CN106339695A - Face similarity detection method, device and terminal - Google Patents
Face similarity detection method, device and terminal Download PDFInfo
- Publication number
- CN106339695A CN106339695A CN201610835787.2A CN201610835787A CN106339695A CN 106339695 A CN106339695 A CN 106339695A CN 201610835787 A CN201610835787 A CN 201610835787A CN 106339695 A CN106339695 A CN 106339695A
- Authority
- CN
- China
- Prior art keywords
- picture
- face
- resolution
- convolutional neural
- neural networks
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 93
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 100
- 238000000034 method Methods 0.000 claims abstract description 65
- 230000001537 neural effect Effects 0.000 claims 1
- 230000008569 process Effects 0.000 abstract description 25
- 230000006870 function Effects 0.000 description 11
- 238000004891 communication Methods 0.000 description 10
- 238000012545 processing Methods 0.000 description 9
- 238000007689 inspection Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 230000004927 fusion Effects 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 5
- 230000005236 sound signal Effects 0.000 description 4
- 230000000712 assembly Effects 0.000 description 3
- 238000000429 assembly Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000001133 acceleration Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 230000007935 neutral effect Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 241000209140 Triticum Species 0.000 description 1
- 235000021307 Triticum Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
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
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/167—Detection; Localisation; Normalisation using comparisons between temporally consecutive images
-
- 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
- G06V40/168—Feature extraction; Face representation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a face similarity detection method, a device and a terminal. The method comprises a step of using a first convolutional neural network to carry out face detection on a first image and obtaining the first face characteristic of the first image, wherein the first face characteristic is an output characteristic of a first preset layer in the first convolutional neural network of the first image in a first resolution, a step of using the first convolutional neural network to carry out face detection on a second image in the first resolution, and obtaining the second face characteristic of the first preset layer in the first convolutional neural network of the second image in the first resolution, a step of judging whether the similarity between the first face characteristic and the second face characteristic is larger than a preset threshold value or not, and determining the first image and the second image as face similarity images if so. The face similarity detection is realized in the face detection process, and the efficiency of the face similarity detection is improved greatly.
Description
Technical field
It relates to image processing field, particularly to a kind of human face similarity detection method, device and terminal.
Background technology
In image processing field it is sometimes desirable to carry out human face similarity detection to two width pictures, for example, user's using terminal is clapped
Multiple photos are taken the photograph, the photo array of same people can out be carried out photo cluster by human face similarity detection by terminal.
In correlation technique, when two width pictures are carried out with human face similarity detection, first respectively using convolutional neural networks
(convolutional neural network, abbreviation cnn) carries out Face datection to two width pictures, obtains in two width pictures
Face characteristic, and then, then approx imately-detecting is carried out to the face characteristic of two width in figures based on specific algorithm.
Content of the invention
The embodiment of the present disclosure provides a kind of human face similarity detection method, device and terminal.Described technical scheme is as follows:
According to the embodiment of the present disclosure in a first aspect, providing a kind of human face similarity detection method, the method includes:
Using the first convolutional neural networks, the first picture is carried out with Face datection, obtain the first face of described first picture
Feature, wherein, described first face is characterized as described first picture under first resolution, in described first convolutional neural networks
In the first default layer output characteristic；
Using described first convolutional neural networks, Face datection is carried out under described first resolution to second picture, obtain
Under described first resolution, in described first convolutional neural networks first presets the second face of layer to described second picture
Feature；
Judge whether described first face characteristic and the similarity of described second face characteristic are more than predetermined threshold value, if so,
Then determine that described first picture and described second picture are face similar pictures.
The technical scheme that the embodiment of the present disclosure provides can include following beneficial effect:
After complete Face datection is carried out to the first picture, only the people under a kind of resolution is carried out to second picture
Face detects, i.e. in the process, the Face datection of second picture does not simultaneously complete, but just can be according to second picture in one kind
Face datection result under resolution detects come the human face similarity to carry out the first picture and second picture, i.e. the present embodiment is realized
Carried out human face similarity detection during Face datection simultaneously, will human face similarity detection fusion in Face datection process
In, complete the Face datection of the first picture and the human face similarity inspection of the first picture and second picture using same algorithm simultaneously
Survey.And in correlation technique, Face datection and human face similarity are detected as two separate processes, human face similarity detection must be
Independently carry out after the completion of the Face datection of picture, therefore, compared to correlation technique, the present embodiment achieves Face datection and face
The fusion of approx imately-detecting, thus greatly improve the efficiency of human face similarity detection.
Further, described first convolutional neural networks are full convolutional neural networks fcn.
Further, described using described first convolutional neural networks, second picture is carried out under described first resolution
Face datection, obtains described second picture under described first resolution, first in described first convolutional neural networks is pre-
If before the second face characteristic of layer, also including:
Judge the shooting time of described second picture and the difference of the shooting time of described first picture whether less than default
Threshold value, if not it is determined that described first picture and described second picture are not face similar pictures.
The technical scheme that the embodiment of the present disclosure provides can include following beneficial effect:
Primary election mode during human face similarity detection is used as by the shooting time of the first picture and second picture, if not
Meet the condition of shooting time, then can directly determine that the first picture and second picture are not similar pictures, without holding again
The follow-up Face datection of row and face characteristic compare, and therefore, improve the efficiency of human face similarity detection further.
Further, described using described fcn, Face datection is carried out under described first resolution to second picture, obtain
Described second picture, under described first resolution, before first in described fcn presets the second face characteristic of layer, also wraps
Include:
Judge whether the length and width of described second picture is consistent respectively with the length and width of described first picture, if
No it is determined that described first picture and described second picture are not face similar pictures.
The technical scheme that the embodiment of the present disclosure provides can include following beneficial effect:
Another kind of primary election side during human face similarity detection is used as by the length and width of the first picture and second picture
Formula, if not meeting the consistent condition of length and width, can directly determine that the first picture and second picture are not similar
Picture, without executing follow-up Face datection and face characteristic comparison again, therefore, improves human face similarity detection further
Efficiency.
Further, also include:
If described first picture and described second picture are face similar pictures, according to the second people of described second picture
Face feature, determines other face similar pictures of described first picture.
The technical scheme that the embodiment of the present disclosure provides can include following beneficial effect:
After determining second picture and the first picture is face similar pictures, carrying out the human face similarity with next width picture
In this way it is no longer necessary to the participation of the first picture during detection, but only need second picture is compared with next width picture, compare
Result can be equally applicable to the first picture, it is achieved thereby that be only performed once human face similarity detection be assured that several figures
The human face similarity of piece, thus improve the efficiency of human face similarity detection further.
Further, also include:
If described first picture and described second picture are not face similar pictures:
Using described first convolutional neural networks to second picture under other resolution outside described first resolution
Carry out Face datection, obtain the face characteristic of described second picture；
Using described first convolutional neural networks, Face datection is carried out under described second resolution to the 3rd picture, obtain
Under described second resolution, in described first convolutional neural networks second presets the 3rd face of layer to described 3rd picture
Feature；
Face characteristic according to described second picture and the 3rd face characteristic of described 3rd picture, determine described second
Whether picture and described 3rd picture are face similar pictures.
The technical scheme that the embodiment of the present disclosure provides can include following beneficial effect:
After determining that second picture and the first picture are not face similar pictures, second picture is carried out with complete face inspection
Surveying, on this basis, then judging whether second picture and its next width picture are face similar pictures, thus realizing in the second figure
The approx imately-detecting of second picture and other pictures is completed while the Face datection of piece.
Further, described first resolution is the minimum resolution that the corresponding picture of described first picture scales in resolution
Rate.
Further, the in described first convolutional neural networks first default layer is last layer in described fcn network
Convolutional layer.
According to the second aspect of the embodiment of the present disclosure, provide a kind of human face similarity detection means, this device includes:
First detection module, is configured with the first convolutional neural networks and carries out Face datection to the first picture, obtains
First face characteristic of described first picture, wherein, described first face is characterized as described first picture under first resolution,
The output characteristic of the in described first convolutional neural networks first default layer；
Second detection module, is configured with described first convolutional neural networks to second picture in the described first resolution
Carry out Face datection under rate, obtain described second picture under described first resolution, in described first convolutional neural networks
The first default layer the second face characteristic；
First determining module, is configured to judge that the similarity of described first face characteristic and described second face characteristic is
No more than predetermined threshold value, if it is determined that described first picture and described second picture are face similar pictures.
Further, described first convolutional neural networks are full convolutional neural networks fcn.
Further, also include:
First judge module, be configured to judge the shooting time of described second picture and described first picture shooting when
Between difference whether be less than predetermined threshold value, if not it is determined that described first picture and described second picture are not face similar diagram
Piece.
Further, also include:
Second judge module, is configured to judge the length with described first picture for the length and width of described second picture
Consistent with whether width is distinguished, if not it is determined that described first picture and described second picture are not face similar pictures.
Further, also include:
Second determining module, is configured to when described first picture and described second picture are face similar pictures, root
According to the second face characteristic of described second picture, determine other face similar pictures of described first picture.
Further, also include:
3rd detection module, is configured to when described first picture and described second picture are not face similar pictures,
Using described first convolutional neural networks, under other resolution outside described first resolution, face is carried out to second picture
Detection, obtains the face characteristic of described second picture；
4th detection module, is configured with described first convolutional neural networks to the 3rd picture in the described second resolution
Carry out Face datection under rate, obtain described 3rd picture under described second resolution, in described first convolutional neural networks
The second default layer the 3rd face characteristic；
3rd determining module, is configured to the 3rd of face characteristic according to described second picture and described 3rd picture the
Face characteristic, determines whether described second picture and described 3rd picture are face similar pictures.
Further, described first resolution is the minimum resolution that the corresponding picture of described first picture scales in resolution
Rate.
Further, the in described first convolutional neural networks first default layer is last layer in described fcn network
Convolutional layer.
According to the third aspect of the embodiment of the present disclosure, provide a kind of terminal, this terminal includes:
Processor；
For storing the memorizer of the executable instruction of described processor；
Wherein, described processor is configured to:
Using the first convolutional neural networks, the first picture is carried out with Face datection, obtain the first face of described first picture
Feature, wherein, described first face is characterized as described first picture under first resolution, in described first convolutional neural networks
In the first default layer output characteristic；
Using described first convolutional neural networks, Face datection is carried out under described first resolution to second picture, obtain
Under described first resolution, in described first convolutional neural networks first presets the second face of layer to described second picture
Feature；
Judge whether described first face characteristic and the similarity of described second face characteristic are more than predetermined threshold value, if so,
Then determine that described first picture and described second picture are face similar pictures.
It should be appreciated that above general description and detailed description hereinafter are only exemplary and explanatory, not
The disclosure can be limited.
Brief description
Accompanying drawing herein is merged in description and constitutes the part of this specification, shows the enforcement meeting the disclosure
Example, and be used for explaining the principle of the disclosure together with description.
Fig. 1 is a kind of flow chart of the human face similarity detection method according to an exemplary embodiment；
Fig. 2 is a kind of flow chart of the human face similarity detection method according to an exemplary embodiment；
Fig. 3 is a kind of entire flow figure of the human face similarity detection method according to an exemplary embodiment；
Fig. 4 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment；
Fig. 5 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment；
Fig. 6 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment；
Fig. 7 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment；
Fig. 8 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment；
Fig. 9 is a kind of block diagram of the entity of terminal according to an exemplary embodiment；
Figure 10 is a kind of block diagram of the terminal 1300 according to an exemplary embodiment.
By above-mentioned accompanying drawing it has been shown that the clear and definite embodiment of the disclosure, hereinafter will be described in more detail.These accompanying drawings
It is not intended to limit the scope of disclosure design by any mode with word description, but by reference to specific embodiment be
Those skilled in the art illustrate the concept of the disclosure.
Specific embodiment
Here will in detail exemplary embodiment be illustrated, its example is illustrated in the accompanying drawings.Explained below is related to
During accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represent same or analogous key element.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the disclosure.On the contrary, they be only with such as appended
The example of the consistent apparatus and method of some aspects being described in detail in claims, the disclosure.
Following for disclosure device embodiment, can be used for executing method of disclosure embodiment.Real for disclosure device
Apply the details not disclosed in example, refer to method of disclosure embodiment.
Fig. 1 is a kind of flow chart of the human face similarity detection method according to an exemplary embodiment, the holding of the method
Row main body can be any equipment with picture recognition, processing function, such as mobile phone, panel computer etc..As shown in figure 1, should
Method includes:
In step s101, using the first convolutional neural networks, the first picture is carried out with Face datection, obtain the first picture
The first face characteristic, wherein, the first face is characterized as the first picture under first resolution, in the first convolutional neural networks
The first default layer output characteristic.
Wherein, above-mentioned Face datection refers to go out the position of the face in a width picture by the first convolution neural network recognization
And size information, generally, need for a width picture to zoom to multiple different resolution, and to multiple different resolutions
Picture under rate using position and the size of the first convolution neural network recognization face therein, divides in conjunction with multiple differences one by one
Under resolution, the position of obtained face and size information are analyzed, thus obtaining the accurate location of the face in a secondary picture
Information with size.
In this step, when carrying out Face datection to the first picture, refer to carry out complete Face datection to the first picture, that is,
According to above-mentioned method for detecting human face, set the picture scaling resolution of the first picture first, for example can be by the figure of the first picture
Piece scaling resolution sets gradually as 0.3,0.5,0.7,1.And then, using the first convolutional neural networks under every kind of resolution
First picture carries out Face datection, obtains the size and location information of the face in the first picture under every kind of resolution.And then,
By the face size and location information under every kind of resolution is determined with the accurate size and location letter of face in the first picture
Breath.Wherein, when Face datection being carried out to the first picture under every kind of resolution using the first convolutional neural networks, the first convolution god
Each layer through network all can export corresponding eigenvalue.In this step, after completing the Face datection of the first picture, can select
The output characteristic of a certain certain layer in the first convolutional neural networks under first resolution carries out human face similarity inspection as following
The first face characteristic during survey, wherein, the certain layer in first resolution and the first convolutional neural networks can be as needed
Flexibly it is configured.
In step s102, using the first convolutional neural networks, face inspection is carried out under first resolution to second picture
Survey, obtain second picture under first resolution, the second face characteristic of the default layer of first in the first convolutional neural networks.
In hypothesis above-mentioned steps s101, the first resolution corresponding to the first face characteristic of the first picture is resolution a,
The layer in the first convolutional neural networks corresponding to first face characteristic is layer b, then, after selecting resolution a and layer b, this
In step, using the first convolutional neural networks, Face datection can be carried out under resolution a to second picture, when completing resolution a
Under the Face datection of second picture after, then obtain second picture in this Face datection, the layer b of the first convolutional neural networks
Output characteristic value, using this output characteristic value as second picture the second face characteristic.
In step s103, judge whether above-mentioned first face characteristic and the similarity of above-mentioned second face characteristic are more than in advance
If threshold value, if it is determined that the first picture and second picture are face similar pictures.
The threshold value of face characteristic similarity can be pre-set, when getting the first of the first picture through abovementioned steps
After second face characteristic of face feature and second picture, the first face characteristic and the second face characteristic are compared, if ratio
The threshold value pre-setting is more than to the similarity going out, then can determine that the first picture and second picture are face similar pictures, no
Then it is possible to determine that the first picture and second picture are not human face similarity pictures.
In above process, after complete Face datection is carried out to the first picture, only one has been carried out to second picture
Plant the Face datection under resolution, i.e. in the process, the Face datection of second picture does not simultaneously complete, but just can basis
A kind of Face datection result under resolution for the second picture detects come the human face similarity to carry out the first picture and second picture,
That is, the present embodiment achieves and has carried out human face similarity detection during Face datection simultaneously, will human face similarity detection fusion
During Face datection, complete Face datection and the first picture and second of the first picture using same algorithm simultaneously
The human face similarity detection of picture.And in correlation technique, Face datection and human face similarity are detected as two separate processes, people
Face approx imately-detecting independently must be carried out after the completion of the Face datection of picture, and therefore, compared to correlation technique, the present embodiment is realized
Face datection and the fusion of human face similarity detection, thus greatly improve the efficiency of human face similarity detection.
As one kind preferred embodiment, above-mentioned first convolutional neural networks can be full convolutional neural networks (fully
Convolutional networks, abbreviation fcn), wherein, fcn network is by the full articulamentum in traditional convolutional neural networks
Make convolutional layer into, to form a kind of full convolution Rotating fields such that it is able to obtain more more accurate than traditional convolutional neural networks
Training result.
In the above-described embodiments, first resolution can scale minimum point in resolution for the corresponding picture of the first picture
Resolution.
For the first picture, all pictures being set are needed to scale the Face datection under resolution, that is, first
What picture was carried out is complete Face datection, and it is only necessary to the set all pictures of execution scale for second picture
The Face datection under lowest resolution in resolution, selects lowest resolution to carry out Face datection, can not only protect
Card obtains the result of follow-up human face similarity detection, meanwhile, compared to holding of the Face datection under other resolution, lowest resolution
Scanning frequency degree faster, therefore can select lowest resolution can also lift the efficiency of human face similarity detection.
In the above-described embodiments, the in the first convolutional neural networks first default layer is last layer volume in fcn network
Lamination.
Multiple convolutional layers are included in fcn network, when Face datection is carried out using fcn network to specific picture,
Each layer of fcn network all can output characteristic value, the information of the eigenvalue that layer more rearward is exported can more enrich and accurately.
Therefore, select last layer of convolutional layer as default layer, for the first picture and second picture, all select the output characteristic of this layer
As the feature compared, ensure that the result of comparison is also more accurate.
It should be noted that for actual needs, in the layer selecting to carry out aspect ratio pair it is also possible to select fcn net
Other layers in network, layer for example second from the bottom.
On the basis of above-described embodiment, the present embodiment is related to a kind of processing procedure before face characteristic compares, i.e.
Before above-mentioned steps s102, also include:
Judge whether the shooting time of second picture and the difference of the shooting time of the first picture are less than predetermined threshold value, if
No it is determined that the first picture and second picture are not face similar pictures.
User, when shooting picture, is often continuously shot several pictures in same position, to ensure to photograph more preferably
Effect.That is, several pictures of human face similarity can be formed at short notice, the common feature of these pictures except human face similarity it
Outward, shooting time is also very close to.Therefore, it is possible to be entered using these pictures dependency in time in the present embodiment
Row human face similarity judges.Specifically, before second picture is carried out with the Face datection under a kind of resolution, obtain respectively first
Take the shooting time of the first picture and second picture, if it is judged that the difference of the shooting time of the first picture and second picture surpasses
Cross certain threshold value, such as difference exceedes half a minute or 1 minute, then can directly determine that this two width picture is not human face similarity figure
Piece is it is no longer necessary to carry out the Face datection of follow-up second picture and the comparison of face characteristic.And if it is judged that the first picture
It is less than or equal to set threshold value with the difference of the shooting time of second picture, then proceed the face of follow-up second picture
Detection and the comparison of face characteristic.
In the present embodiment, primary election during human face similarity detection is used as by the shooting time of the first picture and second picture
Mode, if not meeting the condition of shooting time, can directly determine that the first picture and second picture are not similar pictures, and
Do not need to execute follow-up Face datection again and face characteristic compares, therefore, improve the efficiency of human face similarity detection further.
On the basis of above-described embodiment, the present embodiment is related to another kind of processing procedure before face characteristic compares, i.e.
Before above-mentioned steps s102, also include:
Judge whether the length and width of second picture is consistent respectively with the length and width of the first picture, if it is not, then true
Fixed first picture and second picture are not face similar pictures.
If same user or the multiple user of identical are continuously shot several pictures within a short period of time, these pictures
Length and width should be consistent, therefore, in the present embodiment, before carrying out Face datection to second picture, first determines whether
Whether the first picture is consistent respectively with the length and width of second picture, i.e. whether the length of the first picture is equal to second picture
Length, whether the width of the first picture be equal to the width of second picture, if having one to be unsatisfactory in this two conditions so that it may
To determine that the first picture and second picture are not face similar pictures；If this two conditions all meet, proceed follow-up
The Face datection of second picture and face characteristic comparison.
In the present embodiment, it is used as another during human face similarity detection by the length and width of the first picture and second picture
A kind of primary election mode, if not meeting the consistent condition of length and width, can directly determine the first picture and the second figure
Piece is not similar pictures, without executing follow-up Face datection and face characteristic comparison again, therefore, improves people further
The efficiency of face approx imately-detecting.
It should be noted that during actual face approx imately-detecting, the method for the present embodiment and aforementioned judgement picture are clapped
The method taking the photograph the time can be simultaneously using it is also possible to select one of which to use, when both approaches use simultaneously, this
Bright its sequencing is not limited, can flexibly be configured as needed.
On the basis of above-described embodiment, the present embodiment is related to judge that the first picture and second picture are face similar diagram
Processing procedure after piece, i.e. after above-mentioned steps s103, also includes:
If the first picture and second picture are face similar pictures, according to the second face characteristic of second picture, determine
Other face similar pictures of first picture.
That is, when determining after the first picture and second picture are face similar pictures it is possible to second picture as base
Standard is identifying other similar pictures of the first picture.Specifically it is assumed that the second face characteristic of second picture is in resolution c
Under, the output characteristic that obtained in the d layer of the first convolutional neural networks, then when the human face similarity carrying out next width picture detects
When, next width picture is narrowed down to resolution c, and using the first convolutional neural networks, next the width picture under resolution c is entered
Row Face datection, obtains the output characteristic of wherein d layer, the second feature of this output characteristic and second picture is compared,
If comparison result is more than the predetermined threshold value in above-mentioned steps s103, can determine that next width picture is the first picture and second
The human face similarity picture of picture.That is, only by the 3rd picture and second picture are compared feature it is possible to determine the 3rd picture with
First picture and the human face similarity of second picture, the like, every time only need to be by next width picture and last width figure before
Piece carries out aspect ratio to it is possible to determine the human face similarity of next width picture and all face similar pictures before.
It should be noted that it is also possible to pass through before the human face similarity detection carrying out second picture and next width picture
Aforesaid judge shooting time and/or judge the method for length and width and to carry out primary election, detailed process referring to previous embodiment,
Here is omitted.
In the present embodiment, after determining second picture and the first picture is face similar pictures, carrying out and next width figure
In this way it is no longer necessary to the participation of the first picture during the human face similarity detection of piece, but only need second picture and next width picture
Compare, the result of comparison can be equally applicable to the first picture, it is achieved thereby that be only performed once human face similarity just detecting
Can determine the human face similarity of several pictures, thus improving the efficiency of human face similarity detection further.
On the basis of above-described embodiment, the present embodiment is related to judge that the first picture and second picture are not human face similarity
Processing procedure after picture, i.e. Fig. 2 is a kind of flow process of the human face similarity detection method according to an exemplary embodiment
Figure, as shown in Fig. 2 after above-mentioned steps s103, also including:
If the first picture and second picture are not face similar pictures, execute following processes:
In step s201, using other resolutions outside first resolution to second picture of the first convolutional neural networks
Carry out Face datection under rate, obtain the face characteristic of second picture.
After determining the first picture and second picture is not face similar pictures, for the first picture it is necessary to again
Select next width picture to carry out human face similarity detection, can select to enter with the first picture according to the shooting time order of picture
The human face similarity detection process of next width picture of pedestrian's face approx imately-detecting, the first picture and next width picture and the first picture and
The human face similarity detection process of second picture is identical, is referred to previous embodiment, repeats no more in the present embodiment.The present embodiment
Relate generally to second picture and its next width picture, i.e. the human face similarity detection process of the 3rd picture.
In this step, carrying out second picture and its next width picture, that is, before the human face similarity detection of the 3rd picture, needing
To complete the complete Face datection of second picture first, the follow-up aspect ratio pair with the 3rd picture of guarantee.Due to
One picture carries out having carried out the Face datection under first resolution, therefore, this step to second picture during human face similarity detection
In rapid, it is continuing with the first convolutional neural networks and carries out face under other resolution outside first resolution of second picture
Detection, thus getting the face characteristic of second picture, that is, completes the complete Face datection of second picture, wherein, accessed
Face characteristic can be regarded as the testing result that second picture carried out obtained after complete Face datection.
In step s202, using the first convolutional neural networks, face inspection is carried out under second resolution to the 3rd picture
Survey, obtain the 3rd picture under first resolution, the 3rd face characteristic of the default layer of second in the first convolutional neural networks.
After the face characteristic getting second picture, select a resolution, i.e. second resolution, and the first volume
In long-pending neutral net one layer, i.e. the second default layer, the output in the second default layer under second resolution by second picture
Feature is as feature to be compared.
And then, using the first convolutional neural networks, under second resolution, Face datection is carried out to the 3rd picture, and obtain
3rd face characteristic of the second default layer.
It should be noted that above-mentioned second resolution can be equal to aforesaid first resolution it is also possible to aforesaid the
One resolution is different, and the above-mentioned second default layer can be equal to the aforesaid first default layer it is also possible to preset layer with aforesaid first
Different.That is, the human face similarity detection of second picture is the human face similarity detection being totally independent of the first picture, but, as
A kind of optional mode, can pre-set a resolution and default layer, then by the resolution of setting and the application of default layer
To in the detection of all of picture human face similarity, or, every width picture carries out being applied to certainly it is also possible to set during human face similarity detection
Oneself resolution and default layer.
In step s203, the face characteristic according to second picture and the 3rd face characteristic of the 3rd picture, determine
Whether two pictures and the 3rd picture are face similar pictures.
As it was previously stated, after the face characteristic of second picture can be regarded as carrying out complete Face datection to second picture
The testing result being obtained, is analyzed the face characteristic obtaining afterwards to the face characteristic under each resolution, and with
When three pictures carry out face characteristic comparison, will be special in the output of the second default layer under second resolution corresponding in this face characteristic
Levy as comparing feature, compare with the 3rd face characteristic of the 3rd picture, if comparison result is more than abovementioned steps s103
In predetermined threshold value, then can determine the 3rd picture be second picture human face similarity picture.
In the present embodiment, after determining that second picture and the first picture are not face similar pictures, second picture is carried out
Complete Face datection, on this basis, then judges whether second picture and its next width picture are face similar pictures, thus
Realize completing the approx imately-detecting of second picture and other pictures while the Face datection of second picture.
Fig. 3 is a kind of entire flow figure of the human face similarity detection method according to an exemplary embodiment, as Fig. 3 institute
Show, the complete implementation procedure of the method is:
In step s301, using the first convolutional neural networks, the first picture is carried out with Face datection, obtain the first picture
The first face characteristic, wherein, the first face is characterized as the first picture under first resolution, in the first convolutional neural networks
The first default layer output characteristic.
In step s302, judge whether the shooting time of second picture is less than with the difference of the shooting time of the first picture
Predetermined threshold value, if not it is determined that the first picture and second picture are not face similar pictures.If so, then execution step s303.
In step s303, judge whether the length and width of second picture distinguishes one with the length and width of the first picture
Cause, if not it is determined that the first picture and second picture are not face similar pictures.If so, then execution step s304.
In step s304, using the first convolutional neural networks, face inspection is carried out under first resolution to second picture
Survey, obtain second picture under first resolution, the second face characteristic of the default layer of first in the first convolutional neural networks.
In step s305, judge whether above-mentioned first face characteristic and the similarity of above-mentioned second face characteristic are more than in advance
If threshold value, if it is determined that the first picture and second picture are face similar pictures.
Following for disclosure device embodiment, can be used for executing method of disclosure embodiment.Real for disclosure device
Apply the details not disclosed in example, refer to method of disclosure embodiment.
Fig. 4 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment, as Fig. 4 institute
Show, this device includes:
First detection module 401, is configured with the first convolutional neural networks and carries out Face datection to the first picture, obtain
Take the first face characteristic of the first picture, wherein, the first face is characterized as the first picture under first resolution, in the first convolution
The output characteristic of the in neutral net first default layer.
Second detection module 402, is configured with the first convolutional neural networks to second picture under first resolution
Carry out Face datection, obtain second picture under first resolution, the of the default layer of first in the first convolutional neural networks
Two face characteristics.
First determining module 403, is configured to judge whether the similarity of the first face characteristic and the second face characteristic is big
In predetermined threshold value, if it is determined that the first picture and second picture are face similar pictures.
In another embodiment, above-mentioned first convolutional neural networks of stating are fcn network.
Fig. 5 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment, as Fig. 5 institute
Show, this device also includes:
First judge module 404, is configured to judge the shooting time of the shooting time of second picture and the first picture
Whether difference is less than predetermined threshold value, if not it is determined that the first picture and second picture are not face similar pictures.
Fig. 6 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment, as Fig. 6 institute
Show, this device also includes:
Second judge module 405, is configured to judge the length and width of second picture and the length of the first picture and width
Whether degree is consistent respectively, if not it is determined that the first picture and second picture are not face similar pictures.
Fig. 7 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment, as Fig. 7 institute
Show, this device also includes:
Second determining module 406, is configured to when the first picture and second picture are face similar pictures, according to second
Second face characteristic of picture, determines other face similar pictures of the first picture.
Fig. 8 is a kind of function structure chart of the human face similarity detection means according to an exemplary embodiment, as Fig. 8 institute
Show, this device also includes:
3rd detection module 407, is configured to when the first picture and second picture are not face similar pictures, using
One convolutional neural networks carry out Face datection under other resolution outside first resolution of second picture, obtain the second figure
The face characteristic of piece.
4th detection module 408, is configured with the first convolutional neural networks to the 3rd picture under second resolution
Carry out Face datection, obtain the 3rd picture under described second resolution, second in described first convolutional neural networks is pre-
If the 3rd face characteristic of layer；
3rd determining module 409, is configured to face characteristic according to described second picture and described 3rd picture
3rd face characteristic, determines whether described second picture and described 3rd picture are face similar pictures.
In another embodiment, above-mentioned first resolution is minimum in described first picture corresponding picture scaling resolution
Resolution.
In another embodiment, the in above-mentioned first convolutional neural networks first default layer is last in described fcn network
One layer of convolutional layer.
In sum, the human face similarity detection means that the disclosure is provided, is carrying out complete face inspection to the first picture
After survey, only the Face datection under a kind of resolution is carried out to second picture, i.e. in the process, the face of second picture
Detect and do not complete, but just the first picture can be carried out according to a kind of Face datection result under resolution for the second picture
Human face similarity detection with second picture, i.e. the present embodiment achieves and carried out human face similarity during Face datection simultaneously
Detection, will human face similarity detection fusion during Face datection, complete the first picture using same algorithm simultaneously
The human face similarity detection of Face datection and the first picture and second picture.And in correlation technique, Face datection and human face similarity
It is detected as two separate processes, human face similarity detection independently must be carried out after the completion of the Face datection of picture, therefore,
Compared to correlation technique, the present embodiment achieves Face datection and the fusion of human face similarity detection, thus greatly improving people
The efficiency of face approx imately-detecting.
With regard to the device in above-described embodiment, wherein the concrete mode of modules execution operation is in relevant the method
Embodiment in be described in detail, explanation will be not set forth in detail herein.
Fig. 9 is a kind of block diagram of the entity of terminal according to an exemplary embodiment, as shown in figure 9, this terminal bag
Include:
Memorizer 91 and processor 92.
Memorizer 91 is used for storing the executable instruction of processor 92.
Processor 92 is configured to:
Using the first convolutional neural networks, the first picture is carried out with Face datection, obtain the first face of described first picture
Feature, wherein, described first face is characterized as described first picture under first resolution, in described first convolutional neural networks
In the first default layer output characteristic；
Using described first convolutional neural networks, Face datection is carried out under described first resolution to second picture, obtain
Under described first resolution, in described first convolutional neural networks first presets the second face of layer to described second picture
Feature；
Judge whether described first face characteristic and the similarity of described second face characteristic are more than predetermined threshold value, if so,
Then determine that described first picture and described second picture are face similar pictures.
It should be appreciated that processor 92 can be central authorities processes submodule (English: central in the embodiment of above-mentioned terminal
Processing unit, referred to as: cpu), can also be other general processors, digital signal processor (English: digital
Signal processor, referred to as: dsp), special IC (English: application specific integrated
Circuit, referred to as: asic) etc..General processor can be microprocessor or this processor can also be any conventional place
Reason device etc., and aforesaid memorizer can be read only memory (English: read-only memory, abbreviation: rom), deposit at random
Access to memory (English: random access memory, referred to as: ram), flash memory, hard disk or solid state hard disc.sim
Card is also referred to as subscriber identification card, smart card, and digital mobile telephone must be loaded onto this card and can use.I.e. in computer chip
On store the information of digital mobile phone client, the content such as the key of encryption and the telephone directory of user.Real in conjunction with the disclosure
The step applying the method disclosed in example can be embodied directly in hardware processor execution and completes, or with the hardware in processor and
Software module combination execution completes.
Figure 10 is a kind of block diagram of the terminal 1300 according to an exemplary embodiment.Wherein, terminal 1300 can be
Mobile phone, computer, tablet device, personal digital assistant etc..
With reference to Figure 10, terminal 1300 can include following one or more assemblies: process assembly 1302, memorizer 1304,
Power supply module 1306, multimedia groupware 1308, audio-frequency assembly 1310, the interface 1312 of input/output (i/o), sensor cluster
1314, and communication component 1316.
The integrated operation of the usual control terminal 1300 of process assembly 1302, such as with display, call, data communication,
Camera operation and record operate associated operation.Process assembly 1302 can include one or more processors 1320 to execute
Instruction, to complete all or part of step of above-mentioned method.Additionally, process assembly 1302 can include one or more moulds
Block, is easy to the interaction between process assembly 1302 and other assemblies.For example, process assembly 1302 can include multi-media module,
To facilitate the interaction between multimedia groupware 1308 and process assembly 1302.
Memorizer 1304 is configured to store various types of data to support the operation in terminal 1300.These data
Example include in terminal 1300 operation any application program or method instruction, contact data, telephone book data,
Message, picture, video etc..Memorizer 1304 can by any kind of volatibility or non-volatile memory device or they
Combination is realized, such as static RAM (sram), Electrically Erasable Read Only Memory (eeprom), erasable can
Program read-only memory (eprom), programmable read only memory (prom), read only memory (rom), magnetic memory, flash memory
Reservoir, disk or CD.
Power supply module 1306 provides electric power for the various assemblies of terminal 1300.Power supply module 1306 can include power management
System, one or more power supplys, and other generate, manage and distribute, with for terminal 1300, the assembly that electric power is associated.
The touch-control that multimedia groupware 1308 includes one output interface of offer between described terminal 1300 and user shows
Display screen.In certain embodiments, touching display screen can include liquid crystal display (lcd) and touch panel (tp).Touch panel
Including one or more touch sensors with the gesture on sensing touch, slip and touch panel.Described touch sensor is permissible
The not only border of sensing touch or sliding action, but also the detection persistent period related to described touch or slide and pressure
Power.In certain embodiments, multimedia groupware 1308 includes a front-facing camera and/or post-positioned pick-up head.When terminal 1300
It is in operator scheme, during as screening-mode or video mode, front-facing camera and/or post-positioned pick-up head can receive the many of outside
Media data.Each front-facing camera and post-positioned pick-up head can be the optical lens system of a fixation or have focal length and light
Learn zoom capabilities.
Audio-frequency assembly 1310 is configured to output and/or input audio signal.For example, audio-frequency assembly 1310 includes a wheat
Gram wind (mic), when terminal 1300 is in operator scheme, such as call model, logging mode and speech recognition mode when, mike quilt
It is configured to receive external audio signal.The audio signal being received can be further stored in memorizer 1304 or via communication
Assembly 1316 sends.In certain embodiments, audio-frequency assembly 1310 also includes a speaker, for exports audio signal.
I/o interface 1312 is for providing interface, above-mentioned peripheral interface module between process assembly 1302 and peripheral interface module
Can be keyboard, click wheel, button etc..These buttons may include but be not limited to: home button, volume button, start button and
Locking press button.
Sensor cluster 1314 includes one or more sensors, for providing the state of various aspects to comment for terminal 1300
Estimate.For example, sensor cluster 1314 can detect/the closed mode of opening of terminal 1300, the relative localization of assembly, such as institute
State the display that assembly is terminal 1300 and keypad, sensor cluster 1314 can be with detection terminal 1300 or terminal 1,300 1
The position change of individual assembly, user is presence or absence of with what terminal 1300 contacted, terminal 1300 orientation or acceleration/deceleration and end
The temperature change at end 1300.Sensor cluster 1314 can include proximity transducer, is configured to do not having any physics
The presence of object nearby is detected during contact.Sensor cluster 1314 can also include optical sensor, such as cmos or ccd image sensing
Device, for using in imaging applications.In certain embodiments, this sensor cluster 1314 can also include acceleration sensing
Device, gyro sensor, Magnetic Sensor, pressure transducer or temperature sensor.
Communication component 1316 is configured to facilitate the communication of wired or wireless way between terminal 1300 and other equipment.Eventually
The wireless network based on communication standard, such as wifi, 2g or 3g can be accessed in end 1300, or combinations thereof.Exemplary at one
In embodiment, communication component 1316 receives related from the broadcast singal of external broadcasting management system or broadcast via broadcast channel
Information.In one exemplary embodiment, described communication component 1316 also includes near-field communication (nfc) module, to promote short distance
Communication.For example, RF identification (rfid) technology, Infrared Data Association (irda) technology, ultra broadband can be based in nfc module
(uwb) technology, bluetooth (bt) technology and other technologies are realizing.
In the exemplary embodiment, terminal 1300 can be by one or more application specific integrated circuits (asic), numeral
Signal processor (dsp), digital signal processing appts (dspd), PLD (pld), field programmable gate array
(fpga), controller, microcontroller, microprocessor or other electronic components are realized, for executing above-mentioned human face similarity detection side
Method.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instruction, example are additionally provided
As included the memorizer 1304 instructing, above-mentioned instruction can be executed by the processor 1320 of terminal 1300 to complete said method.Example
If, described non-transitorycomputer readable storage medium can be rom, random access memory (ram), cd-rom, tape, soft
Disk and optical data storage devices etc..
A kind of non-transitorycomputer readable storage medium, when the instruction in described storage medium is by the process of terminal 1300
So that terminal 1300 is able to carry out a kind of human face similarity detection method during device execution.Methods described includes:
Using the first convolutional neural networks, the first picture is carried out with Face datection, obtain the first face of described first picture
Feature, wherein, described first face is characterized as described first picture under first resolution, in described first convolutional neural networks
In the first default layer output characteristic；
Using described first convolutional neural networks, Face datection is carried out under described first resolution to second picture, obtain
Under described first resolution, in described first convolutional neural networks first presets the second face of layer to described second picture
Feature；
Judge whether described first face characteristic and the similarity of described second face characteristic are more than predetermined threshold value, if so,
Then determine that described first picture and described second picture are face similar pictures.
Those skilled in the art, after considering description and putting into practice invention disclosed herein, will readily occur to its of the disclosure
Its embodiment.The application is intended to any modification, purposes or the adaptations of the disclosure, these modifications, purposes or
Person's adaptations are followed the general principle of the disclosure and are included the undocumented common knowledge in the art of the disclosure
Or conventional techniques.Description and embodiments be considered only as exemplary, the true scope of the disclosure and spirit by following
Claims are pointed out.
It should be appreciated that the disclosure is not limited to be described above and precision architecture illustrated in the accompanying drawings, and
And various modifications and changes can carried out without departing from the scope.The scope of the present disclosure only to be limited by appending claims
System.
Claims (17)
1. a kind of human face similarity detection method is it is characterised in that include:
Using the first convolutional neural networks, the first picture is carried out with Face datection, the first face obtaining described first picture is special
Levy, wherein, described first face is characterized as described first picture under first resolution, in described first convolutional neural networks
The first default layer output characteristic；
Using described first convolutional neural networks, second picture is carried out under described first resolution with Face datection, obtain described
, under described first resolution, the second face of the in described first convolutional neural networks first default layer is special for second picture
Levy；
Judge whether described first face characteristic and the similarity of described second face characteristic are more than predetermined threshold value, if so, then true
Fixed described first picture and described second picture are face similar pictures.
2. method according to claim 1 is it is characterised in that described first convolutional neural networks are full convolutional neural networks
fcn.
3. method according to claim 1 it is characterised in that described using described first convolutional neural networks to the second figure
Piece carries out Face datection under described first resolution, obtains described second picture under described first resolution, described the
Before second face characteristic of the in one convolutional neural networks first default layer, also include:
Judge whether the shooting time of described second picture and the difference of the shooting time of described first picture are less than predetermined threshold value,
If not it is determined that described first picture and described second picture are not face similar pictures.
4. method according to claim 1 it is characterised in that described using described fcn to second picture described first
Carry out Face datection under resolution, obtain described second picture under described first resolution, first in described fcn is preset
Before second face characteristic of layer, also include:
Judge whether the length and width of described second picture is consistent respectively with the length and width of described first picture, if it is not,
Then determine that described first picture and described second picture are not face similar pictures.
5. the method according to any one of claim 1-4 is it is characterised in that also include:
If described first picture and described second picture are face similar pictures, special according to the second face of described second picture
Levy, determine other face similar pictures of described first picture.
6. the method according to any one of claim 1-4 is it is characterised in that also include:
If described first picture and described second picture are not face similar pictures:
Using described first convolutional neural networks, second picture is carried out under other resolution outside described first resolution
Face datection, obtains the face characteristic of described second picture；
Using described first convolutional neural networks, the 3rd picture is carried out under described second resolution with Face datection, obtain described
, under described second resolution, the 3rd face of the in described first convolutional neural networks second default layer is special for 3rd picture
Levy；
Face characteristic according to described second picture and the 3rd face characteristic of described 3rd picture, determine described second picture
Whether it is face similar pictures with described 3rd picture.
7. method according to claim 1 is it is characterised in that described first resolution is the corresponding figure of described first picture
Piece scales the lowest resolution in resolution.
8. method according to claim 2 is it is characterised in that the in described first convolutional neural networks first default layer is
Last layer of convolutional layer in described fcn network.
9. a kind of human face similarity detection means is it is characterised in that include:
First detection module, is configured with the first convolutional neural networks and the first picture is carried out with Face datection, obtains described
First face characteristic of the first picture, wherein, described first face is characterized as described first picture under first resolution, in institute
State the output characteristic of the first default layer in the first convolutional neural networks；
Second detection module, is configured with described first convolutional neural networks to second picture under described first resolution
Carry out Face datection, obtain described second picture under described first resolution, the in described first convolutional neural networks
Second face characteristic of one default layer；
First determining module, is configured to judge whether the similarity of described first face characteristic and described second face characteristic is big
In predetermined threshold value, if it is determined that described first picture and described second picture are face similar pictures.
10. device according to claim 9 is it is characterised in that described first convolutional neural networks are full convolutional Neural net
Network fcn.
11. devices according to claim 9 are it is characterised in that also include:
First judge module, is configured to judge the shooting time of described second picture and the shooting time of described first picture
Whether difference is less than predetermined threshold value, if not it is determined that described first picture and described second picture are not face similar pictures.
12. devices according to claim 9 are it is characterised in that also include:
Second judge module, is configured to judge the length and width of described second picture and the length of described first picture and width
Whether degree is consistent respectively, if not it is determined that described first picture and described second picture are not face similar pictures.
13. devices according to any one of claim 9-12 are it is characterised in that also include:
Second determining module, is configured to when described first picture and described second picture are face similar pictures, according to institute
State the second face characteristic of second picture, determine other face similar pictures of described first picture.
14. devices according to any one of claim 9-12 are it is characterised in that also include:
3rd detection module, is configured to, when described first picture and described second picture are not face similar pictures, use
Described first convolutional neural networks carry out Face datection to second picture under other resolution outside described first resolution,
Obtain the face characteristic of described second picture；
4th detection module, is configured with described first convolutional neural networks to the 3rd picture under described second resolution
Carry out Face datection, obtain described 3rd picture under described second resolution, the in described first convolutional neural networks
3rd face characteristic of two default layers；
3rd determining module, is configured to the 3rd face of face characteristic according to described second picture and described 3rd picture
Feature, determines whether described second picture and described 3rd picture are face similar pictures.
15. devices according to claim 9 are it is characterised in that described first resolution is that described first picture is corresponding
Picture scales the lowest resolution in resolution.
16. devices according to claim 10 are it is characterised in that in described first convolutional neural networks first presets layer
For last layer of convolutional layer in described fcn network.
A kind of 17. terminals are it is characterised in that described terminal includes:
Processor；
For storing the memorizer of the executable instruction of described processor；
Wherein, described processor is configured to:
Using the first convolutional neural networks, the first picture is carried out with Face datection, the first face obtaining described first picture is special
Levy, wherein, described first face is characterized as described first picture under first resolution, in described first convolutional neural networks
The first default layer output characteristic；
Using described first convolutional neural networks, second picture is carried out under described first resolution with Face datection, obtain described
, under described first resolution, the second face of the in described first convolutional neural networks first default layer is special for second picture
Levy；
Judge whether described first face characteristic and the similarity of described second face characteristic are more than predetermined threshold value, if so, then true
Fixed described first picture and described second picture are face similar pictures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610835787.2A CN106339695B (en) | 2016-09-20 | 2016-09-20 | Face similarity detection method, device and terminal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610835787.2A CN106339695B (en) | 2016-09-20 | 2016-09-20 | Face similarity detection method, device and terminal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106339695A true CN106339695A (en) | 2017-01-18 |
| CN106339695B CN106339695B (en) | 2019-11-15 |
Family
ID=57839051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610835787.2A Active CN106339695B (en) | 2016-09-20 | 2016-09-20 | Face similarity detection method, device and terminal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106339695B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107067030A (en) * | 2017-03-29 | 2017-08-18 | 北京小米移动软件有限公司 | The method and apparatus of similar pictures detection |
| CN107291945A (en) * | 2017-07-12 | 2017-10-24 | 上海交通大学 | The high-precision image of clothing search method and system of view-based access control model attention model |
| CN107516105A (en) * | 2017-07-20 | 2017-12-26 | 阿里巴巴集团控股有限公司 | Image processing method and device |
| CN108345847A (en) * | 2018-01-30 | 2018-07-31 | 石数字技术成都有限公司 | A kind of facial image label data generation system and method |
| CN109034119A (en) * | 2018-08-27 | 2018-12-18 | 苏州广目信息技术有限公司 | A kind of method for detecting human face of the full convolutional neural networks based on optimization |
| WO2019024636A1 (en) * | 2017-08-01 | 2019-02-07 | 广州广电运通金融电子股份有限公司 | Identity authentication method, system and apparatus |
| CN110110189A (en) * | 2018-02-01 | 2019-08-09 | 北京京东尚科信息技术有限公司 | Method and apparatus for generating information |
| CN111444899A (en) * | 2020-05-14 | 2020-07-24 | 聚好看科技股份有限公司 | Remote examination control method, server and terminal |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102222232A (en) * | 2011-06-24 | 2011-10-19 | 常州锐驰电子科技有限公司 | Multi-level rapid filtering and matching device and method for human faces |
| CN104899579A (en) * | 2015-06-29 | 2015-09-09 | 小米科技有限责任公司 | Face recognition method and face recognition device |
| CN105069426A (en) * | 2015-07-31 | 2015-11-18 | 小米科技有限责任公司 | Similar picture determining method and apparatus |
| CN105608425A (en) * | 2015-12-17 | 2016-05-25 | 小米科技有限责任公司 | Method and device for sorted storage of pictures |
| CN105631404A (en) * | 2015-12-17 | 2016-06-01 | 小米科技有限责任公司 | Method and device for clustering pictures |
-
2016
- 2016-09-20 CN CN201610835787.2A patent/CN106339695B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102222232A (en) * | 2011-06-24 | 2011-10-19 | 常州锐驰电子科技有限公司 | Multi-level rapid filtering and matching device and method for human faces |
| CN104899579A (en) * | 2015-06-29 | 2015-09-09 | 小米科技有限责任公司 | Face recognition method and face recognition device |
| CN105069426A (en) * | 2015-07-31 | 2015-11-18 | 小米科技有限责任公司 | Similar picture determining method and apparatus |
| CN105608425A (en) * | 2015-12-17 | 2016-05-25 | 小米科技有限责任公司 | Method and device for sorted storage of pictures |
| CN105631404A (en) * | 2015-12-17 | 2016-06-01 | 小米科技有限责任公司 | Method and device for clustering pictures |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107067030A (en) * | 2017-03-29 | 2017-08-18 | 北京小米移动软件有限公司 | The method and apparatus of similar pictures detection |
| CN107291945A (en) * | 2017-07-12 | 2017-10-24 | 上海交通大学 | The high-precision image of clothing search method and system of view-based access control model attention model |
| CN107516105B (en) * | 2017-07-20 | 2020-06-16 | 阿里巴巴集团控股有限公司 | Image processing method and device |
| CN107516105A (en) * | 2017-07-20 | 2017-12-26 | 阿里巴巴集团控股有限公司 | Image processing method and device |
| US11093792B2 (en) | 2017-07-20 | 2021-08-17 | Advanced New Technologies Co., Ltd. | Image processing methods and devices |
| WO2019015645A1 (en) * | 2017-07-20 | 2019-01-24 | 阿里巴巴集团控股有限公司 | Imaging processing method and device |
| US10769490B2 (en) | 2017-07-20 | 2020-09-08 | Alibaba Group Holding Limited | Image processing methods and devices |
| WO2019024636A1 (en) * | 2017-08-01 | 2019-02-07 | 广州广电运通金融电子股份有限公司 | Identity authentication method, system and apparatus |
| CN108345847A (en) * | 2018-01-30 | 2018-07-31 | 石数字技术成都有限公司 | A kind of facial image label data generation system and method |
| CN110110189A (en) * | 2018-02-01 | 2019-08-09 | 北京京东尚科信息技术有限公司 | Method and apparatus for generating information |
| CN109034119A (en) * | 2018-08-27 | 2018-12-18 | 苏州广目信息技术有限公司 | A kind of method for detecting human face of the full convolutional neural networks based on optimization |
| CN111444899A (en) * | 2020-05-14 | 2020-07-24 | 聚好看科技股份有限公司 | Remote examination control method, server and terminal |
| CN111444899B (en) * | 2020-05-14 | 2023-10-31 | 聚好看科技股份有限公司 | Remote examination control method, server and terminal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106339695B (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| TWI724736B (en) | Image processing method and device, electronic equipment, storage medium and computer program | |
| CN106339695A (en) | Face similarity detection method, device and terminal | |
| CN106454336B (en) | The method and device and terminal that detection terminal camera is blocked | |
| CN105512685B (en) | Object identification method and device | |
| US10452890B2 (en) | Fingerprint template input method, device and medium | |
| EP3121557B1 (en) | Method and apparatus for determining spatial parameter based on an image | |
| US10115019B2 (en) | Video categorization method and apparatus, and storage medium | |
| CN106355573A (en) | Target object positioning method and device in pictures | |
| CN107122679A (en) | Image processing method and device | |
| CN106951884A (en) | Gather method, device and the electronic equipment of fingerprint | |
| CN106651955A (en) | Method and device for positioning object in picture | |
| US20170185988A1 (en) | Mpayment method and apparatus | |
| CN105631403A (en) | Method and device for human face recognition | |
| CN106225764A (en) | Based on the distance-finding method of binocular camera in terminal and terminal | |
| CN106228556A (en) | Image quality analysis method and device | |
| CN107038428A (en) | Vivo identification method and device | |
| CN110717399A (en) | Face recognition method and electronic terminal equipment | |
| CN105069426A (en) | Similar picture determining method and apparatus | |
| CN105975961B (en) | The method, apparatus and terminal of recognition of face | |
| CN105279499A (en) | Age recognition method and device | |
| JP2017521742A (en) | Method and apparatus for acquiring iris image, and iris identification device | |
| CN104980662A (en) | Method for adjusting imaging style in shooting process, device thereof and imaging device | |
| CN104933700A (en) | Method and apparatus for image content recognition | |
| CN105843503A (en) | Application starting method and device as well as terminal equipment | |
| CN107169429A (en) | Vivo identification method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |