CN108764333A - One kind being based on the cascade semantic segmentation method of time series, system, terminal and storage medium - Google Patents
One kind being based on the cascade semantic segmentation method of time series, system, terminal and storage medium Download PDFInfo
- Publication number
- CN108764333A CN108764333A CN201810521084.1A CN201810521084A CN108764333A CN 108764333 A CN108764333 A CN 108764333A CN 201810521084 A CN201810521084 A CN 201810521084A CN 108764333 A CN108764333 A CN 108764333A
- Authority
- CN
- China
- Prior art keywords
- branch
- learner
- time period
- area
- learning
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 230000011218 segmentation Effects 0.000 title claims abstract description 30
- 238000001514 detection method Methods 0.000 claims abstract description 11
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000012795 verification Methods 0.000 claims description 37
- 230000008569 process Effects 0.000 claims description 28
- 238000012549 training Methods 0.000 claims description 27
- 238000012360 testing method Methods 0.000 claims description 21
- 230000000694 effects Effects 0.000 claims description 6
- 238000004590 computer program Methods 0.000 claims description 3
- 230000007717 exclusion Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 238000004064 recycling Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 235000013399 edible fruits Nutrition 0.000 claims 1
- JTJMJGYZQZDUJJ-UHFFFAOYSA-N phencyclidine Chemical compound C1CCCCN1C1(C=2C=CC=CC=2)CCCCC1 JTJMJGYZQZDUJJ-UHFFFAOYSA-N 0.000 claims 1
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 230000008859 change Effects 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 230000005055 memory storage Effects 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Electrically Operated Instructional Devices (AREA)
Abstract
The present invention provides one kind and being based on the cascade semantic segmentation method of time series, system, terminal and storage medium, mainly includes the following steps that：It is respectively labeled as branch one and branch two to characteristic pattern I, and establishes learner；The processing time of branch one and branch two is divided into two sections：First time period t1 and second time period t2, first time period t1 learners do learning ability detection, and learner exports learning outcome, which is divided into two parts：Area and unfinished area are completed, Division is completed to branch one, does not complete Division to branch two；In second time period t2, branch two continues to learn not completing area, and the completion area of branch one stops learning and detects whether learning outcome is forgotten；If the unfinished area of branch two remains unfulfilled, the part remained unfulfilled is executed into cycle again.The present invention is based on time series cascades, make learning tasks distribution definitely, greatly reduce performance loss.
Description
Technical field
The present invention relates to vehicle electronics technical fields, and the cascade semantic segmentation side of time series is based on more particularly to one kind
Method, system, terminal and storage medium.
Background technology
Semantic segmentation is one of key technology of image understanding, and image is made of many pixels, and semantic segmentation is just
It is to be grouped pixel according to semantic difference is expressed in image.More and more application scenarios are needed from image in reality
Relevant semanteme is inferred, these applications include automatic Pilot, and human-computer interaction calculates photography, augmented reality etc..
The defect and deficiency of the prior art：First, segmentation precision is inadequate；Second, learning Content redundancy takes more.
Invention content
In order to solve above-mentioned and other potential technical problems, the present invention provides one kind to be cascaded based on time series
Semantic segmentation method, system, terminal and storage medium, first, cascade system in usage time sequence carries out semantic point
It cuts, branch one is not learnt adequately partly to pass to branch two and learn, learning tasks are distributed definitely, are greatly reduced
Performance is lost.
One kind being based on the cascade semantic segmentation method of time series, includes the following steps：
S01：Capture images are obtained, capture images are obtained into characteristic pattern I by front end process of compilation；
S02：It is copied to the characteristic point of characteristic pattern I, becomes identical two parts of characteristic pattern I, is respectively labeled as branch one
With branch two, and learner is established；
S03：The processing time of branch one and branch two is divided into two sections：First time period t1 and second time period t2, the
One period t1 learner does learning ability detection, and learner exports learning outcome, which is divided into two parts：Complete area
With unfinished area, Division is completed to branch one, does not complete Division to branch two；
S04：In second time period t2, branch two continues to learn not completing area, and the completion area of branch one stops study simultaneously
Whether detection learning outcome is forgotten；
S05：In the subsequent time of second time period t2
If the unfinished area study of branch two is completed, terminate to learn；
If the unfinished area of branch two remains unfulfilled, the part remained unfulfilled is executed into step S03 to step S06 again
Cycle, until do not complete be partially completed, just terminate to learn；
S06：The decoding of output learning outcome rear end is reduced to semantic segmentation figure.
Further, the front end process of compilation in the step S01 includes that convolution sum is linearly corrected, further include pond, on
One or more of sampling, fusion.
Further, the number of convolution is the network of convolution at least once in the front end process of compilation in the step S01
Structure is full convolutional network structure or expansion convolutional network structure.
Further, comprising the concrete steps that for learning ability is detected in the step S03：
S031：Data set is divided into test set, verification collection and training set；
S032：Learner, trained learner processing feature figure I is trained to obtain handling result with training set；
S033：Handling result and legitimate reading are compared；
S034：The handling result part consistent with legitimate reading, as completion area；Handling result is inconsistent with legitimate reading
Part, as unfinished area.
Further, the test set is used to assess the study precision of the network model of learner, selects optimize
Practise device network model.
Further, the verification collection is used for determining that network structure or Controlling model are complicated for selecting hyper parameter
The parameter of degree.
Further, the test set in the data set and verification intensive data mutual exclusion, the training set in the data set
Including verification collection.
Further, the data sample in the training set accounts for the 80% of data set, the data sample that the verification is concentrated
Account for the 20% of data set.
Further, when data-oriented collection is too small for simply training, the segmentation tested, verified, it is difficult to generate
When the accurate estimation of extensive error, the performance of learner is judged using cross validation algorithm.
Further, comprising the concrete steps that of whether forgeing of learning outcome is detected in the step S04：
S041：One verification model of setting, test result includes correct, mistake, forgetting, and correct standard, mistake is arranged
One or several in standard, forgetting standard；
S042：Collect the learning outcome that verification branch one completes area's learner with verification；
S043：When verification result index meet one in step S041 in correct standard, error criteria, forgetting standard or
At several, then positive branch one completes the network model of area's learner；When verification result index is not met in step S041 correctly
When one or several in standard, error criteria, forgetting standard, then the network mould that branch one completes area's learner is reselected
Type.
One kind being based on the cascade semantic segmentation system of time series, including front end collector, data set, learner,
It is characterized in that, further includes one module of branch, two module of branch and timer；
The front end collector obtains capture images, and capture images are obtained characteristic pattern I by front end process of compilation；
When the timer is used to divide first time period t1 and second to the learner of one module of branch, two module of branch
Between section t2；
The data set includes test set, training set, verifies collection,
The learner includes learning process management module, and the learning process of the learning process management module is to test
Collection does learning ability detection for one module of branch in first time period t1, finds out the network model of learner；The training set
Network model for the learner of training branch one between first time period t1 and second time period t2；The training set is also
Network model for the learner for training branch two in second time period t2；Verification collection for second time period t2 it
Whether the learning outcome of the learner of verification branch one is forgotten afterwards；
One module of the branch is used to test learning ability in first time period t1, and is learned in second time period t2 verifications
Practise effect；
Two module of the branch is for receiving one module of branch in the unfinished area that first time period t1 learns；And second
The study of time period t 2 does not complete area.
Further, further include loop module, the loop module does not complete study in two second time period t2 of branch
Start under the premise of not completing area's task, loop module is for recycling two part of study branch, the cycle rule of the loop module
Then recycled according to the learning process of learning process management module.
One kind being based on the cascade semantic segmentation terminal of time series, including processor and memory, the memory storage
There are program instruction, the processor operation program instruction to realize the step in above-mentioned method.
A kind of computer readable storage medium, is stored thereon with computer program, which realizes when being executed by processor
Such as the step in above-mentioned method.
As described above, the present invention's has the advantages that：
First, cascade system in usage time sequence carries out semantic segmentation, and branch one is not learnt sufficient part
It passes to branch two to be learnt, learning tasks are distributed definitely, and performance loss is greatly reduced.
Second, branch one obtains learning outcome in first time period t1：After completing area and unfinished area, in the second time
In section t2, one process of branch no longer learns, and only two process of branch learns；Distinguish the period to the different piece of same task into
Row intersects study, and in the parallel progress that the same period learns different piece, verifies, reduces the damage of system performance
Consumption.
Description of the drawings
To describe the technical solutions in the embodiments of the present invention more clearly, make required in being described below to embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the invention, for
For those of ordinary skill in the art, without creative efforts, other are can also be obtained according to these attached drawings
Attached drawing.
Fig. 1 is shown as that the present invention is based on cascade semantic segmentation flow charts.
Specific implementation mode
Illustrate that embodiments of the present invention, those skilled in the art can be by this specification below by way of specific specific example
Disclosed content understands other advantages and effect of the present invention easily.The present invention can also pass through in addition different specific realities
The mode of applying is embodied or practiced, the various details in this specification can also be based on different viewpoints with application, without departing from
Various modifications or alterations are carried out under the spirit of the present invention.It should be noted that in the absence of conflict, following embodiment and implementation
Feature in example can be combined with each other.
It should be clear that structure, ratio, size etc. depicted in this specification institute accompanying drawings, only coordinating specification to be taken off
The content shown is not limited to the enforceable qualifications of the present invention so that those skilled in the art understands and reads, therefore
Do not have technical essential meaning, the modification of any structure, the change of proportionate relationship or the adjustment of size are not influencing the present invention
Under the effect of can be generated and the purpose that can reach, it should all still fall and obtain the model that can cover in disclosed technology contents
In enclosing.Meanwhile cited such as "upper" in this specification, "lower", "left", "right", " centre " and " one " term, be also only
Convenient for being illustrated for narration, rather than to limit the scope of the invention, relativeness is altered or modified, in no essence
It changes under technology contents, when being also considered as the enforceable scope of the present invention.
Referring to Fig. 1,
One kind being based on the cascade semantic segmentation method of time series, includes the following steps：
S01：Capture images are obtained, capture images are obtained into characteristic pattern I by front end process of compilation；
S02：It is copied to the characteristic point of characteristic pattern I, becomes identical two parts of characteristic pattern I, is respectively labeled as branch one
With branch two, and learner is established；
S03：The processing time of branch one and branch two is divided into two sections：First time period t1 and second time period t2, the
One period t1 learner does learning ability detection, and learner exports learning outcome, which is divided into two parts：Complete area
With unfinished area, Division is completed to branch one, does not complete Division to branch two；
S04：In second time period t2, branch two continues to learn not completing area, and the completion area of branch one stops study simultaneously
Whether detection learning outcome is forgotten；
S05：In the subsequent time of second time period t2
If the unfinished area study of branch two is completed, terminate to learn；
If the unfinished area of branch two remains unfulfilled, the part remained unfulfilled is executed into step S03 to step S06 again
Cycle, until do not complete be partially completed, just terminate to learn；
S06：The decoding of output learning outcome rear end is reduced to semantic segmentation figure.
Further, the front end process of compilation in the step S01 includes that convolution sum is linearly corrected, further include pond, on
One or more of sampling, fusion.
Further, the number of convolution is the network of convolution at least once in the front end process of compilation in the step S01
Structure is full convolutional network structure or expansion convolutional network structure.
Further, comprising the concrete steps that for learning ability is detected in the step S03：
S031：Data set is divided into test set, verification collection and training set；
S032：Learner, trained learner processing feature figure I is trained to obtain handling result with training set；
S033：Handling result and legitimate reading are compared；
S034：The handling result part consistent with legitimate reading, as completion area；Handling result is inconsistent with legitimate reading
Part, as unfinished area.
Further, the test set is used to assess the study precision of the network model of learner, selects optimize
Practise device network model.
Further, the verification collection is used for determining that network structure or Controlling model are complicated for selecting hyper parameter
The parameter of degree.
Further, the test set in the data set and verification intensive data mutual exclusion, the training set in the data set
Including verification collection.
Further, the data sample in the training set accounts for the 80% of data set, the data sample that the verification is concentrated
Account for the 20% of data set.
Further, when data-oriented collection is too small for simply training, the segmentation tested, verified, it is difficult to generate
When the accurate estimation of extensive error, the performance of learner is judged using cross validation algorithm.
Further, comprising the concrete steps that of whether forgeing of learning outcome is detected in the step S04：
S041：One verification model of setting, test result includes correct, mistake, forgetting, and correct standard, mistake is arranged
One or several in standard, forgetting standard；
S042：Collect the learning outcome that verification branch one completes area's learner with verification；
S043：When verification result index meet one in step S041 in correct standard, error criteria, forgetting standard or
At several, then positive branch one completes the network model of area's learner；When verification result index is not met in step S041 correctly
When one or several in standard, error criteria, forgetting standard, then the network mould that branch one completes area's learner is reselected
Type.
One kind being based on the cascade semantic segmentation system of time series, including front end collector, data set, learner,
It is characterized in that, further includes one module of branch, two module of branch and timer；
The front end collector obtains capture images, and capture images are obtained characteristic pattern I by front end process of compilation；
When the timer is used to divide first time period t1 and second to the learner of one module of branch, two module of branch
Between section t2；
The data set includes test set, training set, verifies collection,
The learner includes learning process management module, and the learning process of the learning process management module is to test
Collection does learning ability detection for one module of branch in first time period t1, finds out the network model of learner；The training set
Network model for the learner of training branch one between first time period t1 and second time period t2；The training set is also
Network model for the learner for training branch two in second time period t2；Verification collection for second time period t2 it
Whether the learning outcome of the learner of verification branch one is forgotten afterwards；
One module of the branch is used to test learning ability in first time period t1, and is learned in second time period t2 verifications
Practise effect；
Two module of the branch is for receiving one module of branch in the unfinished area that first time period t1 learns；And second
The study of time period t 2 does not complete area.
Further, further include loop module, the loop module does not complete study in two second time period t2 of branch
Start under the premise of not completing area's task, loop module is for recycling two part of study branch, the cycle rule of the loop module
Then recycled according to the learning process of learning process management module.
One kind being based on the cascade semantic segmentation terminal of time series, including processor and memory, the memory storage
There are program instruction, the processor operation program instruction to realize the step in above-mentioned method.
A kind of computer readable storage medium, is stored thereon with computer program, which realizes when being executed by processor
Such as the step in above-mentioned method.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.It is any ripe
The personage for knowing this technology can all carry out modifications and changes to above-described embodiment without violating the spirit and scope of the present invention.Cause
This, technical field includes that institute is complete without departing from the spirit and technical ideas disclosed in the present invention for usual skill such as
At all equivalent modifications or change, should by the present invention claim be covered.
Claims (10)
1. one kind being based on the cascade semantic segmentation method of time series, which is characterized in that include the following steps：
S01：Capture images are obtained, capture images are obtained into characteristic pattern I by front end process of compilation；
S02：It is copied to the characteristic point of characteristic pattern I, becomes identical two parts of characteristic pattern I, be respectively labeled as branch one and is divided
Branch two, and establish learner；
S03：The processing time of branch one and branch two is divided into two sections：First time period t1 and second time period t2, when first
Between section t1 learners do learning ability detection, learner exports learning outcome, which is divided into two parts：Complete Qu Hewei
Area is completed, Division is completed to branch one, does not complete Division to branch two；
S04：In second time period t2, branch two continues to learn not completing area, and the completion area of branch one stops learning and detecting
Whether learning outcome is forgotten；
S05：In the subsequent time of second time period t2
If the unfinished area study of branch two is completed, terminate to learn；
If the unfinished area of branch two remains unfulfilled, the part remained unfulfilled is executed into step S03 following to step S06 again
Ring just terminates to learn until what is do not completed is partially completed；
S06：The decoding of output learning outcome rear end is reduced to semantic segmentation figure.
2. according to claim 1 be based on the cascade semantic segmentation method of time series, which is characterized in that the step
Front end process of compilation in S01 includes that convolution sum is linearly corrected, and further includes one or more of pond, up-sampling, fusion.
3. according to claim 1 be based on the cascade semantic segmentation method of time series, which is characterized in that the step
The number of convolution is that at least once, the network structure of convolution is full convolutional network structure or swollen in front end process of compilation in S01
Swollen convolutional network structure.
4. according to claim 1 be based on the cascade semantic segmentation method of time series, which is characterized in that the step
Detection learning ability comprises the concrete steps that in rapid S03：
S031：Data set is divided into test set, verification collection and training set；
S032：Learner, trained learner processing feature figure I is trained to obtain handling result with training set；
S033：Handling result and legitimate reading are compared；
S034：The handling result part consistent with legitimate reading, as completion area；Handling result and the inconsistent portion of legitimate reading
Point, as unfinished area.
5. according to claim 4 be based on the cascade semantic segmentation method of time series, which is characterized in that the test set
Study precision for the network model for assessing learner selects the learner network model optimized；The verification collection is used for
Hyper parameter is selected, that is, is used for determining the parameter of network structure or Controlling model complexity；Test set in the data set
With verification intensive data mutual exclusion, the training set in the data set includes that verification collects.
6. according to claim 1 be based on the cascade semantic segmentation method of time series, which is characterized in that the step
What whether detection learning outcome was forgotten in S04 comprises the concrete steps that：
S041：One verification model of setting, test result include correct, mistake, forgetting, and be arranged correct standard, error criteria,
One or several in forgetting standard；
S042：Collect the learning outcome that verification branch one completes area's learner with verification；
S043：When verification result index meets one or several in step S041 in correct standard, error criteria, forgetting standard
When, then positive branch one completes the network model of area's learner；When verification result index do not meet correct standard in step S041,
When one or several in error criteria, forgetting standard, then the network model that branch one completes area's learner is reselected.
7. one kind being based on the cascade semantic segmentation system of time series, including front end collector, data set, learner, special
Sign is, further includes one module of branch, two module of branch and timer；
The front end collector obtains capture images, and capture images are obtained characteristic pattern I by front end process of compilation；
The timer is used to divide first time period t1 and second time period to the learner of one module of branch, two module of branch
t2；
The data set includes test set, training set, verifies collection,
The learner includes learning process management module, and the learning process of the learning process management module is that test set is used
Learning ability detection is done in first time period t1 in one module of branch, finds out the network model of learner；The training set is used for
The network model of the learner of training branch one between first time period t1 and second time period t2；The training set is additionally operable to
The network model of the learner of branch two is trained in second time period t2；The verification collection is for the posteriority in second time period t2
Whether the learning outcome for demonstrate,proving the learner of branch one is forgotten；
One module of the branch is used to test learning ability in first time period t1, and in second time period t2 verification study effects
Fruit；
Two module of the branch is for receiving one module of branch in the unfinished area that first time period t1 learns, and in the second time
Section t2 study does not complete area.
8. according to claim 7 be based on the cascade semantic segmentation system of time series, which is characterized in that further include cycle
Module, the loop module, which is not completed in two second time period t2 of branch under the premise of study does not complete area's task, to be started, and is followed
Ring moulds block for recycling study branch two part, the cycline rule of the loop module according to learning process management module study
Flow recycles.
9. one kind being based on the cascade semantic segmentation terminal of time series, which is characterized in that described to deposit including processor and memory
Reservoir has program stored therein instruction, and the processor operation program instruction realizes the step in above-mentioned method.
10. a kind of computer readable storage medium, is stored thereon with computer program, it is characterised in that：The program is by processor
The step in the method as described in claim 1 to 6 any claim is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810521084.1A CN108764333B (en) | 2018-05-28 | 2018-05-28 | Learner training method, system, terminal and storage medium for semantic segmentation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810521084.1A CN108764333B (en) | 2018-05-28 | 2018-05-28 | Learner training method, system, terminal and storage medium for semantic segmentation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764333A true CN108764333A (en) | 2018-11-06 |
| CN108764333B CN108764333B (en) | 2024-07-26 |
Family
ID=64005928
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810521084.1A Active CN108764333B (en) | 2018-05-28 | 2018-05-28 | Learner training method, system, terminal and storage medium for semantic segmentation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764333B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507231A (en) * | 2020-04-10 | 2020-08-07 | 三一重工股份有限公司 | Automatic detection method and system for correctness of process steps |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060106797A1 (en) * | 2004-11-17 | 2006-05-18 | Narayan Srinivasa | System and method for temporal data mining |
| US20170147905A1 (en) * | 2015-11-25 | 2017-05-25 | Baidu Usa Llc | Systems and methods for end-to-end object detection |
| CN107480726A (en) * | 2017-08-25 | 2017-12-15 | 电子科技大学 | A kind of Scene Semantics dividing method based on full convolution and shot and long term mnemon |
-
2018
- 2018-05-28 CN CN201810521084.1A patent/CN108764333B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060106797A1 (en) * | 2004-11-17 | 2006-05-18 | Narayan Srinivasa | System and method for temporal data mining |
| US20170147905A1 (en) * | 2015-11-25 | 2017-05-25 | Baidu Usa Llc | Systems and methods for end-to-end object detection |
| CN107480726A (en) * | 2017-08-25 | 2017-12-15 | 电子科技大学 | A kind of Scene Semantics dividing method based on full convolution and shot and long term mnemon |
Non-Patent Citations (2)
| Title |
|---|
| YANWEI PANG 等: "Cascade Learning by Optimally Partitioning", 《IEEE TRANSACTIONS ON CYBERNETICS》, vol. 47, no. 12, 31 December 2017 (2017-12-31), pages 4148 - 4161, XP011672763, DOI: 10.1109/TCYB.2016.2601438 * |
| 廖大强 等: "基于多分支RNN快速学习算法的混沌时间序列预测", 《计算机应用研究》, vol. 32, no. 2, 28 February 2015 (2015-02-28), pages 403 - 408 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111507231A (en) * | 2020-04-10 | 2020-08-07 | 三一重工股份有限公司 | Automatic detection method and system for correctness of process steps |
| CN111507231B (en) * | 2020-04-10 | 2023-06-23 | 盛景智能科技(嘉兴)有限公司 | Automatic detection method and system for correctness of process steps |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764333B (en) | 2024-07-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108124487B (en) | Cloud meter reading method and device | |
| CN108229588B (en) | Machine learning identification method based on deep learning | |
| CN108805180B (en) | Target object detection method and device | |
| CN117173158B (en) | Intelligent detection method and system for quality of precise connector | |
| CN113240673B (en) | Defect detection method, defect detection device, electronic equipment and storage medium | |
| WO2024066035A1 (en) | Defect detection method and system based on battery surface image, and related device | |
| CN110969600A (en) | Product defect detection method and device, electronic equipment and storage medium | |
| CN109885929B (en) | Automatic driving decision planning data reproduction method and device | |
| CN115860091A (en) | Depth feature descriptor learning method based on orthogonal constraint | |
| CN115512169A (en) | Weak supervision semantic segmentation method and device based on gradient and region affinity optimization | |
| CN115100489A (en) | Image processing method, device and equipment and readable storage medium | |
| CN108764333A (en) | One kind being based on the cascade semantic segmentation method of time series, system, terminal and storage medium | |
| CN111338938A (en) | Slider verification method and device based on dynamic path | |
| CN110427998A (en) | Model training, object detection method and device, electronic equipment, storage medium | |
| CN112308149B (en) | Optimization method and device for image information identification based on machine learning | |
| CN115795900A (en) | Software in-loop simulation test method and device for automatic driving system and electronic equipment | |
| CN113284141A (en) | Model determination method, device and equipment for defect detection | |
| CN109284687B (en) | Scene recognition method and device based on indoor opportunity signal enhancement | |
| CN112883988B (en) | Training and feature extraction method of feature extraction network based on multiple data sets | |
| CN117934478B (en) | Defect detection method, device, equipment and medium | |
| US20240078464A1 (en) | Passive outlier-tolerant and active outlier-tolerant learning method with abnormal samples and system thereof | |
| CN116843572B (en) | Image restoration method based on coordinates | |
| CN112766391B (en) | Method, system, equipment and medium for making document | |
| CN112053393B (en) | Image depth estimation method and device | |
| CN117635547B (en) | Method and device for detecting appearance defects of medical gloves |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |