CN109544515A - A kind of trend determines method and device - Google Patents
A kind of trend determines method and device Download PDFInfo
- Publication number
- CN109544515A CN109544515A CN201811309009.5A CN201811309009A CN109544515A CN 109544515 A CN109544515 A CN 109544515A CN 201811309009 A CN201811309009 A CN 201811309009A CN 109544515 A CN109544515 A CN 109544515A
- Authority
- CN
- China
- Prior art keywords
- flaw
- facial image
- serious
- score
- serious score
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
- G06T2207/30201—Face
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The embodiment of the present invention provides a kind of trend and determines method and device, comprising: passes through camera and acquires the first facial image；First facial image is inputted into the first deep neural network, obtains the first flaw quantity and the first serious score of flaw on face in the first facial image；Obtain the mark of the first facial image；The the second flaw quantity and the second serious score of flaw on face in the second facial image are obtained, the second facial image is the facial image corresponding with the mark of storage；The variation tendency of flaw is determined according to the first flaw quantity and the second flaw quantity, perhaps the variation tendency of flaw is determined according to the first serious score and the second serious score or determines the variation tendency of flaw according to the first flaw quantity, the second flaw quantity, the first serious score and the second serious score.The embodiment of the present invention can accurately determine the situation of change of flaw on facial skin.
Description
Technical field
The present invention relates to field of computer technology, and in particular to a kind of trend determines method and device.
Background technique
The skin quality of face directly affects the beauty and ugliness of a people, and therefore, people seeking beauty, especially young woman especially infuse
The case where weight facial skin, will pass through facial skin the case where can preferably nurse facial skin.And on face
The often flaws such as president's blackhead, small pox, user can observe the situation of change of flaw on facial skin by mirror.However,
Due to the problem of face vision, memory etc., user is difficult to accurately determine the situation of change of flaw on facial skin by mirror.
Summary of the invention
The embodiment of the present invention provides a kind of trend and determines method and device, for accurately determining flaw on facial skin
Situation of change.
First aspect provides a kind of trend and determines method, comprising:
The first facial image is acquired by camera；
First facial image is inputted into the first deep neural network, the flaw on face in acquisition first facial image
The the first flaw quantity and the first serious score of defect；
Obtain the mark of first facial image；
Obtain the second flaw quantity and the second serious score of flaw on face in the second facial image, second face
Image is the facial image corresponding with the mark of storage；
The variation tendency of flaw is determined according to the first flaw quantity and the second flaw quantity, or according to described
First serious score and the second serious score determine the variation tendency of the flaw, or according to the first flaw number
Amount, the second flaw quantity, the first serious score and the second serious score determine the variation tendency of the flaw.
In one embodiment, described that first facial image is inputted into the first deep neural network, obtain described the
The first flaw quantity of flaw and the first serious score include: on face in one facial image
First facial image is inputted into the first deep neural network, institute on face in acquisition first facial image
There are the flaw probability and serious score of flaw to be determined；
In the case where the flaw probability of flaw to be determined is greater than threshold value, determine that the flaw to be determined is flaw, it is described
Flaw to be determined is any flaw to be determined in all flaws to be determined；
The quantity for counting determining flaw is the first flaw quantity of flaw on face in first facial image；
The serious score of the flaw of the determination is determined as in first facial image the first tight of flaw on face
Weight score.
In one embodiment, it is described according to the first flaw quantity, it is the second flaw quantity, described first serious
Score and the second serious score determine that the variation tendency of flaw includes:
In the case where the first flaw quantity is different from the second flaw quantity, according to the first flaw quantity
The variation tendency of flaw is determined with the second flaw quantity；
In the first flaw quantity situation identical with the second flaw quantity, according to the described first serious score
The variation tendency of flaw is determined with the described second serious score.
In one embodiment, the change that flaw is determined according to the described first serious score and the second serious score
Change trend includes:
Calculate the first weighted average of the described first serious score；
Calculate the second weighted average of the described second serious score；
The variation tendency of flaw is determined according to first weighted average and second weighted average.
In one embodiment, the second flaw quantity of flaw and second tight of obtaining in the second facial image on face
Score includes: again
Obtain in the second facial image the second flaw quantity of flaw on face, the second serious score and first identifier letter
Breath；
The variation tendency that flaw is determined according to the described first serious score and the second serious score includes:
According to track algorithm, the first identifier information and the second deep neural network identify first facial image with
It is the flaw of same flaw in second facial image；
Compare the serious score of same flaw in first facial image and second facial image, to obtain flaw
Variation tendency.
In one embodiment, the method also includes:
By the camera acquire different angle multiple facial images, the mark of multiple facial images with it is described
It identifies identical；
The identification information that flaw in multiple described facial images is obtained by track algorithm, obtains second identifier information；
It is described that the first face figure is identified according to track algorithm, the first identifier information and the second deep neural network
As be in second facial image same flaw flaw after, the method also includes:
According to the second deep neural network described in multiple described facial images and the second identifier information re -training, with
Update second deep neural network.
In one embodiment, described multiple facial images for acquiring different angle by camera include:
Pass through multiple different facial images of the relative position of camera acquisition face key position.
Second aspect provides a kind of trend determining device, comprising:
Acquisition unit, for acquiring the first facial image by camera；
Input unit, the first facial image for acquiring the acquisition unit input the first deep neural network, obtain
Obtain the first flaw quantity and the first serious score of flaw on face in first facial image；
First acquisition unit, the mark of the first facial image for obtaining the acquisition unit acquisition；
Second acquisition unit, for obtaining in the second facial image on face the second flaw quantity of flaw and second serious
Score, second facial image are the facial images corresponding with the mark of first acquisition unit acquisition of storage；
Determination unit, the first flaw quantity and the second acquisition unit for being obtained according to the input unit obtain
The second flaw quantity determine the variation tendency of flaw, or it is true according to the described first serious score and the second serious score
The variation tendency of the fixed flaw, or seriously divided according to the first flaw quantity, the second flaw quantity, described first
Several and the described second serious score determines the variation tendency of the flaw.
In one embodiment, the input unit includes:
Subelement is inputted, the first facial image for acquiring the acquisition unit inputs the first deep neural network,
Obtain the flaw probability and serious score of all flaws to be determined on face in first facial image；
First determines subelement, in the case where the flaw probability of flaw to be determined is greater than threshold value, determine it is described to
Determine that flaw is flaw, the flaw to be determined is any to true in all flaws to be determined that the input subelement obtains
Determine flaw；
Subelement is counted, determines that the quantity of the determining flaw of subelement is the first face figure for counting described first
As on face flaw the first flaw quantity；
Second determines subelement, for determining that it is described that the serious score of the determining flaw of subelement is determined as by described first
The first of flaw serious score on face in first facial image.
In one embodiment, the determination unit is according to the first flaw quantity, the second flaw quantity, described
First serious score and the second serious score determine that the variation tendency of flaw includes:
In the case where the first flaw quantity is different from the second flaw quantity, according to the first flaw quantity
The variation tendency of flaw is determined with the second flaw quantity；
In the first flaw quantity situation identical with the second flaw quantity, according to the described first serious score
The variation tendency of flaw is determined with the described second serious score.
In one embodiment, the determination unit is determined according to the described first serious score and the second serious score
The variation tendency of flaw includes:
Calculate the first weighted average of the described first serious score；
Calculate the second weighted average of the described second serious score；
The variation tendency of flaw is determined according to first weighted average and second weighted average.
In one embodiment, the second acquisition unit is specifically used for obtaining in the second facial image flaw on face
The second flaw quantity, the second serious score and first identifier information；
The determination unit determines the variation tendency of flaw according to the described first serious score and the second serious score
Include:
According to track algorithm, the first identifier information and the second deep neural network identify first facial image with
It is the flaw of same flaw in second facial image；
Compare the serious score of same flaw in first facial image and second facial image, to obtain flaw
Variation tendency.
In one embodiment, the acquisition unit is also used to acquire multiple people of different angle by the camera
Face image, the mark of multiple facial images are identical as the mark；
Third acquiring unit, for obtaining flaw in multiple facial images that the acquisition unit acquires by track algorithm
Identification information, obtain second identifier information；
Training unit, for according to the identification of track algorithm, the first identifier information and the second deep neural network
Be in first facial image and second facial image same flaw flaw after, according to the more of acquisition unit acquisition
The second deep neural network described in facial image and the second identifier information re -training of third acquiring unit acquisition is opened, with
Update second deep neural network.
In one embodiment, the acquisition unit includes: by multiple facial images that camera acquires different angle
Pass through multiple different facial images of the relative position of camera acquisition face key position.
The third aspect provides a kind of trend determining device, including processor, memory and camera, the processor, institute
It states memory and the camera is connected with each other, the camera is for acquiring image, and the memory is for storing computer
Program, the computer program include program instruction, the processor for call described program instruction execution first aspect or
The trend that any embodiment of first aspect provides determines method.
Fourth aspect provides a kind of readable storage medium storing program for executing, and the readable storage medium storing program for executing is stored with computer program, described
Computer program includes program instruction, described program instruction make when being executed by a processor the processor execute first aspect or
The trend that any embodiment of first aspect provides determines method.
5th aspect provides a kind of application program, and the application program for executing first aspect or first party at runtime
The trend that any embodiment in face provides determines method.
In the embodiment of the present invention, the first facial image is acquired by camera, the first facial image is inputted into the first depth
Neural network obtains the first flaw quantity and the first serious score of flaw on face in the first facial image, obtains the first face
The mark of image obtains in the second facial image corresponding with the mark of storage the second flaw quantity of flaw and on face
Two serious scores determine the variation tendency of flaw according to the first flaw quantity and the second flaw quantity, or serious according to first
Score and the second serious score determine the variation tendency of flaw, or sternly according to the first flaw quantity, the second flaw quantity, first
Weight score and the second serious score determine the variation tendency of flaw, it is seen then that can be by the image of acquisition and the image of storage
The quantity of flaw and/or serious score determine the situation of change of flaw on facial skin on face, therefore, can accurately determine
The situation of change of flaw on facial skin.
Detailed description of the invention
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to needed in the embodiment
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is the flow diagram that a kind of trend provided in an embodiment of the present invention determines method；
Fig. 2 is a kind of structural schematic diagram of trend determining device provided in an embodiment of the present invention；
Fig. 3 is the structural schematic diagram of another trend determining device provided in an embodiment of the present invention；
Fig. 4 is a kind of structural schematic diagram of first deep neural network provided in an embodiment of the present invention；
Fig. 5 is a kind of structural schematic diagram of second deep neural network provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are some of the embodiments of the present invention, instead of all the embodiments.Based on this hair
Embodiment in bright, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
The embodiment of the present invention provides a kind of trend and determines method and device, for accurately determining flaw on facial skin
Situation of change.It is described in detail separately below.
Referring to Fig. 1, Fig. 1 is the flow diagram that a kind of trend provided in an embodiment of the present invention determines method.Wherein, should
Trend determines that method is suitable for being equipped with the electronic equipments such as mobile phone, the tablet computer of camera.As shown in Figure 1, the trend determines
Method may comprise steps of.
101, the first facial image is acquired by camera.
In the present embodiment, the variation by the application observation facial skin on electronic equipment or electronic equipment is needed in user
In the case where situation, user can be by application input of the operation electronic equipment into electronic equipment or electronic equipment for analyzing
Analysis instruction, when electronic equipment detects analysis instruction, start camera, and by camera acquisition the first facial image.
The camera of starting can be front camera, be also possible to rear camera.It is not that user needs in the camera of starting
In the case where camera, preset icon, the predeterminable area etc. that user can shoot interface by clicking camera are inputted for cutting
The switching command of camera is changed, after electronic equipment detects the switching command, switches camera, before the camera of starting is
In the case where setting camera, it can be switched to rear camera, it, can be in the case where the camera of starting is rear camera
It is switched to front camera.
102, the first facial image is inputted into the first deep neural network, flaw on face in the first facial image of acquisition
First flaw quantity and the first serious score.
It is after collecting the first facial image by camera, the first facial image input first is deep in the present embodiment
Neural network is spent, the first flaw quantity and the first serious score of flaw on face in the first facial image are obtained.Can first by
First facial image inputs the first deep neural network, exports the flaw of all flaws to be determined on face in the first facial image
Probability and serious score judge the flaw of flaw to be determined later to each flaw to be determined in all flaws to be determined of output
Whether defect probability is greater than threshold value, in the case where judging that the flaw probability of flaw to be determined is greater than first threshold, determines to true
Determining flaw is flaw, in the case where judging that the flaw probability of flaw to be determined is less than or equal to first threshold, is determined to true
Determining flaw is not flaw.The quantity for counting determining flaw later is the first flaw number of flaw on face in the first facial image
Amount, is determined as in the first facial image the first of flaw the serious score on face for the serious score of determining flaw.Later may be used
To store the first flaw quantity and the first serious score of flaw on face in the first facial image and the first facial image, with
Just subsequent calls.Flaw probability is the probability that flaw to be determined belongs to flaw, and serious score is flaw severity to be determined
Score, serious score is higher, shows that flaw is more serious.
In the present embodiment, the first facial image is inputted into the first deep neural network, the first facial image can also be exported
Position probability, class probability and/or the coordinate of all flaws to be determined on middle face, determine flaw to be determined be flaw it
Afterwards, flaw can be sorted out to the flaw position of position maximum probability and the maximum flaw classification of class probability.Therefore, Ke Yitong
The flaw quantity for counting out each flaw position on face in the first facial image, can also count face in the first facial image
The flaw quantity of upper difference flaw classification can also count in the first facial image on face the different flaws on each flaw position
The flaw quantity of defect classification.Also the above- mentioned information of statistics be can store, be called so as to subsequent.Flaw position is that there are flaws
Position, may include nose, forehead etc., flaw classification may include blackhead, small pox, spot etc..Position probability is to be determined
Flaw belongs to the probability at that position on face, and class probability is the probability that flaw to be determined belongs to that flaw.Have on face
How many each flaws to be determined in a position will export how many a position probability, after determining that flaw to be determined is flaw, these
That position maximum probability, this flaw will belong to the position of this position maximum probability in the probability of position.Similarly, how many is planted
The each flaw to be determined of flaw classification will export how many a class probabilities, after determining that flaw to be determined is flaw, these classes
That class probability is maximum in other probability, this flaw will belong to the maximum flaw classification of this class probability.
Referring to Fig. 4, Fig. 4 is a kind of structural schematic diagram of first deep neural network provided in an embodiment of the present invention.Such as
Shown in Fig. 4, the first deep neural network includes 9 layers of structure, and 1-4 layers include convolutional layer and maximum pond layer (Maxpool
Layer), the number of plies of convolutional layer and/or convolution kernel is of different sizes but in every layer；5-8 layers are convolutional layer, but every layer of layer
It is several and/or convolution kernel of different sizes；9th layer is Output matrix, can be 7 × 7 × 6 matrix.
103, the mark of the first facial image is obtained.
In the present embodiment, after collecting the first facial image by camera, the mark of the first facial image is obtained, it can
To identify the mark of the first facial image by face recognition technology, which can be with the face figure of one people of unique identification
Picture.Face recognition technology can be FaceNet etc..Wherein, step 102 and step 103 can execute parallel, can also serially hold
Row.
104, the second flaw quantity and the second serious score of flaw on face in the second facial image are obtained.
In the present embodiment, after the mark for getting the first facial image, the mark is obtained from the facial image of storage
Second flaw of corresponding facial image and shooting time and flaw on face in shortest second facial image of current time interval
Defect quantity and the second serious score.In addition it is also possible to obtain in the second facial image the label of flaw, each flaw portion on face
The flaw quantity of different flaw classifications on the flaw quantity of position, the flaw quantity of different flaw classifications and/or each flaw position.
Wherein, the second facial image is the facial image acquired when last trend determines, and acquisition time is than the first facial image
It is early.
105, the variation tendency of flaw is determined.
In the present embodiment, the first flaw quantity and the first serious score of flaw on face in the first facial image are obtained,
And second flaw on face in facial image the second flaw quantity and the second serious score after, according to the first flaw quantity
The variation tendency of flaw is determined with the second flaw quantity and/or the first serious score and the second serious score.According to first flaw
It, can be in the case that defect quantity, the second flaw quantity, the first serious score and the second serious score determine the variation tendency of flaw
Compare the first flaw quantity and the second flaw quantity, in the case where the first flaw quantity is greater than the second flaw quantity, determines the flaw
Defect becomes serious；In the case where the first flaw quantity is less than the second flaw quantity, determine that flaw improves；In the first flaw number
It measures in situation identical with the second flaw quantity, determines that the variation of flaw becomes according to the first serious score and the second serious score
Gesture.It can also determine that flaw tails off, becomes more or do not become by the first flaw quantity and the second flaw quantity, while tight by first
Weight score and the second serious score determine that the same flaw improves, becomes serious or unchanged.
In the present embodiment, the first weighted average of the first serious score can be calculated, calculate the of the second serious score
Two weighted averages determine that flaw becomes serious in the case where comparing the first weighted average greater than the second weighted average
, in the case where comparing the first weighted average equal to the second weighted average, can determine that flaw is unchanged, compare
In the case that the first weighted average is less than the second weighted average out, it can determine that flaw improves.
In the present embodiment, the second flaw quantity of flaw and the second serious score on face are obtained in the second facial image
Meanwhile in available second facial image on face flaw first identifier information.It can also be according to track algorithm, the first mark
It is the flaw of same flaw in knowledge information and the second deep neural network the first facial image of identification and the second facial image, compares
The serious score of same flaw obtains the variation tendency of flaw in first facial image and the second facial image, that is, determines same
The situation of change of flaw.
As a kind of possible embodiment, multiple facial images that different angle can be acquired by camera, pass through
The identification information that track algorithm obtains flaw in this multiple facial image obtains second identifier information, the mark letter of the same flaw
Breath must be identical, and the identification information of different flaws must be different.According to track algorithm, first identifier information and the second depth nerve
Be in the first facial image of Network Recognition and the second facial image same flaw flaw after, according to this multiple facial image and
Second identifier information the second deep neural network of re -training, to update the second deep neural network.Is being used every time
After two deep neural networks, acquisition facial image carries out re -training, so that the second deep neural network used every time is
Newest trained deep neural network, so as to improve the accuracy of identification flaw, so as to the face acquired next time
Image can be identified accurately.Such as: the previous day recognizes on the face there are three flaw A, B and C, according to track algorithm, the flaw
These three flaws are also deposited in the facial image that the identification information of defect A, B and C and the identification of the second deep neural network acquire one day after
Be not present, if it does, flaw A, B and location of C where.The mark and the mark of the first facial image of this multiple facial image
It is identical, to be ensured of the facial image of the same person.The relative position that face key position can be acquired by camera is different
Multiple facial images, can be user's active accommodation position, be also possible to electronic device prompts user adjust position.Tracking is calculated
Method can determine that those flaws are the same flaws in this multiple facial image, track algorithm can be core for tracking flaw
Correlation filtering (Kernel Correlation Filter, KCF) algorithm, or differentiate scale space tracking
(Discriminatiive Scale Space Tracker, DSST) algorithm can also be other track algorithms.Such as: tracking
Algorithm obtains different angle flaw sample, for tracking to flaw A, B and C to optimize the second deep neural network
Parameter enables it to more accurately identify flaw A, B and C.
Referring to Fig. 5, Fig. 5 is a kind of structural schematic diagram of second deep neural network provided in an embodiment of the present invention.Such as
Shown in Fig. 5, nervus opticus network includes 9 layers of structure, and 1-4 layers include convolutional layer and maximum pond layer (Maxpool
Layer), the number of plies of convolutional layer and/or convolution kernel is of different sizes but in every layer；5-8 layers are convolutional layer, but every layer of layer
It is several and/convolution kernel of different sizes；9th layer is Output matrix, can be 7 × 7 × 504 matrix.
The trend described in Fig. 1 determines in method, can pass through the flaw on face in the image of acquisition and the image of storage
The quantity of defect and/or serious score determine the situation of change of flaw on facial skin, therefore, can accurately determine facial skin
The situation of change of upper flaw.
Referring to Fig. 2, Fig. 2 is a kind of structural schematic diagram of trend determining device provided in an embodiment of the present invention.The trend
Determining device can be the electronic equipments such as the mobile phone for being equipped with camera, tablet computer.As shown in Fig. 2, the trend determining device
May include:
Acquisition unit 201, for acquiring the first facial image by camera；
Input unit 202, the first facial image for acquiring acquisition unit 201 input the first deep neural network,
Obtain the first flaw quantity and the first serious score of flaw on face in the first facial image；
First acquisition unit 203, the mark of the first facial image for obtaining the acquisition of acquisition unit 201；
Second acquisition unit 204, for obtaining the second flaw quantity and second of flaw on face in the second facial image
Serious score, the second facial image are the facial images corresponding with the mark of the acquisition of first acquisition unit 203 of storage；
Determination unit 205, the first flaw quantity and second acquisition unit 204 for being obtained according to input unit 202 obtain
The the second flaw quantity taken determines the variation tendency of flaw, and the change of flaw is determined according to the first serious score and the second serious score
Change trend, or flaw is determined according to the first flaw quantity, the second flaw quantity, the first serious score and the second serious score
Variation tendency.
As a kind of possible embodiment, input unit 202 may include:
Subelement 2021 is inputted, the first facial image for acquiring acquisition unit 201 inputs the first depth nerve net
Network obtains the flaw probability and serious score of all flaws to be determined on face in the first facial image；
First determines subelement 2022, in the case where the flaw probability of flaw to be determined is greater than threshold value, determining should
Flaw to be determined is flaw, which is any to true in all flaws to be determined for input the acquisition of subelement 2021
Determine flaw；
Subelement 2023 is counted, determines that the quantity of the determining flaw of subelement 2022 is the first face figure for counting first
As on face flaw the first flaw quantity；
Second determines subelement 2024, for determining that the serious score of the determining flaw of subelement 2022 is determined as first
The first of flaw serious score on face in first facial image.
As a kind of possible embodiment, determination unit 205 is according to the first flaw quantity, the second flaw quantity, first
Serious score and the second serious score determine that the variation tendency of flaw includes:
In the case where the first flaw quantity is different from the second flaw quantity, according to the first flaw quantity and the second flaw number
Measure the variation tendency for determining flaw；
In the first flaw quantity situation identical with the second flaw quantity, according to serious point of the first serious score and second
Number determines the variation tendency of flaw.
As a kind of possible embodiment, determination unit 205 is determined according to the first serious score and the second serious score
The variation tendency of flaw includes:
Calculate the first weighted average of the first serious score；
Calculate the second weighted average of the second serious score；
The variation tendency of flaw is determined according to the first weighted average and the second weighted average.
In one embodiment, second acquisition unit 204, specifically for flaw on face in the second facial image of acquisition
Second flaw quantity, the second serious score and first identifier information；
Determination unit 205 determines that the variation tendency of flaw includes: according to the first serious score and the second serious score
The first identifier information and the identification of the second deep neural network obtained according to track algorithm, second acquisition unit 204
It is the flaw of same flaw in first facial image and the second facial image；
Compare the serious score of same flaw in the first facial image and the second facial image, is become with obtaining the variation of flaw
Gesture.
As a kind of possible embodiment, acquisition unit 201 are also used to acquire different angle multiple by camera
The mark of facial image, multiple facial images is identical as the mark of the first facial image；
Third acquiring unit 206, for obtaining the flaw in multiple facial images that acquisition unit 201 acquires by track algorithm
The identification information of defect obtains second identifier information；
Training unit 207, it is the first for being identified according to track algorithm, first identifier information and the second deep neural network
Be in face image and the second facial image same flaw flaw after, multiple facial images for being acquired according to acquisition unit 201
Second identifier information the second deep neural network of re -training obtained with third acquiring unit 206, to update the second depth mind
Through network.
In one embodiment, acquisition unit 201 includes: by multiple facial images that camera acquires different angle
Pass through multiple different facial images of the relative position of camera acquisition face key position.
In the trend determining device described in Fig. 2, the flaw on face in the image of acquisition and the image of storage can be passed through
The quantity of defect and/or serious score determine the situation of change of flaw on facial skin, therefore, can accurately determine facial skin
The situation of change of upper flaw.
It is understood that the function of the unit of the certain device of the trend of the present embodiment can determine method according to above-mentioned trend
Method specific implementation in embodiment, specific implementation process are referred to above-mentioned trend and determine that the correlation of embodiment of the method is retouched
It states, details are not described herein again.
Referring to Fig. 3, Fig. 3 is the structural schematic diagram of the certain device of another trend provided in an embodiment of the present invention.This becomes
Gesture determining device can be the electronic equipments such as the mobile phone for being equipped with camera, tablet computer.As shown in figure 3, the trend fills really
Set may include at least one processor 301, memory 302, camera 303 and bus 304, processor 301, memory 302
It is connected between camera 303 by bus 304, in which:
Camera 303, for acquiring the first facial image；
Memory 302 includes program instruction for storing computer program, computer program, and processor 301 is for calling
The program instruction that memory 302 stores executes following steps:
First facial image is inputted into the first deep neural network, obtains in the first facial image first of flaw on face
Flaw quantity and the first serious score；
Obtain the mark of the first facial image；
Obtain the second flaw quantity and the second serious score of flaw on face in the second facial image, the second facial image
It is the facial image corresponding with the mark of storage；
The variation tendency that flaw is determined according to the first flaw quantity and the second flaw quantity, according to the first serious score and
Two serious scores determine the variation tendency of flaw, or according to the first flaw quantity, the second flaw quantity, the first serious score and
Second serious score determines the variation tendency of flaw.
As a kind of possible embodiment, the first facial image is inputted the first deep neural network by processor 301, is obtained
The first flaw quantity of flaw and the first serious score include: on face in the first facial image
First facial image is inputted into the first deep neural network, is obtained all to be determined on face in the first facial image
The flaw probability and serious score of flaw；
In the case where the flaw probability of flaw to be determined is greater than threshold value, determine that the flaw to be determined is flaw, it should be to true
Determining flaw is any flaw to be determined in all flaws to be determined；
The quantity for counting determining flaw is the first flaw quantity of flaw on face in the first facial image；
The serious score of determining flaw is determined as in the first facial image the first of flaw the serious score on face.
As a kind of possible embodiment, processor 301 is according to the first flaw quantity, the second flaw quantity, first tight
Weight score and the second serious score determine that the variation tendency of flaw includes:
In the case where the first flaw quantity is different from the second flaw quantity, according to the first flaw quantity and the second flaw number
Measure the variation tendency for determining flaw；
In the first flaw quantity situation identical with the second flaw quantity, according to serious point of the first serious score and second
Number determines the variation tendency of flaw.
As a kind of possible embodiment, processor 301 determines the flaw according to the first serious score and the second serious score
The variation tendency of defect includes:
Calculate the first weighted average of the first serious score；
Calculate the second weighted average of the second serious score；
The variation tendency of flaw is determined according to the first weighted average and the second weighted average.
As a kind of possible embodiment, processor 301 obtains second flaw of flaw on face in the second facial image
Defect quantity and the second serious score include:
Obtain in the second facial image the second flaw quantity of flaw on face, the second serious score and first identifier letter
Breath；
Processor 301 determines the variation tendency packet of flaw according to the described first serious score and the second serious score
It includes:
The first facial image and the second face are identified according to track algorithm, first identifier information and the second deep neural network
It is the flaw of same flaw in image；
Compare the serious score of same flaw in the first facial image and the second facial image, is become with obtaining the variation of flaw
Gesture.
As a kind of possible embodiment, camera 303 are also used to acquire multiple facial images of different angle, should
The mark of multiple facial images is identical as the mark of the first facial image；
The program instruction that processor 301 is also used to that memory 302 is called to store executes following steps:
The identification information of flaw in multiple facial images is obtained by track algorithm, obtains second identifier information；
Processor 301 identifies the first facial image according to track algorithm, first identifier information and the second deep neural network
Be in the second facial image same flaw flaw after, the program that processor 301 is also used to call memory 302 to store refers to
It enables and executes following steps:
It is deep to update second according to multiple facial images and second identifier information the second deep neural network of re -training
Spend neural network.
As a kind of possible embodiment, multiple facial images of the acquisition of camera 303 different angle include:
Acquire multiple different facial images of the relative position of face key position.
In the trend determining device described in Fig. 3, the flaw on face in the image of acquisition and the image of storage can be passed through
The quantity of defect and/or serious score determine the situation of change of flaw on facial skin, therefore, can accurately determine facial skin
The situation of change of upper flaw.
Wherein, step 101 can be executed by the camera 303 in the certain device of trend, and step 102- step 105 can be with
By in the certain device of trend processor 301 and memory 302 execute.
Wherein, acquisition unit 201 can be realized by the camera 303 in the certain device of trend, input unit 202,
One acquiring unit 203, second acquisition unit 204, determination unit 205, tag unit 206 and training unit 207 can be by trend
Processor 301 and memory 302 in certain device are realized.
A kind of readable storage medium storing program for executing is provided in one embodiment, which is stored with computer program,
Computer program includes program instruction, and the trend that program instruction makes processor execute Fig. 1 when being executed by a processor determines method.
A kind of application program is provided in one embodiment, and the application program for executing the trend of Fig. 1 at runtime
Determine method.
Those of ordinary skill in the art will appreciate that all or part of the steps in the various methods of above-described embodiment is can
It is completed with instructing relevant hardware by program, which can be stored in a computer readable storage medium, storage
Medium may include: flash disk, read-only memory (Read-Only Memory, ROM), random access device (Random Access
Memory, RAM), disk or CD etc..
The embodiment of the present invention has been described in detail above, specific case used herein to the principle of the present invention and
Embodiment is expounded, and the above description of the embodiment is only used to help understand the method for the present invention and its core ideas；
At the same time, for those skilled in the art can in specific embodiments and applications according to the thought of the present invention
There is change place, in conclusion the contents of this specification are not to be construed as limiting the invention.
Claims (10)
1. a kind of trend determines method characterized by comprising
The first facial image is acquired by camera；
First facial image is inputted into the first deep neural network, flaw on face in acquisition first facial image
First flaw quantity and the first serious score；
Obtain the mark of first facial image；
Obtain the second flaw quantity and the second serious score of flaw on face in the second facial image, second facial image
It is the facial image corresponding with the mark of storage；
The variation tendency of flaw is determined according to the first flaw quantity and the second flaw quantity, or according to described first
Serious score and the second serious score determine the variation tendency of the flaw, or according to the first flaw quantity, institute
State the variation tendency that the second flaw quantity, the first serious score and the second serious score determine the flaw.
2. the method according to claim 1, wherein described input the first depth mind for first facial image
Through network, the first flaw quantity of flaw and the first serious score include: on face in acquisition first facial image
First facial image is inputted into the first deep neural network, obtains in first facial image and is needed on face
Determine the flaw probability and serious score of flaw；
In the case where the flaw probability of flaw to be determined is greater than threshold value, determine that the flaw to be determined is flaw, it is described to true
Determining flaw is any flaw to be determined in all flaws to be determined；
The quantity for counting determining flaw is the first flaw quantity of flaw on face in first facial image；
The serious score of the flaw of the determination is determined as in first facial image first serious point of flaw on face
Number.
3. according to the method described in claim 2, it is characterized in that, described according to the first flaw quantity, second flaw
Defect quantity, the first serious score and the second serious score determine that the variation tendency of flaw includes:
In the case where the first flaw quantity is different from the second flaw quantity, according to the first flaw quantity and institute
State the variation tendency that the second flaw quantity determines flaw；
In the first flaw quantity situation identical with the second flaw quantity, according to the described first serious score and institute
State the variation tendency that the second serious score determines flaw.
4. method according to claim 1-3, which is characterized in that described according to the described first serious score and institute
It states the second serious score and determines that the variation tendency of flaw includes:
Calculate the first weighted average of the described first serious score；
Calculate the second weighted average of the described second serious score；
The variation tendency of flaw is determined according to first weighted average and second weighted average.
5. method according to claim 1-3, which is characterized in that in the second facial image of the acquisition on face
Second flaw quantity of flaw and the second serious score include:
Obtain in the second facial image the second flaw quantity of flaw, the second serious score and first identifier information on face；
The variation tendency that flaw is determined according to the described first serious score and the second serious score includes:
According to track algorithm, the first identifier information and the second deep neural network identify first facial image with it is described
It is the flaw of same flaw in second facial image；
Compare the serious score of same flaw in first facial image and second facial image, to obtain the change of flaw
Change trend.
6. according to the method described in claim 5, it is characterized in that, the method also includes:
Multiple facial images of different angle, the mark of multiple facial images and the mark are acquired by the camera
It is identical；
The identification information that flaw in multiple described facial images is obtained by track algorithm, obtains second identifier information；
It is described according to track algorithm, the first identifier information and the second deep neural network identify first facial image with
Be in second facial image same flaw flaw after, the method also includes:
According to second deep neural network described in multiple described facial images and the second identifier information re -training, with
Update second deep neural network.
7. according to the method described in claim 6, it is characterized in that, described multiple faces for acquiring different angle by camera
Image includes:
Pass through multiple different facial images of the relative position of camera acquisition face key position.
8. a kind of trend determining device characterized by comprising
Acquisition unit, for acquiring the first facial image by camera；
Input unit, the first facial image for acquiring the acquisition unit input the first deep neural network, obtain institute
State the first flaw quantity and the first serious score of flaw on face in the first facial image；
First acquisition unit, the mark of the first facial image for obtaining the acquisition unit acquisition；
Second acquisition unit, for obtaining, the second flaw quantity of flaw and second is seriously divided on face in the second facial image
Number, second facial image are the facial images corresponding with the mark of first acquisition unit acquisition of storage；
Determination unit, the first flaw quantity for being obtained according to the input unit and the second acquisition unit obtain the
Two flaw quantity determine the variation tendency of flaw, or determine institute according to the described first serious score and the second serious score
State the variation tendency of flaw, or according to the first flaw quantity, the second flaw quantity, the first serious score and
The second serious score determines the variation tendency of the flaw.
9. a kind of trend determining device, which is characterized in that including processor, memory and camera, the processor described is deposited
Reservoir and the camera are connected with each other, and the camera is used to store computer program for acquiring image, the memory,
The computer program includes program instruction, and the processor is for calling described program instruction execution such as claim 1-7 to appoint
Trend described in one determines method.
10. a kind of storage medium, which is characterized in that the storage medium is stored with computer program, the computer program packet
Program instruction is included, described program instruction executes the processor such as any one of claim 1-7 institute
The trend stated determines method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811309009.5A CN109544515B (en) | 2018-11-05 | 2018-11-05 | Trend determination method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811309009.5A CN109544515B (en) | 2018-11-05 | 2018-11-05 | Trend determination method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109544515A true CN109544515A (en) | 2019-03-29 |
| CN109544515B CN109544515B (en) | 2021-08-31 |
Family
ID=65846548
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811309009.5A Active CN109544515B (en) | 2018-11-05 | 2018-11-05 | Trend determination method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109544515B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110116691A1 (en) * | 2009-11-13 | 2011-05-19 | Chung Pao-Choo | Facial skin defect resolution system, method and computer program product |
| CN102479322A (en) * | 2010-11-30 | 2012-05-30 | 财团法人资讯工业策进会 | System, apparatus and method for analyzing facial defect by facial image with angle |
| CN104732195A (en) * | 2013-12-19 | 2015-06-24 | 国际商业机器公司 | Mining social media for ultraviolet light exposure analysis |
| CN106469302A (en) * | 2016-09-07 | 2017-03-01 | 成都知识视觉科技有限公司 | A kind of face skin quality detection method based on artificial neural network |
| CN107679507A (en) * | 2017-10-17 | 2018-02-09 | 北京大学第三医院 | Facial pores detecting system and method |
| CN108323203A (en) * | 2017-07-17 | 2018-07-24 | 深圳和而泰智能控制股份有限公司 | A kind of method, apparatus and intelligent terminal quantitatively detecting face skin quality parameter |
| CN108323204A (en) * | 2017-07-17 | 2018-07-24 | 深圳和而泰智能控制股份有限公司 | A kind of method and intelligent terminal of detection face flaw point |
| CN108647566A (en) * | 2018-03-29 | 2018-10-12 | 维沃移动通信有限公司 | A kind of method and terminal of identification skin characteristic |
-
2018
- 2018-11-05 CN CN201811309009.5A patent/CN109544515B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110116691A1 (en) * | 2009-11-13 | 2011-05-19 | Chung Pao-Choo | Facial skin defect resolution system, method and computer program product |
| CN102479322A (en) * | 2010-11-30 | 2012-05-30 | 财团法人资讯工业策进会 | System, apparatus and method for analyzing facial defect by facial image with angle |
| CN104732195A (en) * | 2013-12-19 | 2015-06-24 | 国际商业机器公司 | Mining social media for ultraviolet light exposure analysis |
| CN106469302A (en) * | 2016-09-07 | 2017-03-01 | 成都知识视觉科技有限公司 | A kind of face skin quality detection method based on artificial neural network |
| CN108323203A (en) * | 2017-07-17 | 2018-07-24 | 深圳和而泰智能控制股份有限公司 | A kind of method, apparatus and intelligent terminal quantitatively detecting face skin quality parameter |
| CN108323204A (en) * | 2017-07-17 | 2018-07-24 | 深圳和而泰智能控制股份有限公司 | A kind of method and intelligent terminal of detection face flaw point |
| CN107679507A (en) * | 2017-10-17 | 2018-02-09 | 北京大学第三医院 | Facial pores detecting system and method |
| CN108647566A (en) * | 2018-03-29 | 2018-10-12 | 维沃移动通信有限公司 | A kind of method and terminal of identification skin characteristic |
Non-Patent Citations (4)
| Title |
|---|
| 吴意平 等: "深度学习方法建立痤疮分级系统的探索", 《中国医药》 * |
| 李玲玉: "基于机器视觉下的皮肤老化评价系统设计", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》 * |
| 李顾全 等: ""基于图像处理的皮肤健康检测研究"", 《探索与研究》 * |
| 牛尔老师: ""你的色斑为什么越来越多"", 《HTTPS://WWW.DOUBAN.COM/GROUP/TOPIC/40528961/》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109544515B (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7058373B2 (en) | Lesion detection and positioning methods, devices, devices, and storage media for medical images | |
| CN109766840A (en) | Facial expression recognizing method, device, terminal and storage medium | |
| CN109815965B (en) | Image filtering method and device and storage medium | |
| CN109472608A (en) | Business confirmation method and terminal device based on Emotion identification | |
| CN109800731A (en) | Fingerprint input method and relevant apparatus | |
| CN111345816A (en) | Multi-lead QRS complex detection method, device, equipment and storage medium | |
| CN109948511A (en) | Gesture identification method and device | |
| CN107590474A (en) | Solve lock control method and Related product | |
| CN111657925A (en) | Electrocardiosignal classification method, system, terminal and storage medium based on machine learning | |
| WO2021073117A1 (en) | Oct image segmentation method, apparatus and device, and storage medium | |
| CN111028218A (en) | Method and device for training fundus image quality judgment model and computer equipment | |
| CN110458875A (en) | Detection method, image split-joint method, related device and the equipment of abnormal point pair | |
| CN116863522A (en) | Acne grading method, device, equipment and medium | |
| CN112306829A (en) | Method and device for determining performance information, storage medium and terminal | |
| CN109240759A (en) | Application program launching method, device, terminal device and readable storage medium storing program for executing | |
| CN110809768A (en) | Data cleansing system and method | |
| CN111340213A (en) | Neural network training method, electronic device, and storage medium | |
| WO2017145172A1 (en) | System and method for extraction and analysis of samples under a microscope | |
| CN109697421A (en) | Evaluation method, device, computer equipment and storage medium based on micro- expression | |
| CN109544515A (en) | A kind of trend determines method and device | |
| CN112818899A (en) | Face image processing method and device, computer equipment and storage medium | |
| CN109003264B (en) | Retinopathy image type identification method and device and storage medium | |
| CN109492602A (en) | A kind of process clocking method and system based on human body limb language | |
| CN104866745B (en) | A kind of electronic equipment and information processing method | |
| CN106445654A (en) | Method and device for determining response priorities of control commands |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 518000 Guangdong science and technology innovation and Research Institute, Shenzhen, Shenzhen, Nanshan District No. 6, science and technology innovation and Research Institute, Shenzhen, D 10, 1004, 10 Patentee after: Shenzhen Hetai intelligent home appliance controller Co.,Ltd. Address before: 518000 Guangdong science and technology innovation and Research Institute, Shenzhen, Shenzhen, Nanshan District No. 6, science and technology innovation and Research Institute, Shenzhen, D 10, 1004, 10 Patentee before: SHENZHEN H&T DATA RESOURCES AND CLOUD TECHNOLOGY Ltd. |