CN109815789A - Real-time multiple dimensioned method for detecting human face and system and relevant device on CPU - Google Patents
Real-time multiple dimensioned method for detecting human face and system and relevant device on CPU Download PDFInfo
- Publication number
- CN109815789A CN109815789A CN201811514183.3A CN201811514183A CN109815789A CN 109815789 A CN109815789 A CN 109815789A CN 201811514183 A CN201811514183 A CN 201811514183A CN 109815789 A CN109815789 A CN 109815789A
- Authority
- CN
- China
- Prior art keywords
- detected
- characteristic pattern
- scale
- face
- characteristic
- 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
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000001514 detection method Methods 0.000 claims abstract description 28
- 230000001629 suppression Effects 0.000 claims abstract description 15
- 239000000284 extract Substances 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 6
- 230000004927 fusion Effects 0.000 claims description 28
- 230000004913 activation Effects 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 6
- 238000007689 inspection Methods 0.000 claims description 3
- 238000010586 diagram Methods 0.000 description 8
- 239000000203 mixture Substances 0.000 description 8
- 230000006870 function Effects 0.000 description 5
- 238000003475 lamination Methods 0.000 description 3
- 238000000605 extraction Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000465 moulding Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to human face detection tech fields, and in particular to one kind multiple dimensioned method for detecting human face real-time on CPU and system and relevant device, it is therefore intended that the hardware cost for reducing Face datection improves the speed and accuracy of Face datection.Face detection system of the invention includes: characteristic extracting module, multiple scale detecting module and non-maxima suppression module.Wherein, characteristic extracting module, which is configured that from image to be detected, extracts key feature, obtains multiple dimensioned characteristic pattern to be detected；Multiple scale detecting module is configured that according to multiple dimensioned characteristic pattern to be detected prediction face score and corresponding position；Non-maxima suppression module, which is configured that, carries out non-maxima suppression according to face score, to obtain testing result.Present invention reduces the hardware costs of Face datection, improve the speed and accuracy of multiple dimensioned Face datection, and the higher multiple dimensioned Face datection function of accuracy rate can be realized on CPU, can then be applied on the platforms such as mobile phone.
Description
Technical field
The present invention relates to human face detection tech fields, and in particular to a kind of multiple dimensioned method for detecting human face real-time on CPU
With system and relevant device.
Background technique
With the development of computer vision, human face detection tech has been made significant headway, and is obtained in reality
It is widely applied very much.But human face detection tech can still encounter very big difficulty, under unconstrained condition, because image background is multiple
Miscellaneous, face scale and posture multiplicity etc. need to guarantee also to guarantee detection accuracy while detecting speed using GPU etc.
The calculating equipment of high-speed parallel, so that hardware cost is higher.If CPU is used only, detection speed and detection accuracy can be encountered not
The case where capable of getting both,
Therefore, it is intended that proposing a kind of new network structure, which has parameter few, the small advantage of calculation amount, Neng Goushi
On present CPU while guaranteeing precision, the multiple dimensioned Face datection of real-time perfoming.
Summary of the invention
In order to solve the above problem in the prior art, the multiple dimensioned face inspection in real time on CPU that the invention proposes one kind
Survey method and system and relevant device, reduce the hardware cost of Face datection, improve the speed of multiple dimensioned Face datection with
Accuracy.
An aspect of of the present present invention proposes a kind of multiple dimensioned face detection system real-time on CPU, comprising: feature extraction mould
Block, multiple scale detecting module and non-maxima suppression module；
The characteristic extracting module, which is configured that from image to be detected, extracts key feature, obtains multiple dimensioned to be detected
Characteristic pattern；
The multiple scale detecting module is configured that according to multiple dimensioned the characteristic pattern to be detected prediction face score and phase
The position answered；
The non-maxima suppression module, which is configured that, carries out non-maxima suppression according to the face score, to obtain
Testing result.
Preferably, the multiple dimensioned characteristic pattern to be detected includes: the characteristic pattern to be detected of the first scale, the second scale
The characteristic pattern to be detected of characteristic pattern to be detected and third scale.
Preferably, the multiple scale detecting module concrete configuration are as follows: according to the characteristic pattern to be detected of first scale, institute
The characteristic pattern to be detected of the second scale, the characteristic pattern to be detected of the third scale are stated, each characteristic pattern to be detected is calculated separately
In face score and corresponding position.
Preferably, the characteristic extracting module includes: convolutional layer Conv1, the first CReLU activation primitive, pond layer
Pool1, convolutional layer Conv2, the 2nd CReLU activation primitive, pond layer Pool2, Inception1, Inception2,
Inception3, convolutional layer Conv3-1, convolutional layer Conv3-2, convolutional layer Conv4-1, convolutional layer Conv4-2, the first fusant
Module and the second fusion submodule；
Wherein,
The convolutional layer Conv1 is for tentatively extracting the local message of described image to be detected, convolution kernel size
7 × 7 are selected as, generating characteristic pattern channel is 24；
The first CReLU activation primitive is used for the output to the convolutional layer Conv1 and carries out Nonlinear Mapping；
The maximum pond layer that the pond layer Pool1 is 3 × 3, replaces present bit confidence with the optimal value in 3 × 3 ranges
Breath, keeps current point information more representative；
The convolutional layer Conv2 is used for the further processing to preceding layer characteristic information and extracts, and convolution kernel size selection 3 ×
3, generating characteristic pattern channel is 24；
The 2nd CReLU activation primitive is used for the output to the convolutional layer Conv2 and carries out Nonlinear Mapping；
The maximum pond layer that the pond layer Pool2 is 3 × 3 is used for specification current characteristic information；
The Inception1, the Inception2 and the Inception3 are used to multichannel and extract feature jointly
Information simultaneously carries out concat fusion, exports third essential characteristic figure by the Inception3；
The convolutional layer Conv3-1 be used for input feature vector carry out individual element information summary, convolution kernel size selection 1 ×
1, generating characteristic pattern channel is 128；
The convolutional layer Conv3-2 generates characteristic pattern for exporting the second essential characteristic figure, convolution kernel size selection 3 × 3
Channel is 256；
The convolutional layer Conv4-1 be used for input feature vector carry out individual element information summary, convolution kernel size selection 1 ×
1, generating characteristic pattern channel is 128；
The convolutional layer Conv4-2 exports first ruler for extracting to the Deep Semantics information of input picture
The characteristic pattern to be detected of degree, convolution kernel size selection 3 × 3, generating characteristic pattern channel is 256；
The first fusion submodule is used for the feature to be detected of the second essential characteristic figure and first scale
The deconvolution feature of figure carries out the characteristic pattern to be detected that fusion generates second scale；
The second fusion submodule is used for the feature to be detected of the third essential characteristic figure and second scale
The deconvolution feature of figure carries out the characteristic pattern to be detected that fusion generates the third scale.
Preferably, the resolution ratio of described image to be detected is 1024*1024；
Correspondingly,
The characteristic pattern to be detected of first scale is used to detect the face greater than 140 pixels；
The characteristic pattern to be detected of second scale is used to detect more than or equal to 40 pixels and be less than or equal to the people of 140 pixels
Face；
The characteristic pattern to be detected of the third scale is for detecting the face less than 40 pixels.
The second aspect of the present invention proposes a kind of multiple dimensioned method for detecting human face real-time on CPU, based on recited above
The real-time multiple dimensioned face detection system on CPU, comprising the following steps:
Key feature is extracted from image to be detected, obtains multiple dimensioned characteristic pattern to be detected；
According to the multiple dimensioned characteristic pattern to be detected prediction face score and corresponding position；
Non-maxima suppression is carried out according to the face score, to obtain testing result.
Preferably, the multiple dimensioned characteristic pattern to be detected includes: the characteristic pattern to be detected of the first scale, the second scale
The characteristic pattern to be detected of characteristic pattern to be detected and third scale.
Preferably, " according to the multiple dimensioned characteristic pattern to be detected prediction face score and corresponding position " the step of, wraps
It includes:
According to the characteristic pattern to be detected of first scale, the characteristic pattern to be detected of second scale, the third ruler
The characteristic pattern to be detected of degree, calculates separately the face score in each characteristic pattern to be detected and corresponding position.
The third aspect of the present invention proposes a kind of storage equipment, wherein being stored with program, described program is suitable for by processor
It loads and executes, to realize multiple dimensioned method for detecting human face real-time on CPU recited above.
The fourth aspect of the present invention proposes a kind of processing equipment, comprising:
Processor is adapted for carrying out program；
Memory is suitable for storing the program；
It is characterized in that, described program is suitable for being loaded and being executed by the processor, it is recited above on CPU to realize
Real-time multiple dimensioned method for detecting human face.
Compared with the immediate prior art, the invention has the following beneficial effects:
Face detection system and method for the invention, reduces the hardware cost of Face datection, improves multiple dimensioned face
The speed and accuracy of detection can realize the higher multiple dimensioned Face datection function of accuracy rate on CPU, can then apply
On the platforms such as mobile phone.
Detailed description of the invention
Fig. 1 is the main composition schematic diagram of the real-time multiple dimensioned face detection system on CPU in the embodiment of the present invention；
Fig. 2 is the specific composition schematic diagram of the real-time multiple dimensioned face detection system on CPU in the embodiment of the present invention；
Fig. 3 is the composition schematic diagram of the first fusion submodule in the embodiment of the present invention；
Fig. 4 is the key step schematic diagram of the real-time multiple dimensioned method for detecting human face on CPU in the embodiment of the present invention.
Specific embodiment
The preferred embodiment of the present invention described with reference to the accompanying drawings.It will be apparent to a skilled person that this
The a little technical principles of embodiment for explaining only the invention, it is not intended that limit the scope of the invention.
Face datection is to detect and position all faces in input picture, is the important direction of computer vision.This
Method uses single full convolutional neural networks, and while guaranteeing precision, multiple dimensioned face inspection in real time can be realized on CPU
It surveys.Several layers of basic networks for being known as present network architecture (include: convolutional layer before network in the face detection system of the present embodiment
Conv1, the first CReLU activation primitive, pond layer Pool1, convolutional layer Conv2, the 2nd CReLU activation primitive, pond layer
Pool2), it is basic structure for image classification.Network it is rear it is several layers of (include: Inception1, Inception2,
Inception3, convolutional layer Conv3-1, convolutional layer Conv3-2, convolutional layer Conv4-1, convolutional layer Conv4-2) constitute similar figure
As pyramidal structure, multiple dimensioned Face datection is realized.Present networks are increased also in two fusion submodules using warp lamination
A large amount of contextual information, the final precision for promoting Face datection, the especially detection accuracy of small size face.
Fig. 1 is the main composition schematic diagram of multiple dimensioned face detection system embodiment real-time on CPU of the invention.Such as figure
Shown in 1, the face detection system 1 of the present embodiment includes: characteristic extracting module 10, multiple scale detecting module 20 and the suppression of non-maximum
Molding block 30.
Wherein, characteristic extracting module 10, which is configured that from image to be detected, extracts key feature, obtains multiple dimensioned to be checked
Survey characteristic pattern；Multiple scale detecting module 20 is configured that according to multiple dimensioned characteristic pattern to be detected prediction face score and corresponding
Position；Non-maxima suppression module 30, which is configured that, carries out non-maxima suppression according to face score, to obtain testing result.
In the present embodiment, multiple dimensioned characteristic pattern to be detected includes: the characteristic pattern to be detected of the first scale, the second scale
The characteristic pattern to be detected of characteristic pattern to be detected and third scale.
Specifically, multiple scale detecting module 20 is configured that the characteristic pattern to be detected according to the first scale, in the present embodiment
The characteristic pattern to be detected of the characteristic patterns to be detected of two scales, third scale, calculates separately the face in each characteristic pattern to be detected
Score and corresponding position.
Fig. 2 is the specific composition schematic diagram of face detection system in the present embodiment.As shown in Fig. 2, feature in the present embodiment
Extraction module 10 includes: convolutional layer Conv1, the first CReLU activation primitive, pond layer Pool1, convolutional layer Conv2, second
CReLU activation primitive, pond layer Pool2, Inception1, Inception2, Inception3, convolutional layer Conv3-1, convolution
Layer Conv3-2, convolutional layer Conv4-1, convolutional layer Conv4-2, the first fusion submodule and the second fusion submodule.
Wherein, convolutional layer Conv1 is for tentatively extracting the local message of image to be detected, the selection of convolution kernel size
It is 7 × 7, generating characteristic pattern channel is 24；It is non-linear that first CReLU activation primitive is used for the output progress to convolutional layer Conv1
Mapping；The maximum pond layer that pond layer Pool1 is 3 × 3, replaces current location information with the optimal value in 3 × 3 ranges, makes to work as
Preceding information is more representative；Convolutional layer Conv2 is used for the further processing to preceding layer characteristic information and extracts, convolution kernel size
Selection 3 × 3, generating characteristic pattern channel is 24；It is non-thread that 2nd CReLU activation primitive is used for the output progress to convolutional layer Conv2
Property mapping；The maximum pond layer that pond layer Pool2 is 3 × 3 is used for specification current characteristic information；Inception1,
Inception2 and Inception3 is used to multichannel and extracts characteristic information jointly and carry out concat fusion, by
Inception3 exports third essential characteristic figure；Convolutional layer Conv3-1 is used to carry out individual element information summary to input feature vector,
Convolution kernel size selection 1 × 1, generating characteristic pattern channel is 128；Convolutional layer Conv3-2 is for exporting the second essential characteristic figure, volume
Product core size selection 3 × 3, generating characteristic pattern channel is 256；Convolutional layer Conv4-1 is used to carry out individual element to input feature vector
Information summary, convolution kernel size selection 1 × 1, generating characteristic pattern channel is 128；Convolutional layer Conv4-2 is used for input picture
Deep Semantics information extracts, and exports the characteristic pattern to be detected of the first scale, and convolution kernel size selection 3 × 3 generates characteristic pattern
Channel is 256；First fusion submodule is used for the deconvolution of the second essential characteristic figure and the characteristic pattern to be detected of the first scale
Feature carries out the characteristic pattern to be detected that fusion generates the second scale；Second fusion submodule is used for third essential characteristic figure and the
The deconvolution feature of the characteristic pattern to be detected of two scales carries out the characteristic pattern to be detected that fusion generates third scale.
The composition for illustrating the first fusion submodule and the second fusion submodule by taking the first fusion submodule as an example below, should
Fusion submodule can integrate the information of shallow-layer characteristic pattern and warp lamination, merge submodule input be current characteristic pattern and
By the characteristic pattern for the identical dimensional that deconvolution obtains, new characteristic pattern to be detected then is constituted by adduction fusion.Fig. 3 is this
The composition schematic diagram of first fusion submodule in embodiment.As shown in figure 3, the first of the present embodiment merges submodule for the first ruler
After the characteristic pattern to be detected of degree successively passes through deconvolution, convolution, BN (Batch Normalization, batch normalize) processing
Input Eltw Product (is also broadcast mul, shallow-layer and the characteristic pattern of deep layer is done multiplication fortune on corresponding channel
Calculate) in layer, the second essential characteristic figure is also inputted in Product layers of Eltw after convolution, BN, convolution, BN processing, finally
By the characteristic pattern to be detected of Product layers of Eltw fused second scale of output.Expanded in the present embodiment using warp lamination
Contextual information.
In the present embodiment, the resolution ratio of image to be detected is 1024*1024；Correspondingly, the feature to be detected of the first scale
Figure is for detecting the face greater than 140 pixels；The characteristic pattern to be detected of second scale is more than or equal to 40 pixels and small for detecting
In the face for being equal to 140 pixels；The characteristic pattern to be detected of third scale is for detecting the face less than 40 pixels.
Multiple dimensioned face detection system carries out module division in real time on CPU in the application, it is only for more preferable geographical
Function involved in technical solution of the present invention is solved, should not be construed as limiting of its scope.
Based on the same technical idea with above-mentioned face detection system, the present invention also proposes a kind of reality of method for detecting human face
Apply example.Fig. 4 is the key step schematic diagram of multiple dimensioned method for detecting human face embodiment real-time on CPU of the invention.Such as Fig. 4 institute
Show, the face detection system of the present embodiment includes the following steps S1-S3:
Step S1, extracts key feature from image to be detected, obtains multiple dimensioned characteristic pattern to be detected.Here more rulers
The characteristic pattern to be detected of degree includes: the characteristic pattern to be detected and third scale of the characteristic pattern to be detected of the first scale, the second scale
Characteristic pattern to be detected.
Step S2, according to multiple dimensioned characteristic pattern to be detected prediction face score and corresponding position.Specifically:
According to the spy to be detected of the characteristic pattern to be detected of the first scale, the characteristic pattern to be detected of the second scale, third scale
Sign figure, calculates separately the face score in each characteristic pattern to be detected and corresponding position.
Step S3 carries out non-maxima suppression according to face score, to obtain testing result.
Although each step is described in the way of above-mentioned precedence in above-described embodiment, this field
Technical staff is appreciated that the effect in order to realize the present embodiment, executes between different steps not necessarily in such order,
It (parallel) execution simultaneously or can be executed with reverse order, these simple variations all protection scope of the present invention it
It is interior.
Based on above-mentioned method for detecting human face, the present invention also proposes a kind of embodiment for storing equipment, wherein it is stored with program,
Described program is suitable for being loaded and being executed by processor, to realize multiple dimensioned method for detecting human face real-time on CPU recited above.
Further, the present invention also proposes a kind of embodiment of processing equipment, comprising: processor and memory.Wherein, locate
Reason device is adapted for carrying out program, and memory is suitable for storing the program, and described program is suitable for being loaded and being executed by the processor, with reality
Existing multiple dimensioned method for detecting human face real-time on CPU recited above.
Those skilled in the art should be able to recognize that, side described in conjunction with the examples disclosed in the embodiments of the present disclosure
Method step, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate electronic hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is executed actually with electronic hardware or software mode, specific application and design constraint depending on technical solution.
Those skilled in the art can use different methods to achieve the described function each specific application, but this reality
Now it should not be considered as beyond the scope of the present invention.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this field
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.Without departing from this
Under the premise of the principle of invention, those skilled in the art can make equivalent change or replacement to the relevant technologies feature, these
Technical solution after change or replacement will fall within the scope of protection of the present invention.
Claims (10)
1. a kind of multiple dimensioned face detection system real-time on CPU characterized by comprising characteristic extracting module, multiple dimensioned inspection
Survey module and non-maxima suppression module；
The characteristic extracting module, which is configured that from image to be detected, extracts key feature, obtains multiple dimensioned feature to be detected
Figure；
The multiple scale detecting module is configured that according to the multiple dimensioned characteristic pattern to be detected prediction face score and corresponding
Position；
The non-maxima suppression module, which is configured that, carries out non-maxima suppression according to the face score, to be detected
As a result.
2. face detection system according to claim 1, which is characterized in that the multiple dimensioned characteristic pattern packet to be detected
It includes: the characteristic pattern to be detected of the characteristic pattern to be detected of the first scale, the characteristic pattern to be detected of the second scale and third scale.
3. face detection system according to claim 2, which is characterized in that the multiple scale detecting module concrete configuration
Are as follows: according to the characteristic pattern to be detected of first scale, the characteristic pattern to be detected of second scale, the third scale to
Characteristic pattern is detected, the face score in each characteristic pattern to be detected and corresponding position are calculated separately.
4. face detection system according to claim 3, which is characterized in that the characteristic extracting module includes: convolutional layer
Conv1, the first CReLU activation primitive, pond layer Pool1, convolutional layer Conv2, the 2nd CReLU activation primitive, pond layer
Pool2, Inception1, Inception2, Inception3, convolutional layer Conv3-1, convolutional layer Conv3-2, convolutional layer
Conv4-1, convolutional layer Conv4-2, the first fusion submodule and the second fusion submodule；
Wherein,
The convolutional layer Conv1 is for tentatively extracting the local message of described image to be detected, the selection of convolution kernel size
It is 7 × 7, generating characteristic pattern channel is 24；
The first CReLU activation primitive is used for the output to the convolutional layer Conv1 and carries out Nonlinear Mapping；
The maximum pond layer that the pond layer Pool1 is 3 × 3, replaces current location information with the optimal value in 3 × 3 ranges, makes
Current point information is more representative；
The convolutional layer Conv2 is used for the further processing to preceding layer characteristic information and extracts, and convolution kernel size selection 3 × 3 is raw
It is 24 at characteristic pattern channel；
The 2nd CReLU activation primitive is used for the output to the convolutional layer Conv2 and carries out Nonlinear Mapping；
The maximum pond layer that the pond layer Pool2 is 3 × 3 is used for specification current characteristic information；
The Inception1, the Inception2 and the Inception3 are used to multichannel and extract characteristic information jointly
And concat fusion is carried out, third essential characteristic figure is exported by the Inception3；
The convolutional layer Conv3-1 is used to carry out input feature vector individual element information summary, and convolution kernel size selection 1 × 1 is raw
It is 128 at characteristic pattern channel；
The convolutional layer Conv3-2 generates characteristic pattern channel for exporting the second essential characteristic figure, convolution kernel size selection 3 × 3
It is 256；
The convolutional layer Conv4-1 is used to carry out input feature vector individual element information summary, and convolution kernel size selection 1 × 1 is raw
It is 128 at characteristic pattern channel；
The convolutional layer Conv4-2 exports first scale for extracting to the Deep Semantics information of input picture
Characteristic pattern to be detected, convolution kernel size selection 3 × 3, generating characteristic pattern channel is 256；
The first fusion submodule is used for the characteristic pattern to be detected of the second essential characteristic figure and first scale
Deconvolution feature carries out the characteristic pattern to be detected that fusion generates second scale；
The second fusion submodule is used for the characteristic pattern to be detected of the third essential characteristic figure and second scale
Deconvolution feature carries out the characteristic pattern to be detected that fusion generates the third scale.
5. the face detection system according to any one of claim 2-4, characteristic value are, described image to be detected
Resolution ratio is 1024*1024；
Correspondingly,
The characteristic pattern to be detected of first scale is used to detect the face greater than 140 pixels；
The characteristic pattern to be detected of second scale is used to detect more than or equal to 40 pixels and be less than or equal to the face of 140 pixels；
The characteristic pattern to be detected of the third scale is for detecting the face less than 40 pixels.
6. a kind of multiple dimensioned method for detecting human face real-time on CPU, which is characterized in that based on any one of claim 1-5 institute
The real-time multiple dimensioned face detection system on CPU stated, comprising the following steps:
Key feature is extracted from image to be detected, obtains multiple dimensioned characteristic pattern to be detected；
According to the multiple dimensioned characteristic pattern to be detected prediction face score and corresponding position；
Non-maxima suppression is carried out according to the face score, to obtain testing result.
7. method for detecting human face according to claim 6, which is characterized in that the multiple dimensioned characteristic pattern packet to be detected
It includes: the characteristic pattern to be detected of the characteristic pattern to be detected of the first scale, the characteristic pattern to be detected of the second scale and third scale.
8. method for detecting human face according to claim 7, which is characterized in that " according to the multiple dimensioned feature to be detected
Figure prediction face score and corresponding position " the step of include:
According to the characteristic pattern to be detected of first scale, the characteristic pattern to be detected of second scale, the third scale
Characteristic pattern to be detected calculates separately the face score in each characteristic pattern to be detected and corresponding position.
9. a kind of storage equipment, wherein being stored with program, which is characterized in that described program is suitable for being loaded and being executed by processor,
To realize described in any one of claim 6-8 the real-time multiple dimensioned method for detecting human face on CPU.
10. a kind of processing equipment, comprising:
Processor is adapted for carrying out program；
Memory is suitable for storing the program；
It is characterized in that, described program is suitable for being loaded and being executed by the processor, to realize any one of claim 6-8 institute
The real-time multiple dimensioned method for detecting human face on CPU stated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811514183.3A CN109815789A (en) | 2018-12-11 | 2018-12-11 | Real-time multiple dimensioned method for detecting human face and system and relevant device on CPU |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811514183.3A CN109815789A (en) | 2018-12-11 | 2018-12-11 | Real-time multiple dimensioned method for detecting human face and system and relevant device on CPU |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109815789A true CN109815789A (en) | 2019-05-28 |
Family
ID=66602838
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811514183.3A Pending CN109815789A (en) | 2018-12-11 | 2018-12-11 | Real-time multiple dimensioned method for detecting human face and system and relevant device on CPU |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109815789A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334602A (en) * | 2019-06-06 | 2019-10-15 | 武汉市公安局视频侦查支队 | A kind of people flow rate statistical method based on convolutional neural networks |
| CN111062324A (en) * | 2019-12-17 | 2020-04-24 | 上海眼控科技股份有限公司 | Face detection method and device, computer equipment and storage medium |
| CN111563466A (en) * | 2020-05-12 | 2020-08-21 | Oppo广东移动通信有限公司 | Face detection method and related product |
| CN112446247A (en) * | 2019-08-30 | 2021-03-05 | 北京大学 | Low-illumination face detection method based on multi-feature fusion and low-illumination face detection network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107527029A (en) * | 2017-08-18 | 2017-12-29 | 卫晨 | A kind of improved Faster R CNN method for detecting human face |
| US20180068198A1 (en) * | 2016-09-06 | 2018-03-08 | Carnegie Mellon University | Methods and Software for Detecting Objects in an Image Using Contextual Multiscale Fast Region-Based Convolutional Neural Network |
| US20180096457A1 (en) * | 2016-09-08 | 2018-04-05 | Carnegie Mellon University | Methods and Software For Detecting Objects in Images Using a Multiscale Fast Region-Based Convolutional Neural Network |
| CN108509978A (en) * | 2018-02-28 | 2018-09-07 | 中南大学 | The multi-class targets detection method and model of multi-stage characteristics fusion based on CNN |
-
2018
- 2018-12-11 CN CN201811514183.3A patent/CN109815789A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180068198A1 (en) * | 2016-09-06 | 2018-03-08 | Carnegie Mellon University | Methods and Software for Detecting Objects in an Image Using Contextual Multiscale Fast Region-Based Convolutional Neural Network |
| US20180096457A1 (en) * | 2016-09-08 | 2018-04-05 | Carnegie Mellon University | Methods and Software For Detecting Objects in Images Using a Multiscale Fast Region-Based Convolutional Neural Network |
| CN107527029A (en) * | 2017-08-18 | 2017-12-29 | 卫晨 | A kind of improved Faster R CNN method for detecting human face |
| CN108509978A (en) * | 2018-02-28 | 2018-09-07 | 中南大学 | The multi-class targets detection method and model of multi-stage characteristics fusion based on CNN |
Non-Patent Citations (6)
| Title |
|---|
| QIAOSONG CHEN ET.AL: "A multi-scale fusion convolutional neural network for face detection", 《2017 IEEE INTERNATIONAL CONFERENCE ON SYSTEMS, MAN, AND CYBERNETICS (SMC)》 * |
| XUDONG SUN ET.AL: "Face detection using deep learning An improved faster RCNN approach", 《NEUROCOMPUTING》 * |
| ZHENHENG YANG ET.AL: "A multi-scale cascade fully convolutional network face detector", 《2016 23RD INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR)》 * |
| 刘宏哲: "基于单一神经网络的多尺度人脸检测", 《电子与信息学报》 * |
| 朱倩: "目标快速检测算法的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王哲峰: "移动端目标检测***的设计与实现", 《万方学位论文全文数据库》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110334602A (en) * | 2019-06-06 | 2019-10-15 | 武汉市公安局视频侦查支队 | A kind of people flow rate statistical method based on convolutional neural networks |
| CN110334602B (en) * | 2019-06-06 | 2021-10-26 | 武汉市公安局视频侦查支队 | People flow statistical method based on convolutional neural network |
| CN112446247A (en) * | 2019-08-30 | 2021-03-05 | 北京大学 | Low-illumination face detection method based on multi-feature fusion and low-illumination face detection network |
| CN112446247B (en) * | 2019-08-30 | 2022-11-15 | 北京大学 | Low-illumination face detection method based on multi-feature fusion and low-illumination face detection network |
| CN111062324A (en) * | 2019-12-17 | 2020-04-24 | 上海眼控科技股份有限公司 | Face detection method and device, computer equipment and storage medium |
| CN111563466A (en) * | 2020-05-12 | 2020-08-21 | Oppo广东移动通信有限公司 | Face detection method and related product |
| CN111563466B (en) * | 2020-05-12 | 2023-10-10 | Oppo广东移动通信有限公司 | Face detection method and related product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109508681B (en) | Method and device for generating human body key point detection model | |
| CN109815789A (en) | Real-time multiple dimensioned method for detecting human face and system and relevant device on CPU | |
| CN108229343B (en) | Target object key point detection method, deep learning neural network and device | |
| CN108460362B (en) | System and method for detecting human body part | |
| CN110866509B (en) | Action recognition method, device, computer storage medium and computer equipment | |
| CN108416327A (en) | A kind of object detection method, device, computer equipment and readable storage medium storing program for executing | |
| CN111881804B (en) | Posture estimation model training method, system, medium and terminal based on joint training | |
| CN111209811B (en) | Method and system for detecting eyeball attention position in real time | |
| CN107220643A (en) | The Traffic Sign Recognition System of deep learning model based on neurological network | |
| CN113111767A (en) | Fall detection method based on deep learning 3D posture assessment | |
| CN113159200B (en) | Object analysis method, device and storage medium | |
| CN108376234B (en) | Emotion recognition system and method for video image | |
| CN110503083A (en) | A kind of critical point detection method, apparatus and electronic equipment | |
| CN113807361A (en) | Neural network, target detection method, neural network training method and related products | |
| CN114120361A (en) | Crowd counting and positioning method based on coding and decoding structure | |
| CN111623797B (en) | Step number measuring method based on deep learning | |
| CN116310757A (en) | Multitasking real-time smoke detection method | |
| Devyatkin et al. | Neural network traffic signs detection system development | |
| Gao et al. | Safety helmet detection based on YOLOV4-M | |
| CN115223181A (en) | Text detection-based method and device for recognizing characters of seal of report material | |
| CN114399628A (en) | Insulator high-efficiency detection system under complex space environment | |
| KR20180012638A (en) | Method and apparatus for detecting object in vision recognition with aggregate channel features | |
| CN110889894A (en) | Three-dimensional face reconstruction method and device and terminal equipment | |
| CN112712015B (en) | Human body key point identification method and device, intelligent terminal and storage medium | |
| CN115147508B (en) | Training of clothing generation model and method and device for generating clothing image |
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: 20190528 |