jenkins server. Therefore, the code updating triggering mode can be configured in Jenkins automated testing projects. And automatic triggering in other modes is realized through the test management platform module, and when the model, the data set, the equipment and the like change, the Jenkins service is actively provided with an instruction for triggering automatic testing.
In the embodiment, in order to provide a friendly user interaction interface to configure the relevant parameters, the relevant parameters which are not provided by Jenkins are stored in the platform database, and the parameters are accessed and executed through the interface when the automation code is executed.
Optionally, step S12 includes: executing a preset test flow based on the test data set and the test parameters, further comprising: performing data synchronization based on the test parameters to obtain synchronous data; extracting model data in the test parameters, and converting the model data to obtain an execution model; and performing data recharging test according to the synchronous data and the execution model, and outputting a test result of the algorithm model.
In the embodiment, the script code synchronization is that the automatic test script code synchronization is performed on the specified test equipment instance through git or other modes; the model file synchronization is the synchronization that the test equipment instance receives the latest script code and checks the model file version of the platform; converting the model file into a format required by the test equipment environment, such as a pitorch, mxnet, caffe and the like; the test data synchronization is to carry out version detection and synchronization with the test data of the platform;
the data recharging test is that a specified model is used for recharging the data set in test equipment based on the corresponding software and hardware environment, and a detection result is output.
In this embodiment, the test time can be long or short according to the difference of the sizes of the model and the data set, but the whole process is automatically triggered to run, so that the time at night can be fully utilized for calculation, the running state and the index can be observed at any time, and the manual intervention is timely performed.
Optionally, step S13 includes: obtaining at least 2 test results; and carrying out comparative analysis based on the test result to determine an optimal algorithm model.
In the embodiment, because the test process automatically runs, the model can be tested in a large batch, a part of the early termination threshold value is set, and when the index performance of the model is too poor, the test process can be terminated in advance so as to save the test time and resources. Therefore, after all test processes of a certain task are finished, the test report can be contrasted and analyzed on the platform, and various screening and selection of the required optimal model can be performed by combining factors such as historical indexes, subdivided scene indexes, comprehensive indexes, calculation efficiency and the like.
In the embodiment, Jenkins is used for deep learning algorithm testing, so that configuration management of models, data and testing tasks is facilitated, and a large number of tests of algorithm models under multiple platforms can be completed; by the aid of the method for constructing the subdivided scene data set and calculating the comprehensive indexes, the screened model is closer to a real scene through weighting calculation and has higher comprehensive performance, an automatic test process is triggered through Jenkins, the script file is distributed to designated equipment, the model and the data are automatically synchronized, and then the model and the data are recharged for testing and the report is uploaded.
The embodiment of the present invention provides a screening device for an algorithm model, as shown in fig. 3, the screening device for an algorithm model includes:
the obtaining module 10 is configured to obtain a scene detail data set, and the details refer to the related description of step S10 of the foregoing method embodiment.
The setting module 11 is configured to set the test parameters of the algorithm model, and the details refer to the related description of step S11 of the above method embodiment.
The execution module 12 is configured to execute a preset test procedure based on the scene detail data set and the test parameters to obtain a test result of the algorithm model, and the detailed content refers to the related description of step S12 in the foregoing method embodiment.
The screening module 13 is configured to perform evaluation screening on the algorithm model according to the test result of the algorithm model, and the details refer to the related description of step S13 of the above method embodiment.
The screening device for the algorithm model, provided by the embodiment, comprises an acquisition module, a setting module, an execution module, a screening module and a screening module, wherein the acquisition module acquires a scene detail data set, the setting module is used for setting test parameters of the algorithm model, the execution module is used for testing the scene detail data set according to a preset flow to obtain a test result, and finally the test result is sent to the screening module to determine an optimal screening result, so that the algorithm model with optimal comprehensiveness is rapidly screened.
The test platform provided by the embodiment of the invention can comprise: the screening device can also be provided with a test management module, wherein the test management module is connected with the software interface, and the software interface is also connected with the test equipment. The environment in which the test equipment operates can be at least one of GPU, CPU, NPU chip equipment and ARM chip equipment. An alternative software interface may be Jenkins. And the test management module is provided with a test flow submodule, a model management submodule, a data management submodule and a test report management submodule.
In the embodiment, by constructing a test data set for subdividing a scene, setting algorithm models, test data and test parameters in a screening device, adding an automatic test task through a Jenkins software interface, configuring various trigger conditions and test flows, automatically synchronizing test scripts, models and data to a specified device example, automatically completing the test, and uploading results to a platform for comparative analysis, the algorithm models with the optimal comprehensive performance are rapidly screened.
In addition, an electronic device is further provided in an embodiment of the present invention, as shown in fig. 4, the electronic device may include a processor 20 and a memory 21, where the processor 20 and the memory 21 may be connected by a bus or in another manner, and fig. 4 illustrates an example of connection by a bus.
The processor 20 may be a Central Processing Unit (CPU). The Processor 20 may also be other general purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or any combination thereof.
The memory 21, which is a non-transitory computer-readable storage medium, may be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as program instructions/modules (e.g., the obtaining module 10, the setting module 11, the executing module 12, and the filtering module 13 shown in fig. 3) corresponding to the key shielding method of the in-vehicle display device in the embodiment of the present invention. The processor 20 executes various functional applications and data processing of the processor, namely, a screening method of the algorithm model in the above method embodiment, by running the non-transitory software program, instructions and modules stored in the memory 21.
The memory 21 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created by the processor 20, and the like. Further, the memory 21 may include high speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, memory 21 optionally includes memory located remotely from processor 20, which may be connected to processor 20 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The one or more modules are stored in the memory 21 and, when executed by the processor 20, perform a screening method of an algorithmic model as in the embodiment shown in fig. 1-2.
The details of the electronic device may be understood by referring to the corresponding descriptions and effects in the embodiments shown in fig. 1 to fig. 3, and are not described herein again.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic Disk, an optical Disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a Flash Memory (Flash Memory), a Hard Disk (Hard Disk Drive, abbreviated as HDD), a Solid State Drive (SSD), or the like; the storage medium may also comprise a combination of memories of the kind described above.
Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.
Claims (10)
1. A method for screening algorithm models is characterized by comprising the following steps:
acquiring a scene detail data set;
setting test parameters of the algorithm model;
executing a preset test flow based on the scene detail data set and the test parameters to obtain a test result of the algorithm model;
and evaluating and screening the algorithm model according to the test result of the algorithm model.
2. The method of claim 1, wherein the obtaining the scene detail data set comprises:
acquiring a test data set;
and carrying out scene subdivision on the basis of the test data set to obtain a scene subdivision data set.
3. The method of claim 2, wherein said performing scene segmentation based on said test dataset results in a scene segmentation dataset comprising:
carrying out weight distribution according to dimension information based on the test data set to obtain a dimension data set;
and carrying out weight distribution on the data sets under the same dimensionality according to the dimensionality data sets to obtain subdata sets.
4. The method of claim 1, wherein setting test parameters of the algorithm model comprises: setting test execution conditions, configuring running equipment, a model to be tested and data to be tested.
5. The method of claim 1, wherein performing a predetermined test procedure based on the test data set and the test parameters further comprises:
performing data synchronization based on the test parameters to obtain synchronous data;
extracting model data in the test parameters, and converting the model data to obtain an execution model;
and performing data recharging test according to the synchronous data and the execution model, and outputting the test result of the algorithm model.
6. The method according to claim 1, wherein the evaluation screening of the algorithm model according to the test result of the algorithm model comprises:
obtaining at least 2 test results;
and carrying out comparative analysis based on the test result to determine an optimal algorithm model.
7. An apparatus for screening an algorithm model, comprising:
an acquisition module for acquiring a scene detail data set;
the setting module is used for setting the test parameters of the algorithm model;
the execution module is used for executing a preset test flow based on the scene detail data set and the test parameters to obtain a test result of the algorithm model;
and the screening module is used for evaluating and screening the algorithm model according to the test result of the algorithm model.
8. A test platform, comprising: screening means for performing the method of screening an algorithmic model as set forth in any of claims 1 to 6.
9. An electronic device, comprising:
a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the method of screening an algorithmic model as defined in any of claims 1 to 6.
10. A computer-readable storage medium storing computer instructions for causing a computer to perform the method of filtering an algorithm model of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110379695.9A CN113010435A (en) | 2021-04-08 | 2021-04-08 | Method and device for screening algorithm model and test platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110379695.9A CN113010435A (en) | 2021-04-08 | 2021-04-08 | Method and device for screening algorithm model and test platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113010435A true CN113010435A (en) | 2021-06-22 |
Family
ID=76388265
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110379695.9A Pending CN113010435A (en) | 2021-04-08 | 2021-04-08 | Method and device for screening algorithm model and test platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113010435A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116521344A (en) * | 2023-05-12 | 2023-08-01 | 广州卓勤信息技术有限公司 | AI algorithm scheduling method and system based on resource bus |
| WO2023179133A1 (en) * | 2022-03-22 | 2023-09-28 | 深圳云天励飞技术股份有限公司 | Target algorithm selection method and apparatus, and electronic device and storage medium |
-
2021
- 2021-04-08 CN CN202110379695.9A patent/CN113010435A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023179133A1 (en) * | 2022-03-22 | 2023-09-28 | 深圳云天励飞技术股份有限公司 | Target algorithm selection method and apparatus, and electronic device and storage medium |
| CN116521344A (en) * | 2023-05-12 | 2023-08-01 | 广州卓勤信息技术有限公司 | AI algorithm scheduling method and system based on resource bus |
| CN116521344B (en) * | 2023-05-12 | 2023-10-03 | 广州卓勤信息技术有限公司 | AI algorithm scheduling method and system based on resource bus |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11733693B2 (en) | Data acquisition method and apparatus | |
| CN110659173B (en) | Operation and maintenance system and method | |
| CN113010435A (en) | Method and device for screening algorithm model and test platform | |
| CN111177416A (en) | Event root cause analysis model construction method, event root cause analysis method and device | |
| WO2022001918A1 (en) | Method and apparatus for building predictive model, computing device, and storage medium | |
| CN110109899B (en) | Internet of things data filling method, device and system | |
| CN115049057B (en) | Model deployment method and device, electronic equipment and storage medium | |
| CN111582323A (en) | Power transmission line channel detection method, device and medium | |
| CN104427547A (en) | Business and network association test method, device and system | |
| CN116189172A (en) | 3D target detection method, device, storage medium and chip | |
| CN113660687B (en) | Network difference cell processing method, device, equipment and storage medium | |
| EP4339843A1 (en) | Neural network optimization method and apparatus | |
| CN116452850A (en) | Road ponding area identification method based on data mining and deep learning | |
| CN113298822B (en) | Point cloud data selection method and device, equipment and storage medium | |
| CN114489574B (en) | SVM-based automatic optimization method for stream processing framework | |
| CN114661571A (en) | Model evaluation method, model evaluation device, electronic equipment and storage medium | |
| CN114118440A (en) | Model iteration method, model iteration device, electronic equipment and computer readable storage medium | |
| CN114168657A (en) | Method, system, equipment and medium for detecting wind power abnormal data in real time | |
| CN115866235A (en) | Video quality evaluation method and device, electronic equipment and storage medium | |
| CN110705627A (en) | Target detection method, target detection system, target detection device and readable storage medium | |
| CN117081666B (en) | Fault prediction method, device, electronic equipment, storage medium and program product | |
| CN110084511B (en) | Unmanned aerial vehicle configuration method, device, equipment and readable storage medium | |
| CN117422944A (en) | Equipment end model optimization method, device, equipment and storage medium | |
| CN112837040A (en) | Power data management method and system applied to smart power grid | |
| CN117861198A (en) | Method, client, server and system for adjusting resolution of client application |
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 |