CN107358143A - Sweep forward model integrated method, apparatus, storage device and face identification system - Google Patents
Sweep forward model integrated method, apparatus, storage device and face identification system Download PDFInfo
- Publication number
- CN107358143A CN107358143A CN201710350359.5A CN201710350359A CN107358143A CN 107358143 A CN107358143 A CN 107358143A CN 201710350359 A CN201710350359 A CN 201710350359A CN 107358143 A CN107358143 A CN 107358143A
- Authority
- CN
- China
- Prior art keywords
- model
- human face
- face recognition
- models sets
- recognition model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- 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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Collating Specific Patterns (AREA)
Abstract
The embodiment of the invention discloses a kind of sweep forward model integrated method and apparatus, and a kind of storage device and a kind of face identification system, methods described to include：Obtain the N number of human face recognition model for having completed training；With the element of N number of human face recognition model alternatively supplementary set, using i initial value as 1, and a human face recognition model is selected as the i-th Models Sets from the selection supplementary set；Then following sweep forward operation is performed：Supplementary set of i-th Models Sets in the selection supplementary set is updated to the selection supplementary set；A human face recognition model and i-th Models Sets and integrated i+1 Models Sets are selected from the selection supplementary set；When judging that the i+1 Models Sets meet to stop the condition of sweep forward, the i+1 Models Sets are exported, otherwise, returns to i plus in the lump and performs the sweep forward operation.Using the embodiment of the present invention, model can be fast and effeciently filtered out, avoids search space excessive while redundancy is reduced.
Description
Technical field
The present invention relates to computer face identification technology field, more particularly to a kind of sweep forward model integrated method and dress
Put, and a kind of storage device and a kind of face identification system.
Background technology
Found in face recognition algorithms research, the accuracy of individual human face identification model in general than relatively limited, compared with
Conventional way is to train different human face recognition models, then by these model combination of sets into using together.After integrated
Model can typically have higher recognition accuracy than single model.This thinking is applicable not only to face recognition algorithms, general
Area of pattern recognition it is equally applicable.
Existing way is that directly the multiple models trained are put together use, is exported according to polyalgorithm model,
The determination of final recognition result is carried out by the way of average or ballot.When inventor will implement the present invention, existing do is found
Problems be present in method：
1st, we are it is generally desirable to the model trained to same task is The more the better, but there may be between multiple models
Redundancy, therefore we need quickly and efficiently to be screened in the model trained, carry out model selection；
2nd, multiple models do not account for the information transmission between each model and exchange in the training process, therefore most
The cooperative and complementarity of whole cognitive phase can be weak, and on the one hand can not accomplish the global information of the multiple models of Comprehensive,
On the other hand there is very big redundancy in the feature of multiple model outputs, aggravate amount of storage and amount of calculation.
The content of the invention
The sweep forward model integrated method and apparatus that the embodiment of the present invention proposes, and storage device and recognition of face system
System, can fast and effeciently filter out model, avoid search space excessive while redundancy is reduced.
The embodiment of the present invention provides a kind of sweep forward model integrated method, including：
Obtain the N number of human face recognition model for having completed training；
With the element of N number of human face recognition model alternatively supplementary set, using i initial value as 1, and mended from the selection
Concentration selects a human face recognition model as the i-th Models Sets；Then following sweep forward operation is performed：
Supplementary set of i-th Models Sets in the selection supplementary set is updated to the selection supplementary set；
A human face recognition model and i-th Models Sets and integrated i+1 model are selected from the selection supplementary set
Collection；
When judging that the i+1 Models Sets meet to stop the condition of sweep forward, the i+1 Models Sets are exported, with
Recognition of face is carried out according to the integrated model of the human face recognition model by being included in the i+1 Models Sets for face identification system
Work；Otherwise, returned to i plus in the lump and perform the sweep forward operation；i<N.
Further, it is described to select a human face recognition model as the i-th Models Sets from the selection supplementary set, specifically
For：
Test discrimination highest human face recognition model is selected from the selection supplementary set as the i-th Models Sets；Wherein,
The test discrimination refers to the success rate tested using human face recognition model the picture of standard faces test set.
Further, it is described to select a human face recognition model and i-th Models Sets simultaneously from the selection supplementary set
Integrated i+1 Models Sets, it is specially：
Each element in the selection supplementary set is gathered with i-th Models Sets and integrated one point respectively；
Each point of set formed for union, calculate the recognition of face mould that all elements in being gathered by this point integrate
The test discrimination of type；
Diversity cooperation corresponding to choosing test discrimination highest is i+1 Models Sets.
Yet further, it is described to judge that the i+1 Models Sets meet the condition for stopping sweep forward, be specially：
By the integrated human face recognition model of all elements in the i+1 Models Sets test discrimination with by described the
The difference of the test discrimination for the human face recognition model that all elements in i Models Sets integrate is more than sweep forward threshold value.
Further, the process of the test discrimination of human face recognition model is calculated, is specially：
Using principal component analytical method from the human face recognition model extraction model feature；The array length of the aspect of model
Spend for preset group length；
The picture of the standard faces test set is tested according to the aspect of model extracted；
Success rate in statistical test procedures, and identified the success rate as the test of the human face recognition model
Rate.
Correspondingly, the embodiment of the present invention also provides a kind of sweep forward model integrated device, including：
Model acquisition module, N number of human face recognition model of training is completed for obtaining；
Initialization module, for the element of N number of human face recognition model alternatively supplementary set, using i initial value as
1, and a human face recognition model is selected as the i-th Models Sets from the selection supplementary set；
Sweep forward module is used to perform following sweep forward operation, specifically includes with lower unit：
Supplementary set updating block, for supplementary set of i-th Models Sets in the selection supplementary set to be updated into the selection
Supplementary set；
Union unit is chosen, for selecting a human face recognition model and i-th model from the selection supplementary set
Collect and integrate i+1 Models Sets；
Cycle criterion unit, for when judging that the i+1 Models Sets meet to stop the condition of sweep forward, exporting institute
I+1 Models Sets are stated, are integrated for face identification system according to the human face recognition model by being included in the i+1 Models Sets
Model carries out recognition of face work；Otherwise, returned to i plus in the lump and perform the sweep forward operation；i<N.
Further, the initialization module is specifically used for：
Test discrimination highest human face recognition model is selected from the selection supplementary set as the i-th Models Sets；Wherein,
The test discrimination refers to the success rate tested using human face recognition model the picture of standard faces test set.
Further, the selection union unit, it is specially：
Union subelement, for by each element in the selection supplementary set respectively with i-th Models Sets and integrating
One point of set；
Gather computation subunit, for each point of set formed for union, calculate all in being gathered by this point
The test discrimination of the integrated human face recognition model of element；
Subelement is chosen, is i+1 Models Sets for diversity cooperation corresponding to selection test discrimination highest.
In addition, the embodiment of the present invention also provides a kind of storage device, wherein being stored with a plurality of instruction, the instruction is processed
Device realizes any embodiment of foregoing sweep forward model integrated method when performing.
And the embodiment of the present invention also provides a kind of face identification system, including storage device, processor and it is stored in institute
The a plurality of instruction that can be run in storage device and on the processor is stated, wherein, it is real when computing device institute art instructs
The now any embodiment of foregoing sweep forward model integrated method.
Implement the embodiment of the present invention, have the advantages that：
A kind of sweep forward model integrated method and apparatus provided in an embodiment of the present invention, and a kind of storage device and one
Kind face identification system, a model is selected from multiple human face recognition models, is then selected again from remaining multiple models
Take out a model and select the model come before and integrated, and then it is preceding to searching to judge whether integrated model meets to stop
The condition of rope, if it is satisfied, then directly export the integrated human face recognition model as identification instrument, otherwise return continue from
Choose a model in remaining multiple models to be integrated with the models come that select all before again, until meeting before stopping
It is that extremely, such a pattern search mode is based only on current situation and carries out decision-making, as long as meeting that condition is then jumped to the condition of search
Go out circulation, amount of calculation can be substantially reduced, without the possibility that worry about is more remote.And chosen from remaining model
The standard of model is the height of the test discrimination of the model after integrating, and improves the effective percentage of screening model, and then is improved integrated
Model test discrimination.In addition, using principal component analysis mode to human face recognition model extraction model feature, can be to mould
Type feature carries out fusion compression, reduces the redundancy between multiple models.
Brief description of the drawings
Fig. 1 is the schematic flow sheet of one embodiment of sweep forward model integrated method provided by the invention；
Fig. 2 is the structural representation of one embodiment of sweep forward model integrated device provided by the invention；
Fig. 3 is the structure of one embodiment of the sweep forward module of sweep forward model integrated device provided by the invention
Schematic diagram；
Fig. 4 is the one of the selection union unit of the sweep forward module of sweep forward model integrated device provided by the invention
The structural representation of individual embodiment；
Fig. 5 is the knot of one embodiment of the discrimination computing module of sweep forward model integrated device provided by the invention
Structure schematic diagram.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, rather than whole embodiments.It is based on
Embodiment in the present invention, those of ordinary skill in the art are obtained every other under the premise of creative work is not made
Embodiment, belong to the scope of protection of the invention.
It is the schematic flow sheet of one embodiment of sweep forward model integrated method provided by the invention referring to Fig. 1；Should
Sweep forward model integrated method, including step S1 to S7, it is specific as follows：
S1, obtain the N number of human face recognition model for having completed training；N number of human face recognition model is to train in advance, example
A face picture such as is split into 25 face region units to go to identify respectively, then just have 25 different recognition of face moulds
Type.
S2, with the element of N number of human face recognition model alternatively supplementary set, using i initial value as 1, and from the choosing
Select and a human face recognition model is selected in supplementary set as the i-th Models Sets, the i-th current Models Sets are the 1st Models Sets；Then hold
The following sweep forward operation of row：
S3, supplementary set of i-th Models Sets in the selection supplementary set is updated to the selection supplementary set；Purpose is so that
It is never to be selected in the model come to choose so that in i+1 Models Sets subsequently from the selection supplementary set during Selection Model
Human face recognition model it is different, avoid it is integrated after model have the composition of obvious redundancy.
S4, a human face recognition model and i-th Models Sets and integrated i+1 mould are selected from the selection supplementary set
Type collection；
S5, judges whether the i+1 Models Sets meet the condition for stopping sweep forward；
S6, if so, the i+1 Models Sets are exported,
S7, the sweep forward operation i.e. step S3 to step S5 is performed if it is not, adding to i and returning in the lump；i<N.
It should be noted that during sweep forward operation is performed, after i plus one, when the numerical value for detecting i is N,
Direct defeated 1st Models Sets.
Further, above-mentioned steps S2 Criterion of Selecting is the height of test discrimination, thus, in implementation column of the present invention
In, it can apply for testing discrimination highest model, then step S2 is embodied as：
Test discrimination highest human face recognition model is selected from the selection supplementary set as the i-th Models Sets；Wherein,
The test discrimination refers to the success rate tested using human face recognition model the picture of standard faces test set.Herein
I Models Sets are the 1st Models Sets, only include an element.
Further, above-mentioned steps S4's is that selection one can be with having chosen before from the model of remaining unselected mistake
The test discrimination highest model for the model that all human face recognition models crossed integrate, by above-mentioned steps S4 specific implementation
Cheng Wei：
Each element in the selection supplementary set is gathered with i-th Models Sets and integrated one point respectively；
Each point of set formed for union, calculate the recognition of face mould that all elements in being gathered by this point integrate
The test discrimination of type；
Diversity cooperation corresponding to choosing test discrimination highest is i+1 Models Sets.
It should be noted that the model chosen before current i numerical identity is as a set, as the i-th mould
Type collection.By being circularly set for above-mentioned steps S4, as long as obtaining the Models Sets in current i numerical identities, the Models Sets
Element is that the model come and integrated test discrimination highest are selected from N number of human face recognition model, follow-up above-mentioned step
The process of rapid S4 unions, the element for also causing i+1 Models Sets are that the model come is selected from N number of human face recognition model
And integrated test discrimination highest, then represented so that the comparison screening of follow-up i+1 Models Sets and the i-th Models Sets has
Property, can effectively filter out the multiple models high to face recognition accuracy integrates, and improves the efficiency of search.
Specifically, the judgement i+1 Models Sets in above-mentioned steps S5 meet the condition for stopping sweep forward, specifically
For：
By the integrated human face recognition model of all elements in the i+1 Models Sets test discrimination with by described the
The difference of the test discrimination for the human face recognition model that all elements in i Models Sets integrate is more than sweep forward threshold value.
The process of the test discrimination for the calculating human face recognition model that above-described embodiment is related to, it is specially：
Using principal component analytical method from the human face recognition model extraction model feature；The array length of the aspect of model
Spend for preset group length；
The picture of the standard faces test set is tested according to the aspect of model extracted；
Success rate in statistical test procedures, and identified the success rate as the test of the human face recognition model
Rate.
In embodiments of the present invention, it is for the application scenarios of feature, the feature array length of human face recognition model output
Predefined is good, thus assume that the number of the primitive character of any one human face recognition model of N number of human face recognition model
Group length is 4096, and the aspect of model array length extracted through principal component analytical method is 400, there is obvious compression and superfluous
The reduction of remaining, in follow-up sweep forward process, as the model quantity in Models Sets increases, the fusion between the aspect of model
Decline with redundancy can be more obvious.And integrated between multiple models is that the aspect of model of each model is serially connected in into one
Block, such as the array length of the primitive character of first model are 4096, then the primitive character of the model after two model integrateds
Array length be 4096*2=8192, knot principal component analysis can will it is integrated after model Feature Compression to length be 400,
The information fusion between model is added, reduces the characteristic information redundancy inside model between model again.
A kind of sweep forward model integrated method provided in an embodiment of the present invention, is selected from multiple identification models of having the face
One model, then select a model from remaining multiple models again and integrated with selecting the model come before,
And then judge whether integrated model meets the condition for stopping sweep forward, if it is satisfied, then directly exporting the integrated face
Identification model as identification instrument, otherwise return after from remaining multiple models choose a model again with all choosings before
The model taken out is integrated, until the condition for meeting to stop sweep forward is extremely, such a pattern search mode is based only on
Current situation does plan, as long as meeting that condition then jumps out circulation, can substantially reduce amount of calculation, more remote without worry about
Possibility.And the standard of Selection Model is the height of the test discrimination of the model after integrating from remaining model, is carried
The effective percentage of high screening model, and then improve the test discrimination of integrated model.In addition, using principal component analysis mode to people
Face identification model extraction model feature, fusion compression can be carried out to the aspect of model, reduce the redundancy between multiple models.
In addition, the embodiment of the present invention also provides a kind of storage device, wherein being stored with a plurality of instruction, the instruction is processed
Device realizes any embodiment of foregoing sweep forward model integrated method when performing.
And the embodiment of the present invention also provides a kind of face identification system, including storage device, processor and it is stored in institute
The a plurality of instruction that can be run in storage device and on the processor is stated, wherein, it is real when computing device institute art instructs
The now any embodiment of foregoing sweep forward model integrated method.
Referring to Fig. 2 and Fig. 3, Fig. 2 is the structure of one embodiment of sweep forward model integrated device provided by the invention
Schematic diagram, Fig. 3 are the structures of one embodiment of the sweep forward module of sweep forward model integrated device provided by the invention
Schematic diagram；The model integrated device can implement whole flows of the method for above-described embodiment, specifically include：
Model acquisition module 10, N number of human face recognition model of training is completed for obtaining；
Initialization module 20, for the element of N number of human face recognition model alternatively supplementary set, with i initial value
For 1, and a human face recognition model is selected as the i-th Models Sets from the selection supplementary set；
Sweep forward module 30 is used to perform following sweep forward operation, specifically includes with lower unit：
Supplementary set updating block 31, for supplementary set of i-th Models Sets in the selection supplementary set to be updated into the choosing
Select supplementary set；
Union unit 32 is chosen, for selecting a human face recognition model and i-th mould from the selection supplementary set
Type collection and integrated i+1 Models Sets；
Cycle criterion unit 33, for when judging that the i+1 Models Sets meet to stop the condition of sweep forward, exporting
The i+1 Models Sets, otherwise, return to i plus in the lump and perform the sweep forward operation；i<N.
Further, the initialization module 20 is specifically used for：
Test discrimination highest human face recognition model is selected from the selection supplementary set as the i-th Models Sets；Wherein,
The test discrimination refers to the success rate tested using human face recognition model the picture of standard faces test set.
It is the selection union list of the sweep forward module of sweep forward model integrated device provided by the invention referring to Fig. 4
The structural representation of one embodiment of member；
Further, the selection union unit 32, it is specially：
Union subelement 321, for by it is described selection supplementary set in each element respectively with the i-th Models Sets union
Into a point of set；
Gather computation subunit 322, for each point of set formed for union, calculate the institute in being gathered by this point
There is the test discrimination of the integrated human face recognition model of element；
Subelement 323 is chosen, is i+1 Models Sets for diversity cooperation corresponding to selection test discrimination highest.
Further, it is described to judge that the i+1 Models Sets meet the condition for stopping sweep forward, be specially：
By the integrated human face recognition model of all elements in the i+1 Models Sets test discrimination with by described the
The difference of the test discrimination for the human face recognition model that all elements in i Models Sets integrate is more than sweep forward threshold value.
It is that one of the discrimination computing module of sweep forward model integrated device provided by the invention implements referring to Fig. 5
The structural representation of example；
Further, the test that the sweep forward model integrated device also includes being used to calculate human face recognition model is known
The not discrimination computing module 40 of rate, is specifically included：
Feature extraction unit 41, for using principal component analytical method from the human face recognition model extraction model feature；
The array length of the aspect of model is preset group length；
Test cell 42, for being surveyed according to the aspect of model extracted to the picture of the standard faces test set
Examination；
Statistic unit 43, for the success rate in statistical test procedures, and using the success rate as the recognition of face
The test discrimination of model.
Implement the embodiment of the present invention, have the advantages that：
A kind of sweep forward model integrated device provided in an embodiment of the present invention, is selected from multiple identification models of having the face
One model, then select a model from remaining multiple models again and integrated with selecting the model come before,
And then judge whether integrated model meets the condition for stopping sweep forward, if it is satisfied, then directly exporting the integrated face
Identification model as identification instrument, otherwise return after from remaining multiple models choose a model again with all choosings before
The model taken out is integrated, until the condition for meeting to stop sweep forward is extremely, such a pattern search mode is based only on
Current situation does plan, as long as meeting that condition then jumps out circulation, can substantially reduce amount of calculation, more remote without worry about
Possibility.And the standard of Selection Model is the height of the test discrimination of the model after integrating from remaining model, is carried
The effective percentage of high screening model, and then improve the test discrimination of integrated model.In addition, using principal component analysis mode to people
Face identification model extraction model feature, fusion compression can be carried out to the aspect of model, reduce the redundancy between multiple models.
One of ordinary skill in the art will appreciate that realize all or part of flow in above-described embodiment method, being can be with
The hardware of correlation is instructed to complete by computer program, described program can be stored in a computer read/write memory medium
In, the program is upon execution, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, described storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Described above is the preferred embodiment of the present invention, it is noted that for those skilled in the art
For, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications are also considered as
Protection scope of the present invention.
Claims (10)
- A kind of 1. sweep forward model integrated method, it is characterised in that including：Obtain the N number of human face recognition model for having completed training；With the element of N number of human face recognition model alternatively supplementary set, using i initial value as 1, and from the selection supplementary set A human face recognition model is selected as the i-th Models Sets；Then following sweep forward operation is performed：Supplementary set of i-th Models Sets in the selection supplementary set is updated to the selection supplementary set；A human face recognition model and i-th Models Sets and integrated i+1 Models Sets are selected from the selection supplementary set；When judging that the i+1 Models Sets meet to stop the condition of sweep forward, the i+1 Models Sets are exported, for people Face identifying system carries out recognition of face work according to the integrated model of the human face recognition model by being included in the i+1 Models Sets Make；Otherwise, returned to i plus in the lump and perform the sweep forward operation；i<N.
- 2. sweep forward model integrated method as claimed in claim 1, it is characterised in that described to be selected from the selection supplementary set A human face recognition model is taken out as the i-th Models Sets, is specially：Test discrimination highest human face recognition model is selected from the selection supplementary set as the i-th Models Sets；Wherein, it is described Test discrimination refers to the success rate tested using human face recognition model the picture of standard faces test set.
- 3. sweep forward model integrated method as claimed in claim 2, it is characterised in that described to be selected from the selection supplementary set A human face recognition model and i-th Models Sets and integrated i+1 Models Sets are taken out, is specially：Each element in the selection supplementary set is gathered with i-th Models Sets and integrated one point respectively；Each point of set formed for union, calculate the human face recognition model that all elements in being gathered by this point integrate Test discrimination；Diversity cooperation corresponding to choosing test discrimination highest is i+1 Models Sets.
- 4. sweep forward model integrated method as claimed in claim 3, it is characterised in that described to judge the i+1 model Collection meets the condition for stopping sweep forward, is specially：By the test discrimination of the integrated human face recognition model of all elements in the i+1 Models Sets and by i-th mould The difference of the test discrimination for the human face recognition model that all elements that type is concentrated integrate is more than sweep forward threshold value.
- 5. the sweep forward model integrated method as described in any one of claim 2 to 4, it is characterised in that calculate recognition of face The process of the test discrimination of model, it is specially：Using principal component analytical method from the human face recognition model extraction model feature；The array length of the aspect of model is Preset group length；The picture of the standard faces test set is tested according to the aspect of model extracted；Success rate in statistical test procedures, and the test discrimination using the success rate as the human face recognition model.
- A kind of 6. sweep forward model integrated device, it is characterised in that including：Model acquisition module, N number of human face recognition model of training is completed for obtaining；Initialization module, for the element of N number of human face recognition model alternatively supplementary set, using i initial value as 1, and A human face recognition model is selected as the i-th Models Sets from the selection supplementary set；Sweep forward module is used to perform following sweep forward operation, specifically includes with lower unit：Supplementary set updating block, for supplementary set of i-th Models Sets in the selection supplementary set to be updated into the selection supplementary set；Union unit is chosen, for selecting a human face recognition model and i-th Models Sets simultaneously from the selection supplementary set Integrated i+1 Models Sets；Cycle criterion unit, for when judging that the i+1 Models Sets meet to stop the condition of sweep forward, output described the I+1 Models Sets, so that face identification system is according to the integrated model of the human face recognition model by being included in the i+1 Models Sets Carry out recognition of face work；Otherwise, returned to i plus in the lump and perform the sweep forward operation；i<N.
- 7. sweep forward model integrated device as claimed in claim 6, it is characterised in that the initialization module is specifically used In：Test discrimination highest human face recognition model is selected from the selection supplementary set as the i-th Models Sets；Wherein, it is described Test discrimination refers to the success rate tested using human face recognition model the picture of standard faces test set.
- 8. sweep forward model integrated device as claimed in claim 7, it is characterised in that the selection union unit, specifically For：Union subelement, for by each element in the selection supplementary set respectively with i-th Models Sets and integrated one Divide set；Gather computation subunit, for each point of set formed for union, calculate all elements in being gathered by this point The test discrimination of integrated human face recognition model；Subelement is chosen, is i+1 Models Sets for diversity cooperation corresponding to selection test discrimination highest.
- 9. a kind of storage device, wherein being stored with a plurality of instruction, it is characterised in that the instruction is realized such as when being executed by processor Sweep forward model integrated method described in any one of claim 1 to 5.
- 10. a kind of face identification system, it is characterised in that including storage device, processor and be stored in the storage device And a plurality of instruction that can be run on the processor, wherein, such as claim 1 is realized when computing device institute art instructs To the sweep forward model integrated method described in 5 any one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710350359.5A CN107358143A (en) | 2017-05-17 | 2017-05-17 | Sweep forward model integrated method, apparatus, storage device and face identification system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710350359.5A CN107358143A (en) | 2017-05-17 | 2017-05-17 | Sweep forward model integrated method, apparatus, storage device and face identification system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107358143A true CN107358143A (en) | 2017-11-17 |
Family
ID=60272182
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710350359.5A Pending CN107358143A (en) | 2017-05-17 | 2017-05-17 | Sweep forward model integrated method, apparatus, storage device and face identification system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107358143A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197660A (en) * | 2018-01-17 | 2018-06-22 | 中国科学院上海高等研究院 | Multi-model Feature fusion/system, computer readable storage medium and equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050144149A1 (en) * | 2001-12-08 | 2005-06-30 | Microsoft Corporation | Method for boosting the performance of machine-learning classifiers |
| US20070100618A1 (en) * | 2005-11-02 | 2007-05-03 | Samsung Electronics Co., Ltd. | Apparatus, method, and medium for dialogue speech recognition using topic domain detection |
| CN104348652A (en) * | 2013-08-06 | 2015-02-11 | 南京理工大学常熟研究院有限公司 | Method and device for evaluating system security based on correlation analysis |
-
2017
- 2017-05-17 CN CN201710350359.5A patent/CN107358143A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050144149A1 (en) * | 2001-12-08 | 2005-06-30 | Microsoft Corporation | Method for boosting the performance of machine-learning classifiers |
| US20070100618A1 (en) * | 2005-11-02 | 2007-05-03 | Samsung Electronics Co., Ltd. | Apparatus, method, and medium for dialogue speech recognition using topic domain detection |
| CN104348652A (en) * | 2013-08-06 | 2015-02-11 | 南京理工大学常熟研究院有限公司 | Method and device for evaluating system security based on correlation analysis |
Non-Patent Citations (1)
| Title |
|---|
| 韩建光 等: "《基于多特征子集组合分类器的企业财务困境预测》", 《系统管理学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197660A (en) * | 2018-01-17 | 2018-06-22 | 中国科学院上海高等研究院 | Multi-model Feature fusion/system, computer readable storage medium and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A simple pooling-based design for real-time salient object detection | |
| CN106780612B (en) | Object detecting method and device in a kind of image | |
| CN110807385B (en) | Target detection method, target detection device, electronic equipment and storage medium | |
| CN103065134A (en) | Fingerprint identification device and method with prompt information | |
| CN110427970A (en) | Image classification method, device, computer equipment and storage medium | |
| CN106295502A (en) | A kind of method for detecting human face and device | |
| CN104463141A (en) | Fingerprint template input method and device based on intelligent card | |
| CN109828919A (en) | Test report automatic generation method, device, computer equipment and storage medium | |
| CN110083531A (en) | It improves the shared multi-goal path coverage test method of individual information and realizes system | |
| CN109858476A (en) | The extending method and electronic equipment of label | |
| CN108875795A (en) | A kind of feature selecting algorithm based on Relief and mutual information | |
| CN109740549A (en) | SAR image object detection system and method based on semi-supervised CNN | |
| KR20190114384A (en) | Apparatus and method for skin lesion diagnosis based on neural network | |
| CN106331848B (en) | The recognition methods of panoramic video and equipment play video method and equipment | |
| CN109710240A (en) | Flow chart decomposition method and system | |
| CN112419306A (en) | Lung nodule detection method based on NAS-FPN | |
| CN107358143A (en) | Sweep forward model integrated method, apparatus, storage device and face identification system | |
| CN105654095A (en) | Feature selection method and device | |
| CN108595312A (en) | A kind of automatic performance method and device of modelling customer behavior | |
| CN114942947A (en) | Follow-up visit data processing method and system based on intelligent medical treatment | |
| CN104915961A (en) | Lump image region display method and system based on mammary X-ray image | |
| CN107330358A (en) | Sweep backward model integrated method and device, storage device and face identification system | |
| CN104504399A (en) | Multi-spectrum data supervised classification method based on relevant linear information entropy | |
| Guo et al. | Pathological Detection of Micro and Fuzzy Gastric Cancer Cells Based on Deep Learning. | |
| CN102117380B (en) | System and method for simplification of matrix based Boosting algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171117 |