CN1511303A - Scalable expandable system and method for optimizing random system of algorithms for image quality - Google Patents

Scalable expandable system and method for optimizing random system of algorithms for image quality Download PDF

Info

Publication number
CN1511303A
CN1511303A CNA028009525A CN02800952A CN1511303A CN 1511303 A CN1511303 A CN 1511303A CN A028009525 A CNA028009525 A CN A028009525A CN 02800952 A CN02800952 A CN 02800952A CN 1511303 A CN1511303 A CN 1511303A
Authority
CN
China
Prior art keywords
video
module
oiq
tolerance
optimization
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
Application number
CNA028009525A
Other languages
Chinese (zh)
Inventor
W・阿利
W·阿利
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Koninklijke Philips NV
Original Assignee
Koninklijke Philips Electronics NV
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Priority claimed from US09/912,468 external-priority patent/US6813390B2/en
Application filed by Koninklijke Philips Electronics NV filed Critical Koninklijke Philips Electronics NV
Publication of CN1511303A publication Critical patent/CN1511303A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N17/00—Diagnosis, testing or measuring for television systems or their details

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • General Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Testing, Inspecting, Measuring Of Stereoscopic Televisions And Televisions (AREA)
  • Image Processing (AREA)
  • Picture Signal Circuits (AREA)

Abstract

An optimizing video processing method and system selects algorithms for best obtainable video quality for the available computation resources. A video processing module, which processes an input of a video stream, architectural parameters for identifying an order of cascaded video functions and determining a bit precision between data of any consecutive cascaded functions according to an associated complexity level which correlates with a value of available computational resources. An optimizer module optimizes processing of the video stream and includes a plurality of optimization engines each having an associated complexity level. The optimizer module selects an optimization engine according a complexity level which correlates with the value of available computational resources. An Object Image Quality (OIQ) evaluator module evaluates an image quality of an output of the video stream from the video processing module. The OIQ evaluator module includes a plurality of objective image quality metrics having an associated complexity level. The OIQ evaluator module selects a metric according to a correlation factor and a complexity level for said value of available computation resources.

Description

The scalable extendible system and method for optimize image quality random algorithm system
Technical field
The present invention relates to optimize the method and system of video quality.Particularly, but the present invention relates to a kind of expansion scheme with the video algorithm that improves picture quality.
Technical background
In processing system for video, handle vision signal (for example improve acutance, reduce noise, correct color or the like) with a plurality of video functions.Controlled variable that these functions all need a plurality of (a small amount of or a large amount of).
But, there are some very big influence to be arranged in these parameters to picture quality, other then influences very little.In addition, the order of using each function can be a parameter (before the hardware that constitutes the video processing function of realizing software simulation, if perhaps we have the reconfigurable hardware of a high flexible), and perhaps this video system is static, can not change.
Except Video processing function they itself, also have two modules in addition, their complexity can determine the final mass of video system.
The complicacy of objective image quality (OIQ) assessment unit can conform to the principle of simplicity single-signal simple measurement (resembling the rise time of luminance signal) to the very complicated system of imitation human visual system (HVS) psychophysics process.The complexity of optimizing process can be less to calculated amount from the exhaustive type search engine (it needs a large amount of computational resources, almost can't adopt in many practical situation) of greediness, intelligentized exploration type searching method.
Summary of the invention
According to the present invention, a kind of system that optimizes video quality comprises a scalable optimization example, is used to utilize available computational resources that best objective image quality is provided.
A kind of optimization processing system for video comprises:
A video processing module, be used to handle the video flowing of input, this video processing module comprises some structural parameters, is used for determining the order of cascade video functions, according to a complexity that value is relevant of available computational resources, determine arbitrarily the bit accuracy between the cascade function continuously;
An optimal module, be used to optimize processing to video flowing, this optimal module is communicated by letter with video processing module, this optimal module comprises a plurality of optimization engines, each optimizes engine all a relevant complexity, and this optimal module comprises that device is used for selecting to optimize engine according to a complexity relevant with a value of available computational resources; With
An objective image quality (OIQ) evaluation module, be used to assess the picture quality of video processing module video flowing output, this OIQ evaluator includes a plurality of objective image quality metrics of a related complicated degree, and this OIQ evaluation module comprises that device is used for selecting a tolerance according to a complexity of the value of a correlation coefficient r i and available computational resources.
The device of OIQ evaluation module selection tolerance can comprise according to following formula determines a coefficient R:
F = max R { W i n f i } ,
Wherein F is (video quality that system evaluation comes out) final tolerance, by seeking one group of power w iDetermine F, each tolerance f of this group power and a plurality of objective metrics iWhen (its value 1 between n) multiplies each other, can make coefficient R reach maximum with predetermined subjective impression.
Can also there be a computational resource analyzer in this system, is used for selecting at least one complexity of video processing module, optimal module and OIQ evaluation module.
This optimal module can comprise determinacy and uncertainty optimization engine.
This optimal module can include genetic algorithm (GA), simulated annealing (SA) exploration type search engine, forbids search (TS), the exploration type search engine of at least one among artificial evolution (SE) and evolve at random (SE).
In video processing module, optimal module and the OIQ evaluation module at least one is telescopic.
The computational resource analysis module can be that in video processing module, optimal module and the OIQ evaluation module at least one selected complexity by detecting available computational resources.
The method of optimizing video algorithm for available computational resources comprises:
(a) video processing module is according to the order of the cascade video functions of the video flowing of the definite processing of the related complicated degree relevant with a value of available computational resources input video processing module;
(b) select a kind of optimization method, be used for optimizing the processing of video system, this optimization method is to choose from a plurality of optimization methods according to the complexity relevant with that value of available computational resources;
(c) after video flowing is exported from video processing module, the objective image quality of assessment video flowing;
Wherein the assessment of video flowing objective image quality is a tolerance that chooses from a plurality of tolerance by according to a related coefficient and the complexity relevant with that value of computational resource.
The assessment of the middle objective image quality of step (c) comprises according to following formula determines a coefficient R:
F = max R { W i n f i } ,
Wherein F is (video quality that system evaluation comes out) final tolerance, by seeking one group of power w iDetermine F, each tolerance f of this group power and a plurality of objective metrics iWhen (its value 1 between n) multiplies each other, can make coefficient R reach maximum with predetermined subjective impression.
This method can also comprise:
(d) the computational resource analyzer select the step (a) and (b) and (c) at least one related complicated degree.
A plurality of optimization methods that step (b) chooses can comprise deterministic and optimization method uncertainty.
These a plurality of optimization methods include genetic algorithm (GA), simulated annealing (SA), forbid search (TS), the exploration type search engine of at least one among artificial evolution (SE) and evolve at random (SE).
The related complicated degree that chooses in the step (d) can comprise detect the step (a) and (b) and (c) in the step of available computational resources of at least one step.
The video processing module of mentioning in the step (a) is telescopic.
Step (b) can comprise a telescopic optimizer, is used to select optimization method.
Step (c) can provide a telescopic objective image quality evaluation device, is used for assessing the objective image quality.
This system can also comprise a video processing module, an optimal module, scalable objective image quality (OIQ) evaluation module and a computational resource analyzer.
This video processing module comprises a plurality of Video processing function F 1, F 2..., F nEach function all has one group of parameter P i, 1≤i≤n arranges by ascending the influence of the picture quality that obtains according to them.This video processing module has its one group of structural parameters, and it describes the bit accuracy of data bus between the order of cascade Video processing function and any two continuous functions.
This optimal module is a scalable optimizer that multiple possibility optimization mechanism is arranged.This optimal module can comprise a plurality of optimization searching engines that complexity is different with resource requirement.These search engines can be exhaustive types, also can be the exploration types.
Scalable OIQ evaluation module includes a plurality of OIQ tolerance of differing complexity.The complexity table is safeguarded that by the OIQ evaluation module it comprises that all constitute the complexity of measure and each tolerance supposition.
The computational resource analysis module is an arbiter, and it determines on the basis of available computational resources which complexity is all other modules should adopt.
Description of drawings
Fig. 1 is a total synoptic diagram of scalable optimization system among the present invention.
Fig. 2 is a detailed diagram of optimal module shown in Figure 1.
Fig. 3 A is a detailed diagram of objective image quality evaluation device shown in Figure 1.
Fig. 3 B is a scalable dynamic objective metric stream.
Fig. 4 is a method flow diagram of the present invention.
Fig. 5 is the continuation of process flow diagram shown in Figure 4.
Embodiment
Fig. 1's total figure of scalable optimization system among the present invention has drawn.As shown in Figure 1,200, one objective image quality assessment modules 300 of 100, one system optimization modules of a video processing module and an optional computational resource analysis module 400 are arranged.
This video processing module 100 comprises some structural parameters, is used for determining the order of cascade video functions, determines arbitrarily a bit accuracy between the data of cascade function continuously.
As shown in Figure 1, (scope is F a plurality of Video processing functions 102 1~ F n), each function all has one group of structural parameters P i105, scope is P 1~ P nThis group parameter P i(1≤i≤n) wherein, their are by according to they ascending orders that sequences of influence to the picture quality that obtains.
A drawn detailed example of optimal module 110 shown in Figure 1 of Fig. 2.This module comprises a plurality of optimization engines (m search engine), they can be called optimization method 220, and their complexity, statement are different with the computational resource that needs.Optimization method 220 can comprise that (it can upset all predefine parameters in the span to a simple exhaustive type searching method, and a plurality of exploration type engine.
Optimal module is also preserved a record of the predetermined complexity of each method in form 230.This optimal module can be expanded, because any engine that finds can add its back to, as long as define its relative component degree with respect to the complexity of other method.On the basis of the suitable complexity that available resources can be born, parameter in the optimal module and control signal scheduler 235 can call suitable optimization engine.This scheduler obtains control signal, calls suitable method (engine just) and structural parameters.In this embodiment, the complexity that computational resource analyzer 130 (as shown in Figure 1) is selected and/or offered suggestions, but the computational resource analyzer is an optional function.The complexity of suggestion can be selected by for example optimal module.
As an illustration rather than the restriction, certain methods in the optimal module can be a tentative approach, they can be from a kind of greedy method, good result wherein constructs by cascade, become more the exploration type searching method of localization, for example genetic algorithm (GA), simulated annealing (SA), forbid search (TS), artificial evolution (SE) and the mixing of any amount in evolution (SE) and these methods at random.
About this optimal module, provide the example of above-mentioned several exploration type searching methods here in further detail.But those of skill in the art understand these methods.
When using with exploration type method, video processnig algorithms can adopt for example genetic algorithm.Genetic algorithm can be towards the system architecture development that allows optimum picture quality.
Genetic algorithm is to keep the iterative program that a group potential " candidate " separates, and assesses these " candidates " and separates, and distribute an appropriateness value to them.Genetic algorithm is the famous algorithm that solves challenge, " genetic algorithm in optimization and the self-adaptation " this part is the 227th ~ 276 page in the Kronsjo that nineteen ninety publishes and " parallel algorithm progress " book of Shumshesuddin, it is introduced material as a setting here.
Genetic algorithm is a kind of iterative program, and it is safeguarding in a large number the candidate solution with the form coding of chromosome string.Each chromosome has all provided the specific connected mode of different video processing module, and the processing sequence of sequence.Each chromosome then comprises a plurality of genes, and they are Video processing function and their order for video optimized process.
It is a kind of method that emulation is developed, rather than a fixing algorithm, and wherein the complexity that will adopt for optimal module separates the calculating global minimum.
The TABU search is a kind of adaptive program, is used to solve combinatorial optimization problem, and it can be soundd out, and continuation is explored descent path and can not got back to the place that lives through before it.
Artificial evolution is a kind of like this method, and it utilizes series of equations to determine the appropriateness of complexity.
Evolve at random and then utilize the hereditary stochastic variable of a parameter that depends on the explanation genetic program time usually.
Fig. 3 A describes the OIQ evaluation module in detail.OIQ evaluation module 300 comprises the different a plurality of objective image quality metrics of complexity (K tolerance).The OIQ module is preserved the record that it constitutes measure in form 330, and the complexity of each method hypothesis.The OIQ module is extendible, because any tolerance that proposes can be replenished into, as long as provide its complexity in advance.On the basis of the suitable complexity that available resources can be born, the video flowing scheduler 310 in the OIQ module starts suitable OIQ tolerance.
About tolerance, according to the complexity of needed performance and permission, each objective metric 320 all has an evaluation, and it is called quality factor.In other words, quality factor is illustrated on the basis of this tolerance, the quality of vision signal.The people allows to adopt scalable model to the related coefficient of video quality perception, new objective metric can be added in this system and go, and perhaps removes objective metric from this system, as long as provide the correlativity of it and human perception.
Scalable objective metric 320 shown in Fig. 3 B key diagram 3A, it illustrates in greater detail complexity table 330.Each tolerance all has a related coefficient, and (R, 1≤i≤n), 1 corresponding to first tolerance f iN is corresponding to n tolerance f nOn the basis of each related coefficient, evaluator provides a power w for each quality factor i, total coefficient R maximum of attempting making the final composite metric F that following formula represents for predetermined subjective result:
F = max R { W i n f i }
For rapid system (in real time), can close the complicated tolerance of measuring, judge and under the situation of the quality factor that does not have it, carry out.In order to carry out emulation and the optimization of video chain, can spend the longer time in this case, just open more complicated tolerance, their result is considered in the final objective measurement go.
Optionally computational resource analysis module 400 can detect available computational resources, and determines suitable analyzer complexity and suitable OIQ module complexity.
The resource that needs like this,, can provide the computational resource availability, from a series of selections, remove some tolerance, because might surpass active volume to the OIQ evaluation module owing in order to consider the real-time tolerance that requires to have closed some complexity.This value also can be received by system optimization module 200, and the requirement of selected 220 pairs of resources of optimization method must be compatible with given available resources.
Net result is that the algorithm of choosing is optimum according to utilizing operable available resources to reach best objective image quality.This objective image quality then is associated with human visual system's subjective picture quality.On the basis of arbitrfary point resource availability, can select different algorithms and/or different tolerance at any time for given image.This method flexibly makes that picture quality is the highest, because for static system, thereby can not surpass the angle that the resource availability limits from selection algorithm or tolerance, needs a conservative thresholding.If resource is transshipped because of the requirement of algorithm or tolerance, when selecting one to replace algorithm and adapt to the resource of given time, system break will appear, and the time interruption that people's vision can be discovered also can appear in bottom line.
Fig. 4 and Fig. 5 are method flow diagrams of the present invention.
In step (a), need to determine the order of cascade video functions.
In step (b), select an optimization method, be used for optimizing the processing of video flowing.
In step (c) at video flowing after video processing module output, by being that the value of computational resource selects a tolerance to assess the objective image quality of this video flowing according to related coefficient and relevant complexity.
In Fig. 5 description of step (c) how by determining that according to the formula of front a related coefficient assesses the objective image quality.
As previously mentioned, the computational resource module can be reached specific complexity if desired by bypass in optimal module and/or OIQ module.
Those of skill in the art can carry out various improvement and can not depart from the spirit and scope of the invention above system and method.

Claims (14)

1. optimize processing system for video for one kind, comprising:
A video processing module (100), be used to handle the video flowing of input, described video processing module comprises some structural parameters (105), be used for determining the order of cascade video functions (102), determine the bit accuracy between the data of any continuous cascade function according to the relevant complexity relevant with a value of available computational resources;
An optimal module (200), be used to optimize the processing of described video flowing, this optimal module is communicated by letter with described video processing module, this optimal module comprises a plurality of optimization engines (220), each all has a relevant complexity (230), and this optimal module comprises that device is used for selecting an optimization engine (235) according to the complexity relevant with the described value of available computational resources;
An objective image quality (OIQ) evaluation module (300), be used to assess the picture quality of the video flowing (310) of video processing module output, this OIQ evaluator comprises a plurality of objective image quality metrics (320) with related complicated degree (330), and this OIQ evaluation module comprises that device is used for selecting a tolerance (310) according to the coefficient R and the complexity of the described value of available computational resources from a plurality of objective metrics.
2. the system of claim 1, wherein said OIQ evaluation module is selected the device of tolerance to comprise according to following formula to determine described coefficient R:
F = max R { W i n f i } ,
Wherein F is a final tolerance of the video quality that comes out of system evaluation, by seeking one group of power w iDetermine F, each tolerance f of this group power and a plurality of objective metrics i(f iValue 1 between n) multiply each other in, can make coefficient R reach maximum with predetermined subjective impression.
3. the optimization processing system for video of claim 1 also comprises a computational resource analyzer (400), is used to select described video processing module, the related complicated degree of at least one in described optimal module and the described OIQ evaluation module.
4. the system of claim 1, optimal module wherein comprise deterministic and optimization engine uncertainty.
5. the system of claim 4, optimal module wherein include genetic algorithm (GA), simulated annealing (SA), forbid search (TS), artificial evolution (SE) and the exploration type search engine of at least one in the evolution at random.
6. the system of claim 2, optimal module wherein include genetic algorithm (GA), simulated annealing (SA), forbid search (TS), artificial evolution (SE) and the exploration type search engine of at least one in the evolution at random.
7. the processing system for video of claim 1, optimal module wherein is telescopic.
8. the processing system for video of claim 2, optimal module wherein is telescopic.
9. the processing system for video of claim 1, OIQ evaluation module wherein is telescopic.
10. the processing system for video of claim 2, OIQ evaluation module wherein is telescopic.
11. the processing system for video of claim 7, OIQ evaluation module wherein is telescopic.
12. the processing system for video of claim 2, computational resource analysis module wherein is at least one the selection complexity in described video processing module, described optimal module and the described OIQ evaluation module by being at least one the detection available computational resources in described video processing module, described optimal module and the described OIQ evaluation module.
13. be a kind of method of available computational resources optimization video algorithm, this method comprises:
(a) video processing module is according to the order of the cascade video functions of the video flowing of the definite processing of the related complicated degree relevant with a value of available computational resources input video processing module;
(b) select a kind of optimization method, be used for optimizing the processing of video flowing, this optimization method is to choose from a plurality of optimization methods according to the complexity relevant with the described value of available computational resources;
(c) after video flowing is exported from video processing module, the objective image quality of assessment video flowing;
Wherein the assessment of video flowing objective image quality is to determine by choose a tolerance from a plurality of tolerance according to a coefficient R and the complexity relevant with a value of computational resource.
14. the system of claim 13 wherein comprises according to following formula the assessment of described objective image quality in the step (c) and determines described coefficient R:
F = max R { W i n f i } ,
Wherein F is a final tolerance of the video quality that comes out of system evaluation, by seeking one group of power w iDetermine F, each tolerance f of this group power and a plurality of objective metrics i(f iValue 1 between n) multiply each other in, can make coefficient R reach maximum with predetermined subjective impression.
CNA028009525A 2001-03-29 2002-03-28 Scalable expandable system and method for optimizing random system of algorithms for image quality Pending CN1511303A (en)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US27981501P 2001-03-29 2001-03-29
US60/279,815 2001-03-29
US09/912,468 2001-07-25
US09/912,468 US6813390B2 (en) 2001-07-25 2001-07-25 Scalable expandable system and method for optimizing a random system of algorithms for image quality

Publications (1)

Publication Number Publication Date
CN1511303A true CN1511303A (en) 2004-07-07

Family

ID=26959901

Family Applications (1)

Application Number Title Priority Date Filing Date
CNA028009525A Pending CN1511303A (en) 2001-03-29 2002-03-28 Scalable expandable system and method for optimizing random system of algorithms for image quality

Country Status (5)

Country Link
EP (1) EP1433134A2 (en)
JP (1) JP2004527172A (en)
KR (1) KR20030005409A (en)
CN (1) CN1511303A (en)
WO (1) WO2002080563A2 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101345875B (en) * 2008-09-03 2013-08-07 北京中星微电子有限公司 Video algorithm development platform and its development method

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP4257333B2 (en) * 2003-08-22 2009-04-22 日本電信電話株式会社 Image quality evaluation apparatus, image quality evaluation method, image quality evaluation program, image alignment apparatus, image alignment method, and image alignment program
US8422795B2 (en) 2009-02-12 2013-04-16 Dolby Laboratories Licensing Corporation Quality evaluation of sequences of images
CN102170581B (en) * 2011-05-05 2013-03-20 天津大学 Human-visual-system (HVS)-based structural similarity (SSIM) and characteristic matching three-dimensional image quality evaluation method
US9934043B2 (en) 2013-08-08 2018-04-03 Linear Algebra Technologies Limited Apparatus, systems, and methods for providing computational imaging pipeline
US11768689B2 (en) 2013-08-08 2023-09-26 Movidius Limited Apparatus, systems, and methods for low power computational imaging

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5835627A (en) * 1995-05-15 1998-11-10 Higgins; Eric W. System and method for automatically optimizing image quality and processing time
WO1997039417A2 (en) * 1996-03-29 1997-10-23 Sarnoff Corporation Method and apparatus for training a neural network to use a fidelity metric

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101345875B (en) * 2008-09-03 2013-08-07 北京中星微电子有限公司 Video algorithm development platform and its development method

Also Published As

Publication number Publication date
WO2002080563A2 (en) 2002-10-10
EP1433134A2 (en) 2004-06-30
JP2004527172A (en) 2004-09-02
KR20030005409A (en) 2003-01-17
WO2002080563A3 (en) 2004-03-11

Similar Documents

Publication Publication Date Title
JP6978519B2 (en) Predicting the quality of sequencing results using deep neural networks
US10691996B2 (en) Hardware accelerator for compressed LSTM
CN108122032B (en) Neural network model training method, device, chip and system
CN109740731B (en) Design method of self-adaptive convolution layer hardware accelerator
JP7461344B2 (en) Convolution processing method, device, and storage medium for convolutional neural network
RU2007116053A (en) METHOD FOR COMPUTERIZED TRAINING ONE OR MORE NEURAL NETWORKS
CN1416651A (en) Reducable and enlargable objective measurement of evaluating automatic video quality
CN110264984B (en) Model training method, music generation method and device and electronic equipment
CN105302903A (en) Search method, apparatus and system and search result sequence adjustment basis determination method
RU2004139195A (en) IMPROVED APPROACH TO M-ARRAY DECODING AND ERROR CORRECTION
CN113705775A (en) Neural network pruning method, device, equipment and storage medium
Reggiani et al. Pareto optimal design space exploration for accelerated CNN on FPGA
CN1511303A (en) Scalable expandable system and method for optimizing random system of algorithms for image quality
CN108111252B (en) Sequence generation method, data decoding method and device
CN110826707A (en) Acceleration method and hardware accelerator applied to convolutional neural network
KR20200066143A (en) Method for lighting neural network using ruduction-repetition block and apparatus using the same
US6813390B2 (en) Scalable expandable system and method for optimizing a random system of algorithms for image quality
KR101989793B1 (en) An accelerator-aware pruning method for convolution neural networks and a recording medium thereof
Tesser et al. Selecting efficient VM types to train deep learning models on Amazon SageMaker
Mazet et al. Simulation of postive normal variables using several proposal distributions
EP4348511A1 (en) Dynamic activation sparsity in neural networks
CN114676629A (en) Multi-means composite light weight processing method for modulation type recognition model
RU2610012C1 (en) System of innovation project personnel formation
CN104536761A (en) Method and device for evaluating service module processing time
CN112668702B (en) Fixed-point parameter optimization method, system, terminal and storage medium

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication