CN110070049A - Facial image recognition method and device, electronic equipment and storage medium - Google Patents
Facial image recognition method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN110070049A CN110070049A CN201910330767.3A CN201910330767A CN110070049A CN 110070049 A CN110070049 A CN 110070049A CN 201910330767 A CN201910330767 A CN 201910330767A CN 110070049 A CN110070049 A CN 110070049A
- Authority
- CN
- China
- Prior art keywords
- parameter
- similarity
- feature vector
- obtains
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
This disclosure relates to a kind of facial image recognition method and device, electronic equipment and storage medium, wherein the described method includes: obtaining multiple facial images;The corresponding obtained multiple feature vectors of feature extraction are carried out according to the multiple facial image, obtain multiple target objects to be identified;According to classification belonging to the multiple feature vector, sorting parameter is obtained;Classified according to the sorting parameter to the multiple target object to be identified, obtains classification results.Recognition effect to facial image can be improved using the embodiment of the present disclosure.
Description
Technical field
This disclosure relates to technical field of computer vision more particularly to a kind of facial image recognition method and device, electronics
Equipment and storage medium.
Background technique
Facial image identification is the vital task of computer vision application, is schemed by the face that optical element will be used to acquire
As being mapped to identity separating capacity, computable data characteristics, the verifying or identification of identity are carried out.Facial image identification
Technology has extremely important application value under the application scenarios such as security protection, criminal investigation and authorization, is protection the people's lives and property
The important technology for being inviolable, functional department being facilitated to carry out the work.In facial image identification, the identification parameter of use is usually people
Accuracy for the hyper parameter of setting, facial image identification is low.
Summary of the invention
The present disclosure proposes a kind of facial image identification technology schemes.
According to the one side of the disclosure, a kind of facial image recognition method is provided, comprising:
Obtain multiple facial images;
The corresponding obtained multiple feature vectors of feature extraction are carried out according to the multiple facial image, are obtained multiple wait know
Other target object;
According to classification belonging to the multiple feature vector, sorting parameter is obtained;
Classified according to the sorting parameter to the multiple target object to be identified, obtains classification results.
In possible implementation, the classification according to belonging to the multiple feature vector obtains sorting parameter, packet
It includes:
According to the class number of the multiple feature vector generic, similarity parameter is obtained;
Processing is iterated to the similarity parameter, obtains adaptive similarity parameter;
The sorting parameter is obtained according to the adaptive similarity parameter.
It is described that the multiple target object to be identified is divided according to the sorting parameter in possible implementation
Class, comprising:
Sorter network according to sorting parameter training sorter network, after being trained;
According to the sorter network after the training, classify to the multiple target object to be identified.
In possible implementation, point according to the sorting parameter training sorter network, after being trained
Class network, comprising:
According to the sorting parameter, loss function is obtained;
The sorter network is trained according to the loss function backpropagation, the sorter network after obtaining the training.
In possible implementation, also wrapped according to the training process before the sorting parameter training sorter network
It includes:
The training data sample being made of the multiple feature vector is obtained, by described point of training data sample input
Similarity processing module in class network obtains adaptive similarity parameter;
The sorting parameter is obtained according to the adaptive similarity parameter.
In possible implementation, the training data sample is inputted into the processing mould of the similarity in the sorter network
Block obtains adaptive similarity parameter, comprising:
The similarity parameter and this training data sample iteration obtained according to last training data sample iteration obtains
COS distance, obtain the sum of cosine similarity;
According to characteristic feature vector Category Relevance in the sum of described cosine similarity and this training data sample
Median obtains the adaptive similarity parameter.
In possible implementation, the COS distance, according to feature vector in this training data sample iteration and its
Distance between non-corresponding category feature vector obtains.
In possible implementation, the median is right with it according to feature vector in this training data sample iteration
The angle between category feature vector is answered to obtain.
According to the one side of the disclosure, a kind of facial image identification device is provided, comprising:
Image collection module, for obtaining multiple facial images;
Semantic object extraction module, for multiple according to being obtained to the multiple facial image progress feature extraction correspondence
Feature vector obtains multiple target objects to be identified;
Parameter acquisition module obtains sorting parameter for the classification according to belonging to the multiple feature vector;
Categorization module is obtained for being classified according to the sorting parameter to the multiple target object to be identified
Classification results.
In possible implementation, the parameter acquisition module is used for:
According to the class number of the multiple feature vector generic, similarity parameter is obtained;
Processing is iterated to the similarity parameter, obtains adaptive similarity parameter;
The sorting parameter is obtained according to the adaptive similarity parameter.
In possible implementation, the categorization module is used for:
Sorter network according to sorting parameter training sorter network, after being trained;
According to the sorter network after the training, classify to the multiple target object to be identified.
In possible implementation, the categorization module is used for:
According to the sorting parameter, loss function is obtained;
The sorter network is trained according to the loss function backpropagation, the sorter network after obtaining the training.
In possible implementation, described device further include:
Sample acquisition module, for obtaining the training data sample being made of the multiple feature vector;
Processing module is obtained for the training data sample to be inputted the similarity processing module in the sorter network
To adaptive similarity parameter;
Sorting parameter determining module, for obtaining the sorting parameter according to the adaptive similarity parameter.
In possible implementation, the processing module is used for:
The similarity parameter and this training data sample iteration obtained according to last training data sample iteration obtains
COS distance, obtain the sum of cosine similarity;
According to characteristic feature vector Category Relevance in the sum of described cosine similarity and this training data sample
Median obtains the adaptive similarity parameter.
In possible implementation, the COS distance, according to feature vector in this training data sample iteration and its
Distance between non-corresponding category feature vector obtains.
In possible implementation, the median is right with it according to feature vector in this training data sample iteration
The angle between category feature vector is answered to obtain.
According to the one side of the disclosure, a kind of electronic equipment is provided, comprising:
Processor;
Memory for storage processor executable instruction;
Wherein, the processor is configured to: execute above-mentioned facial image recognition method.
According to the one side of the disclosure, a kind of computer readable storage medium is provided, computer program is stored thereon with
Instruction, the computer program instructions realize above-mentioned facial image recognition method when being executed by processor.
In the embodiments of the present disclosure, multiple facial images are obtained;Feature extraction is carried out according to the multiple facial image
Corresponding obtained multiple feature vectors, obtain multiple target objects to be identified;According to class belonging to the multiple feature vector
Not, sorting parameter is obtained;Classified according to the sorting parameter to the multiple target object to be identified, obtains classification knot
Fruit.Since sorting parameter (such as class probability) is being obtained according to multiple feature vector generics (such as class number), that is, to divide
Class parameter is non-artificial setting, therefore, can be more accurate in recognition of face, to improve the identification effect of facial image
Fruit.
It should be understood that above general description and following detailed description is only exemplary and explanatory, rather than
Limit the disclosure.
According to below with reference to the accompanying drawings to detailed description of illustrative embodiments, the other feature and aspect of the disclosure will become
It is clear.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and those figures show meet this public affairs
The embodiment opened, and together with specification it is used to illustrate the technical solution of the disclosure.
Fig. 1 shows the flow chart of the facial image recognition method according to the embodiment of the present disclosure.
Fig. 2 shows the schematic diagrames according to the training data sample of the embodiment of the present disclosure.
Fig. 3 shows the flow chart of the facial image recognition method according to the embodiment of the present disclosure.
Fig. 4 is according to obtaining the training process of adaptive similarity parameter in the embodiment of the present disclosure by training data sample
Schematic diagram;
Fig. 5 is the schematic diagram according to relationship between characteristic feature vector in the embodiment of the present disclosure;
Fig. 6 shows the class probability schematic diagram obtained according to the initialization similarity of the embodiment of the present disclosure.
Fig. 7 shows the schematic diagram that repetitive exercise is distributed according to the similarity of the embodiment of the present disclosure.
Fig. 8 shows the block diagram of the face identification method according to the embodiment of the present disclosure.
Fig. 9 shows the block diagram of the electronic equipment according to the embodiment of the present disclosure.
Figure 10 shows the block diagram of the electronic equipment according to the embodiment of the present disclosure.
Specific embodiment
Various exemplary embodiments, feature and the aspect of the disclosure are described in detail below with reference to attached drawing.It is identical in attached drawing
Appended drawing reference indicate element functionally identical or similar.Although the various aspects of embodiment are shown in the attached drawings, remove
It non-specifically points out, it is not necessary to attached drawing drawn to scale.
Dedicated word " exemplary " means " being used as example, embodiment or illustrative " herein.Here as " exemplary "
Illustrated any embodiment should not necessarily be construed as preferred or advantageous over other embodiments.
The terms "and/or", only a kind of incidence relation for describing affiliated partner, indicates that there may be three kinds of passes
System, for example, A and/or B, can indicate: individualism A exists simultaneously A and B, these three situations of individualism B.In addition, herein
Middle term "at least one" indicate a variety of in any one or more at least two any combination, it may for example comprise A,
B, at least one of C can indicate to include any one or more elements selected from the set that A, B and C are constituted.
In addition, giving numerous details in specific embodiment below to better illustrate the disclosure.
It will be appreciated by those skilled in the art that without certain details, the disclosure equally be can be implemented.In some instances, for
Method, means, element and circuit well known to those skilled in the art are not described in detail, in order to highlight the purport of the disclosure.
Recognition of face is the vital task of computer vision application, by reflecting the face-image for using optical element to acquire
With identity separating capacity, computable data characteristics is penetrated into, the verifying or identification of identity are carried out.Face recognition technology is being pacified
There is extremely important application value under the scenes such as anti-, criminal investigation and authorization, is that protection the people's lives and property is inviolable, conveniently
The important technology that functional department carries out the work.The concrete application of recognition of face may include Face datection and the inspection of face key point
It surveys, it may be assumed that provide piece image, return out the frame coordinate and facial main feature (such as eye, oral area of face position
Deng) coordinate.These concrete applications of recognition of face are directed to depth representative learning, cosine cross entropy loss function, large-spacing
Cross entropy loss function etc..
Wherein, depth representative learning is the foundation stone of face recognition algorithms, i.e., alignment facial image is mapped to a higher-dimension
Feature vector, this feature vector contain the identity information of face.A mechanism is introduced in the training of sorter network model: being utilized
Cosine cross entropy loss function calculates the mechanism of similarity between high dimensional feature vector using COS distance.Large-spacing intersects
Entropy loss function then from the decision boundary angle of classification, by introduce interval hyper parameter come so that different identity face
There is bigger interval, to obtain preferable Generalization Capability between feature.
Cosine cross entropy loss function and large-spacing cross entropy loss function are currently the hyper parameters being manually set, artificially
The hyper parameter of setting can largely influence the performance of final mask, for sorter network model, the model that trains
Classifying quality is undesirable, and the setting of hyper parameter there is no suitable theoretical direction.
In embodiment of the disclosure, sorter network model is trained using the hyper parameter of dynamic adjustment, is trained
Category of model effect it is even more ideal.By sorter network model can real-time perception current signature in characteristic vector space
Similarity distribution, to generate suitable class probability according to similarity distribution to carry out the training of sorter network model.
Fig. 1 shows the flow chart of the facial image recognition method according to the embodiment of the present disclosure, the facial image recognition method
Applied to facial image identification device, for example, facial image identification device can be by terminal device or server or other processing
Equipment executes, wherein terminal device can be user equipment (UE, User Equipment), mobile device, cellular phone, nothing
Rope phone, handheld device, calculates equipment, vehicle-mounted sets personal digital assistant (PDA, Personal Digital Assistant)
Standby, wearable device etc..In some possible implementations, facial image identification can call memory by processor
The mode of the computer-readable instruction of middle storage is realized.As shown in Figure 1, the process includes:
Step S101, multiple facial images are obtained.
In the possible implementation of the disclosure one, multiple facial images be can be from the same image, can also distinguish
From multiple images.
Step S102, the corresponding obtained multiple feature vectors of feature extraction are carried out according to multiple facial images, obtained more
A target object to be identified.
The disclosure one may in implementation, to multiple facial images carry out the corresponding obtained multiple features of feature extraction to
Amount, obtains multiple target objects to be identified, can carry out feature extraction to multiple facial images according to feature extraction network, obtain
To the corresponding multiple feature vectors of multiple facial images.In addition to feature extraction network, other networks can also be used, it can be real
Existing feature extraction, it is included in the protection scope of the disclosure.
Step S103, the classification according to belonging to the multiple feature vector, obtains sorting parameter.
Fig. 2 shows the schematic diagram according to the training data sample of the embodiment of the present disclosure, the disclosure one may in implementation,
Shown in Fig. 2, a vector cluster is made of same class feature vector, multiple vector clusters are constituted by vector cluster 11- vector cluster 13, to
The feature vector for including in amount cluster is identified target object.Feature vector 21 ..., feature vector 2m ..., feature
Vector 2n constitutes multiple target objects to be identified,.It needs to classify to multiple target object to be identified.To multiple
Target object to be identified, which carries out classification, has classification correct and two kinds of possibility of classification error, correct for classifying, and there is classification just
True class probability;For classification error, there are the class probabilities of classification error.
In the possible implementation of the disclosure one, according to classification belonging to the multiple feature vector, sorting parameter is obtained, is wrapped
It includes: according to the class number of the multiple feature vector generic, obtaining similarity parameter;The similarity parameter is carried out
Iterative processing obtains adaptive similarity parameter;The sorting parameter is obtained according to the adaptive similarity parameter.
In one example, according to the class number of multiple feature vector generics, the similarity parameter is obtained.In Fig. 2,
3 vector clusters, class number 3 are constituted by vector cluster 11- vector cluster 13.Adaptive training sorter network model, such as scaling
Similarity in training sorter network.By taking cosine similarity as an example, sorter network is inputted after initializing to cosine similarity
In, in each repetitive exercise of sorter network, the iteration for carrying out cosine similarity updates, and obtains the remaining of adaptive dynamic change
String similarity.It is updated according to the iteration of cosine similarity, dynamically generates suitable sorting parameter (such as class probability).According to classification
Parameter, as the distribution of class probability carries out the training of sorter network.Wherein, " cosine similarity " is used as hyper parameter, is adaptive
Dynamic change, it is not manually set, therefore, the classification dynamically generated will be updated according to the iteration of cosine similarity and is joined
Number, for multiple target objects to be identified (feature vector 21 in such as Fig. 2 ..., feature vector 2m ..., feature vector
2n constitutes multiple target objects to be identified) classification, more can accurately classify in recognition of face, to improve
The recognition effect of facial image.
In one example, similarity parameter can be used for characterizing cosine similarity, and similarity parameter is hyper parameter, such as phase
Like degree parameter be cosine similarity scale parameter or other are used to characterize the parameter of cosine similarity.For cosine similarity ruler
The initialization for spending parameter, can be calculated by the class number in training data.In Fig. 2, by vector cluster 11- vector cluster
13 constitute multiple vector clusters and feature vector 21 ..., feature vector 2m ..., feature vector 2n constitute it is multiple to be identified
Target object form training data.Wherein, m and n is positive integer, m < n, the feature vector separately included in multiple vector clusters
For identified target object.Corresponding 3 vector clusters being made of vector cluster 11- vector cluster 13, class number 3.Its
In, for the vector cluster for having completed to classify and sort out, the vector cluster being made of same class feature vector, the center of vector cluster
It is properly termed as class center.Similarity parameter can be the feature vector in training data and the ginseng of the similarity between all class centers
Number.
Step S104, classified according to the sorting parameter to the multiple target object to be identified, classified
As a result.
Sorter network in the possible implementation of the disclosure one, according to sorting parameter training sorter network, after being trained;
According to the sorter network after the training, classify to the multiple target object to be identified.
In one example, according to sorting parameter, loss function is obtained;The classification is trained according to loss function backpropagation
Network, the sorter network after being trained.According to the sorter network after the training, to the multiple target object to be identified
Classify.
The facial image recognition method of above-described embodiment, due to sorting parameter (such as class probability) be according to multiple features to
Amount generic (such as class number) obtains, i.e., sorting parameter is non-artificial setting, therefore, can be more in recognition of face
Precisely, to improve the recognition effect of facial image.
Fig. 3 shows the flow chart of the facial image recognition method according to the embodiment of the present disclosure, the facial image recognition method
Applied to facial image identification device, for example, facial image identification device can be by terminal device or server or other processing
Equipment executes, wherein terminal device can be user equipment (UE, User Equipment), mobile device, cellular phone, nothing
Rope phone, handheld device, calculates equipment, vehicle-mounted sets personal digital assistant (PDA, Personal Digital Assistant)
Standby, wearable device etc..In some possible implementations, facial image identification can call memory by processor
The mode of the computer-readable instruction of middle storage is realized.As shown in figure 3, the process includes:
Step S201, multiple facial images are obtained.
In the possible implementation of the disclosure one, multiple facial images be can be from the same image, can also distinguish
From multiple images.
Step S202, feature extraction is carried out to multiple facial images, obtains the corresponding multiple spies of multiple facial images
Levy vector.
In the possible implementation of the disclosure one, feature can be carried out to multiple facial images according to feature extraction network and mentioned
It takes, obtains the corresponding multiple feature vectors of multiple facial images.In addition to feature extraction network, other nets can also be used
Network is able to achieve feature extraction, is included in the protection scope of the disclosure.
Step S203, multiple target objects to be identified are obtained according to multiple feature vectors.
Fig. 2 shows the schematic diagrames that vector cluster is made of multiple feature vectors according to the embodiment of the present disclosure, and the disclosure one can
It is able to achieve in mode, shown in Fig. 2, a vector cluster is made of same class feature vector, be made of vector cluster 11- vector cluster 13 more
A vector cluster, the feature vector for including in vector cluster are identified target object.Feature vector 21 ..., feature to
It measures 2m ..., feature vector 2n and constitutes multiple target objects to be identified.It needs to carry out multiple target object to be identified
Classification.
Step S204, according to the class number of multiple feature vector generics, similarity parameter is obtained.
The disclosure one may be in implementation, similarity parameter, in current training data sample feature vector and institute
There is the similarity parameter between class center.Wherein, current training data sample includes: the class for having identified target object and having constituted, described
Multiple target objects to be identified.As shown in Fig. 2, the multiple vector clusters being made of vector cluster 11- vector cluster 13, and from feature to
Amount 21 ..., feature vector 2m ..., feature vector 2n constitute multiple target objects to be identified, one can be referred to as
Training data sample.Wherein, the feature vector for including in vector cluster is identified target object.In the phase of subsequent step
In repetitive exercise like degree distribution, classification can be more and more careful, such as based on the feature vector in above-mentioned training data sample, at present
It is divided into corresponding to and 3 vector clusters, class number 3 is constituted by vector cluster 11- vector cluster 13.It, can be further with repetitive exercise
3 or more vector clusters are classified as, class number is 3 or more, until, multiple target objects to be identified have all been classified
Finish.And/or the identified target object in vector cluster is further categorized into more accurate degree.
Step S205, processing is iterated to similarity parameter, obtains adaptive similarity parameter.
Step S206, the sorting parameter is obtained according to the adaptive similarity parameter.
Step S207, classified according to sorting parameter to multiple target objects to be identified, obtain classification results.
Classification in the possible implementation of the disclosure one, according to sorting parameter training sorter network, after being trained
Network;According to the sorter network after the training, classify to the multiple target object to be identified.
Using the disclosure, sorting parameter (such as class probability) is that the class number according to belonging to multiple feature vectors obtains
, i.e., sorting parameter is that non-artificial setting can be more smart in recognition of face then according to the sorter network of sorting parameter training
Standard after the sorter network after being trained, according to the sorter network after training, divides multiple target objects to be identified
The recognition effect of facial image has can be improved in class.
Fig. 4 be according in the embodiment of the present disclosure by training data sample (the training data sample being made of multiple feature vectors
This) the training process schematic diagram of adaptive similarity parameter is obtained, as shown in Figure 4, comprising:
Step S301, the training data sample being made of multiple feature vectors is obtained.
Step S302, training data sample is inputted into the similarity processing module in sorter network, obtains adaptive phase
Like degree parameter.
The processing stream that following steps S3021- step S3024 is constituted is run in similarity processing module in sorter network
Journey, the available adaptive similarity parameter.
Step S3021, according in this training data sample iteration between feature vector and its non-corresponding category feature vector
Distance, obtain COS distance.
Step S3022, according between the corresponding category feature vector of feature vector in this training data sample iteration
Angle obtains the median of characteristic feature vector Category Relevance.
As shown in Figure 5 for according to the schematic diagram of relationship between characteristic feature vector in the embodiment of the present disclosure.In Fig. 5,It is instruction
Practice some feature vector in data sample,Be withThe feature vector of corresponding classification,Be withNon-corresponding classification
Feature vector.
Step S3023, the similarity parameter obtained according to last training data sample iteration and this training data sample
The COS distance that this iteration obtains, obtains the sum of cosine similarity.
Step S3024, it according to the sum of cosine similarity and the median of characteristic feature vector Category Relevance, obtains certainly
The similarity parameter of adaptation.
After obtaining adaptive similarity parameter by training data sample, it can be obtained according to adaptive similarity parameter
Sorting parameter.
In one example, similarity parameter is initialized first, carries out phase on the basis of the initial value of similarity parameter
Like the repetitive exercise of degree.Secondly, in the training process, perceiving the dynamic change of cosine similarity.For example, in each repetitive exercise
In, it can be obtained according to the similarity parameter that previous iteration obtains and distance parameter (such as COS distance) that current iteration obtains
The sum of cosine similarity is carried out operation with obtained median, finally obtains adaptive similarity by the sum of cosine similarity
Parameter.The sorting parameter can be obtained according to the adaptive similarity parameter.Similarity parameter is with cosine similarity
Example in each repetitive exercise of sorter network, carries out cosine to being inputted in sorter network after cosine similarity initialization
The iteration of similarity updates, and obtains the cosine similarity of adaptive dynamic change.According to the distribution of similarity (cosine after training
The iteration of similarity updates), dynamically generate suitable sorting parameter (such as class probability).
In one example, according to sorting parameter (such as class probability), loss function is obtained.According to loss function, (such as cosine is handed over
Fork entropy loss function) backpropagation trains sorter network, sorter network after being trained.According to the sorter network after training
Classify to multiple target objects to be identified, obtains classification results.
Using the disclosure, it is related to " the classification for calculating cosine similarity (such as cosine similarity scale parameter) initial value
Number ".The updated cosine similarity of iteration is perceived according to the cosine similarity of successive ignition and COS distance, finally, is used
The updated cosine similarity of iteration obtains class probability, trains sorter network with class probability.Training obtains sorter network
Afterwards, in the application scenarios of recognition of face, classified according to the sorter network to multiple target objects to be identified.
Application example:
Using in one application example of the embodiment of the present disclosure, specific implementation is divided into two parts: first part is cosine similarity
The initialization of scale parameter can be calculated by the classification number in training data;Second part is by sorter network
The dynamic computing module of scale parameter calculates the dynamic change of cosine similarity scale parameter, can be and is generated by previous iteration
Scale parameter and the COS distance of high dimensional feature of current iteration calculate newly generated cosine similarity scale parameter.
Fig. 6 shows the class probability schematic diagram obtained according to the initialization similarity of the embodiment of the present disclosure.In Fig. 6, dotted line
16 indicate be initialization cosine similarity scale parameter s generate probability.Shown in the following formula of its calculation (1):
In formula (1),Class probability corresponding to cosine similarity scale parameter for initialization;BiBecome for centre
Amount, is the sum of current training data sample and the sized cosine similarity of all non-corresponding classes;cosθI, kFor similarity;C is
Class number shows that the initialization of cosine similarity scale parameter is only related to class number.After starting repetitive exercise, cosine
The dynamic change of similarity scale parameter calculates shown in following formula (2):
Cosine cross entropy loss function needs to calculate the remaining of feature vector and all class centers in current training data sample
String similarity.BI needs to be calculated when each repetitive exercise, after optimizing by repetitive exercise to sorter network, currently trains number
According to feature vector in sample close to class center, then sorter network has been carried out optimization, terminates repetitive exercise.In formula (2),
For BiAverage value, for indicate mistake classification value;cosθI, kFor similarity;In formula (3),For dynamic change cosine phase
The class probability like corresponding to degree scale parameter;For the angle of feature vector and corresponding class;For
The cosine value of smaller value in current corresponding angle median and pi/4, for characterizing the convergence situation of current class network.
By calculating probability to the second dervative of angle, available " probability-angle " curve changes most fast position,
This position should be reasonably set as the median of currently corresponding class similarity distribution.In view of feature vector and class center
Angle generally changes in 0 to 90 degree ranges, therefore, is advisable when cosine similarity scale parameter initializes using 45 degree.
In conclusion initializing first by class number C to cosine similarity scale parameter s, there is the angle of entry here
The priori that the intermediate value of degree distribution is 45 degree.When cosine similarity scale parameter carries out dynamic change, last cosine is utilized
Similarity scale parameter and training data sample this time, to each training data sample calculate belonging to BiValue recycles
Above-mentioned formula (3) calculates new cosine similarity scale parameter s.In conjunction with above-mentioned formula (2)-formula (3) it is found that dynamic change
Cosine similarity scale parameter s and BiValue and current angular median (in current training data sample feature vector with it is corresponding
The I d median of class center similarity) it is related.
Fig. 7 shows the schematic diagram that repetitive exercise is distributed according to the similarity of the embodiment of the present disclosure, and Fig. 7 is shown with training
The generic angle change situation of carry out.Under normal circumstances, angle is smaller, shows the more abundant of training.Lines 24 are using this
Open embodiment is trained, can allow inside classification assemble on cosine space it is even closer.In Fig. 5,The softmax loss function for indicating two norm constraints, for the side of feature vector and corresponding classification
The angle of parallactic angle (theta);Large-spacing cosine losses function is indicated, for feature vector and corresponding classification
Theta angle;Angle step separation loss function is indicated, for feature vector and corresponding classification
The angle of theta;Indicate dynamic adaptive scaling cosine cross entropy loss function,
For the angle of feature vector and the theta of corresponding classification;θI, j, j≠yiOf Dynamic AdaScale is indicated dynamically certainly
Scaling cosine cross entropy loss function is adapted to, for the angle of feature vector and the theta of corresponding classification.The embodiment of the present disclosure is suitable
For extensive video monitoring, customs, airport etc. carry out authentication, do empowerment management and security department's view for company, mechanism
Frequency monitoring system, identity authorization system, hotel company the application scenarios such as guest system in, calculated using perception COS distance
Scale parameter carries out, so that the class probability of Optimum Classification network is with training this depth sorting model of the sorter network.
It will be understood by those skilled in the art that each step writes sequence simultaneously in the above method of specific embodiment
It does not mean that stringent execution sequence and any restriction is constituted to implementation process, the specific execution sequence of each step should be with its function
It can be determined with possible internal logic.
Above-mentioned each embodiment of the method that the disclosure refers to can phase each other without prejudice to principle logic
The embodiment formed after combining is mutually combined, as space is limited, the disclosure repeats no more.
In addition, the disclosure additionally provides facial image identification device, electronic equipment, computer readable storage medium, program,
The above-mentioned any facial image recognition method that can be used to realize disclosure offer, corresponding technical solution is with description and referring to side
The corresponding record of method part, repeats no more.
Fig. 8 shows the block diagram of the facial image identification device according to the embodiment of the present disclosure, as shown in figure 8, the disclosure is implemented
The facial image identification device of example, comprising: image collection module 41, for obtaining multiple facial images;Semantic object extraction mould
Block 42, for carrying out the corresponding obtained multiple feature vectors of feature extraction according to the multiple facial image, obtain it is multiple to
The target object of identification;Parameter acquisition module 43 obtains classification ginseng for the classification according to belonging to the multiple feature vector
Number;Categorization module 44 is classified for being classified according to the sorting parameter to the multiple target object to be identified
As a result.
In the possible implementation of the disclosure, the parameter acquisition module is used for: according to belonging to the multiple feature vector
The class number of classification obtains similarity parameter;Processing is iterated to the similarity parameter, obtains adaptive similarity
Parameter;The sorting parameter is obtained according to the adaptive similarity parameter.
In the possible implementation of the disclosure, the categorization module is used for: according to sorting parameter training classification net
Network, the sorter network after being trained;According to the sorter network after the training, to the multiple target object to be identified into
Row classification.
In the possible implementation of the disclosure, the categorization module is used for: according to the sorting parameter, obtaining loss letter
Number;The sorter network is trained according to the loss function backpropagation, the sorter network after obtaining the training.
In the possible implementation of the disclosure, described device further include:
Sample acquisition module, for obtaining the training data sample being made of the multiple feature vector;
Processing module is obtained for the training data sample to be inputted the similarity processing module in the sorter network
To adaptive similarity parameter;
Sorting parameter determining module, for obtaining the sorting parameter according to the adaptive similarity parameter.
In the possible implementation of the disclosure, the processing module is used for:
The similarity parameter and this training data sample iteration obtained according to last training data sample iteration obtains
COS distance, obtain the sum of cosine similarity;
According in the sum of described cosine similarity and this training data sample in characteristic feature vector Category Relevance
Digit obtains the adaptive similarity parameter.
In the possible implementation of the disclosure, the COS distance, according to feature in this training data sample iteration to
It measures at a distance between its non-corresponding category feature vector and obtains.
In the possible implementation of the disclosure, the median, according to feature vector in this training data sample iteration
Angle between corresponding category feature vector obtains.
In some embodiments, the embodiment of the present disclosure provides the function that has of device or comprising module can be used for holding
The method of row embodiment of the method description above, specific implementation are referred to the description of embodiment of the method above, for sake of simplicity, this
In repeat no more.
The embodiment of the present disclosure also proposes a kind of computer readable storage medium, is stored thereon with computer program instructions, institute
It states when computer program instructions are executed by processor and realizes the above method.Computer readable storage medium can be non-volatile meter
Calculation machine readable storage medium storing program for executing.
The embodiment of the present disclosure also proposes a kind of electronic equipment, comprising: processor;For storage processor executable instruction
Memory;Wherein, the processor is configured to the above method.
The equipment that electronic equipment may be provided as terminal, server or other forms.
Fig. 9 is the block diagram of a kind of electronic equipment 800 shown according to an exemplary embodiment.For example, electronic equipment 800 can
To be mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, Medical Devices are good for
Body equipment, the terminals such as personal digital assistant.
Referring to Fig. 9, electronic equipment 800 may include following one or more components: processing component 802, memory 804,
Power supply module 806, multimedia component 808, audio component 810, the interface 812 of input/output (I/O), sensor module 814,
And communication component 816.
The integrated operation of the usual controlling electronic devices 800 of processing component 802, such as with display, call, data are logical
Letter, camera operation and record operate associated operation.Processing component 802 may include one or more processors 820 to hold
Row instruction, to perform all or part of the steps of the methods described above.In addition, processing component 802 may include one or more moulds
Block, convenient for the interaction between processing component 802 and other assemblies.For example, processing component 802 may include multi-media module, with
Facilitate the interaction between multimedia component 808 and processing component 802.
Memory 804 is configured as storing various types of data to support the operation in electronic equipment 800.These data
Example include any application or method for being operated on electronic equipment 800 instruction, contact data, telephone directory
Data, message, picture, video etc..Memory 804 can by any kind of volatibility or non-volatile memory device or it
Combination realize, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM) is erasable
Except programmable read only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, fastly
Flash memory, disk or CD.
Power supply module 806 provides electric power for the various assemblies of electronic equipment 800.Power supply module 806 may include power supply pipe
Reason system, one or more power supplys and other with for electronic equipment 800 generate, manage, and distribute the associated component of electric power.
Multimedia component 808 includes the screen of one output interface of offer between the electronic equipment 800 and user.
In some embodiments, screen may include liquid crystal display (LCD) and touch panel (TP).If screen includes touch surface
Plate, screen may be implemented as touch screen, to receive input signal from the user.Touch panel includes one or more touches
Sensor is to sense the gesture on touch, slide, and touch panel.The touch sensor can not only sense touch or sliding
The boundary of movement, but also detect duration and pressure associated with the touch or slide operation.In some embodiments,
Multimedia component 808 includes a front camera and/or rear camera.When electronic equipment 800 is in operation mode, as clapped
When taking the photograph mode or video mode, front camera and/or rear camera can receive external multi-medium data.It is each preposition
Camera and rear camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 810 is configured as output and/or input audio signal.For example, audio component 810 includes a Mike
Wind (MIC), when electronic equipment 800 is in operation mode, when such as call mode, recording mode, and voice recognition mode, microphone
It is configured as receiving external audio signal.The received audio signal can be further stored in memory 804 or via logical
Believe that component 816 is sent.In some embodiments, audio component 810 further includes a loudspeaker, is used for output audio signal.
I/O interface 812 provides interface between processing component 802 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include, but are not limited to: home button, volume button, start button and lock
Determine button.
Sensor module 814 includes one or more sensors, for providing the state of various aspects for electronic equipment 800
Assessment.For example, sensor module 814 can detecte the state that opens/closes of electronic equipment 800, the relative positioning of component, example
As the component be electronic equipment 800 display and keypad, sensor module 814 can also detect electronic equipment 800 or
The position change of 800 1 components of electronic equipment, the existence or non-existence that user contacts with electronic equipment 800, electronic equipment 800
The temperature change of orientation or acceleration/deceleration and electronic equipment 800.Sensor module 814 may include proximity sensor, be configured
For detecting the presence of nearby objects without any physical contact.Sensor module 814 can also include optical sensor,
Such as CMOS or ccd image sensor, for being used in imaging applications.In some embodiments, which may be used also
To include acceleration transducer, gyro sensor, Magnetic Sensor, pressure sensor or temperature sensor.
Communication component 816 is configured to facilitate the communication of wired or wireless way between electronic equipment 800 and other equipment.
Electronic equipment 800 can access the wireless network based on communication standard, such as WiFi, 2G or 3G or their combination.Show at one
In example property embodiment, communication component 816 receives broadcast singal or broadcast from external broadcasting management system via broadcast channel
Relevant information.In one exemplary embodiment, the communication component 816 further includes near-field communication (NFC) module, short to promote
Cheng Tongxin.For example, radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band can be based in NFC module
(UWB) technology, bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, electronic equipment 800 can be by one or more application specific integrated circuit (ASIC), number
Word signal processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic components are realized, for executing the above method.
In the exemplary embodiment, a kind of non-volatile computer readable storage medium storing program for executing is additionally provided, for example including calculating
The memory 804 of machine program instruction, above-mentioned computer program instructions can be executed by the processor 820 of electronic equipment 800 to complete
The above method.
Figure 10 is the block diagram of a kind of electronic equipment 900 shown according to an exemplary embodiment.For example, electronic equipment 900
It may be provided as a server.Referring to Fig.1 0, electronic equipment 900 includes processing component 922, further comprises one or more
A processor, and the memory resource as representated by memory 932, can be by the finger of the execution of processing component 922 for storing
It enables, such as application program.The application program stored in memory 932 may include it is one or more each correspond to
The module of one group of instruction.In addition, processing component 922 is configured as executing instruction, to execute the above method.
Electronic equipment 900 can also include that a power supply module 926 is configured as executing the power supply pipe of electronic equipment 900
Reason, a wired or wireless network interface 950 are configured as electronic equipment 1900 being connected to network and an input and output
(I/O) interface 958.Electronic equipment 900 can be operated based on the operating system for being stored in memory 932, such as Windows
ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or similar.
In the exemplary embodiment, a kind of non-volatile computer readable storage medium storing program for executing is additionally provided, for example including calculating
The memory 932 of machine program instruction, above-mentioned computer program instructions can be executed by the processing component 922 of electronic equipment 900 with complete
At the above method.
The disclosure can be system, method and/or computer program product.Computer program product may include computer
Readable storage medium storing program for executing, containing for making processor realize the computer-readable program instructions of various aspects of the disclosure.
Computer readable storage medium, which can be, can keep and store the tangible of the instruction used by instruction execution equipment
Equipment.Computer readable storage medium for example can be-- but it is not limited to-- storage device electric, magnetic storage apparatus, optical storage
Equipment, electric magnetic storage apparatus, semiconductor memory apparatus or above-mentioned any appropriate combination.Computer readable storage medium
More specific example (non exhaustive list) includes: portable computer diskette, hard disk, random access memory (RAM), read-only deposits
It is reservoir (ROM), erasable programmable read only memory (EPROM or flash memory), static random access memory (SRAM), portable
Compact disk read-only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanical coding equipment, for example thereon
It is stored with punch card or groove internal projection structure and the above-mentioned any appropriate combination of instruction.Calculating used herein above
Machine readable storage medium storing program for executing is not interpreted that instantaneous signal itself, the electromagnetic wave of such as radio wave or other Free propagations lead to
It crosses the electromagnetic wave (for example, the light pulse for passing through fiber optic cables) of waveguide or the propagation of other transmission mediums or is transmitted by electric wire
Electric signal.
Computer-readable program instructions as described herein can be downloaded to from computer readable storage medium it is each calculate/
Processing equipment, or outer computer or outer is downloaded to by network, such as internet, local area network, wide area network and/or wireless network
Portion stores equipment.Network may include copper transmission cable, optical fiber transmission, wireless transmission, router, firewall, interchanger, gateway
Computer and/or Edge Server.Adapter or network interface in each calculating/processing equipment are received from network to be counted
Calculation machine readable program instructions, and the computer-readable program instructions are forwarded, for the meter being stored in each calculating/processing equipment
In calculation machine readable storage medium storing program for executing.
Computer program instructions for executing disclosure operation can be assembly instruction, instruction set architecture (ISA) instructs,
Machine instruction, machine-dependent instructions, microcode, firmware instructions, condition setup data or with one or more programming languages
The source code or object code that any combination is write, the programming language include the programming language-of object-oriented such as
Smalltalk, C++ etc., and conventional procedural programming languages-such as " C " language or similar programming language.Computer
Readable program instructions can be executed fully on the user computer, partly execute on the user computer, be only as one
Vertical software package executes, part executes on the remote computer or completely in remote computer on the user computer for part
Or it is executed on server.In situations involving remote computers, remote computer can pass through network-packet of any kind
It includes local area network (LAN) or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as benefit
It is connected with ISP by internet).In some embodiments, by utilizing computer-readable program instructions
Status information carry out personalized customization electronic circuit, such as programmable logic circuit, field programmable gate array (FPGA) or can
Programmed logic array (PLA) (PLA), the electronic circuit can execute computer-readable program instructions, to realize each side of the disclosure
Face.
Referring herein to according to the flow chart of the method, apparatus (system) of the embodiment of the present disclosure and computer program product and/
Or block diagram describes various aspects of the disclosure.It should be appreciated that flowchart and or block diagram each box and flow chart and/
Or in block diagram each box combination, can be realized by computer-readable program instructions.
These computer-readable program instructions can be supplied to general purpose computer, special purpose computer or other programmable datas
The processor of processing unit, so that a kind of machine is produced, so that these instructions are passing through computer or other programmable datas
When the processor of processing unit executes, function specified in one or more boxes in implementation flow chart and/or block diagram is produced
The device of energy/movement.These computer-readable program instructions can also be stored in a computer-readable storage medium, these refer to
It enables so that computer, programmable data processing unit and/or other equipment work in a specific way, thus, it is stored with instruction
Computer-readable medium then includes a manufacture comprising in one or more boxes in implementation flow chart and/or block diagram
The instruction of the various aspects of defined function action.
Computer-readable program instructions can also be loaded into computer, other programmable data processing units or other
In equipment, so that series of operation steps are executed in computer, other programmable data processing units or other equipment, to produce
Raw computer implemented process, so that executed in computer, other programmable data processing units or other equipment
Instruct function action specified in one or more boxes in implementation flow chart and/or block diagram.
The flow chart and block diagram in the drawings show system, method and the computer journeys according to multiple embodiments of the disclosure
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
One module of table, program segment or a part of instruction, the module, program segment or a part of instruction include one or more use
The executable instruction of the logic function as defined in realizing.In some implementations as replacements, function marked in the box
It can occur in a different order than that indicated in the drawings.For example, two continuous boxes can actually be held substantially in parallel
Row, they can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram and/or
The combination of each box in flow chart and the box in block diagram and or flow chart, can the function as defined in executing or dynamic
The dedicated hardware based system made is realized, or can be realized using a combination of dedicated hardware and computer instructions.
The presently disclosed embodiments is described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or technological improvement to technology in market for best explaining each embodiment, or make the art
Other those of ordinary skill can understand each embodiment disclosed herein.
Claims (10)
1. a kind of facial image recognition method characterized by comprising
Obtain multiple facial images;
The corresponding obtained multiple feature vectors of feature extraction are carried out according to the multiple facial image, are obtained multiple to be identified
Target object;
According to classification belonging to the multiple feature vector, sorting parameter is obtained;
Classified according to the sorting parameter to the multiple target object to be identified, obtains classification results.
2. the method according to claim 1, wherein the classification according to belonging to the multiple feature vector,
Obtain sorting parameter, comprising:
According to the class number of the multiple feature vector generic, similarity parameter is obtained;
Processing is iterated to the similarity parameter, obtains adaptive similarity parameter;
The sorting parameter is obtained according to the adaptive similarity parameter.
3. method according to claim 1 or 2, which is characterized in that it is described according to the sorting parameter to it is the multiple to
The target object of identification is classified, comprising:
Sorter network according to sorting parameter training sorter network, after being trained;
According to the sorter network after the training, classify to the multiple target object to be identified.
4. according to the method described in claim 3, it is characterized in that, described according to the sorting parameter training classification net
Network, the sorter network after being trained, comprising:
According to the sorting parameter, loss function is obtained;
The sorter network is trained according to the loss function backpropagation, the sorter network after obtaining the training.
5. according to the method described in claim 4, it is characterized in that, before according to the sorting parameter training sorter network
Training process further include:
The training data sample being made of the multiple feature vector is obtained, the training data sample is inputted into the classification net
Similarity processing module in network obtains adaptive similarity parameter;
The sorting parameter is obtained according to the adaptive similarity parameter.
6. according to the method described in claim 5, it is characterized in that, the training data sample is inputted in the sorter network
Similarity processing module, obtain adaptive similarity parameter, comprising:
More than the similarity parameter and this training data sample iteration obtained according to last training data sample iteration obtains
Chordal distance obtains the sum of cosine similarity;
According to the middle position of characteristic feature vector Category Relevance in the sum of described cosine similarity and this training data sample
Number, obtains the adaptive similarity parameter.
7. according to the method described in claim 6, it is characterized in that, the COS distance, changes according to this training data sample
Generation in feature vector between its non-corresponding category feature vector at a distance from obtain.
8. a kind of facial image identification device characterized by comprising
Image collection module, for obtaining multiple facial images;
Semantic object extraction module, for carrying out the corresponding obtained multiple features of feature extraction according to the multiple facial image
Vector obtains multiple target objects to be identified;
Parameter acquisition module obtains sorting parameter for the classification according to belonging to the multiple feature vector;
Categorization module is classified for being classified according to the sorting parameter to the multiple target object to be identified
As a result.
9. a kind of electronic equipment characterized by comprising
Processor;
Memory for storage processor executable instruction;
Wherein, the processor is configured to: perform claim require any one of 1 to 7 described in method.
10. a kind of computer readable storage medium, is stored thereon with computer program instructions, which is characterized in that the computer
Method described in any one of claim 1 to 7 is realized when program instruction is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910330767.3A CN110070049A (en) | 2019-04-23 | 2019-04-23 | Facial image recognition method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910330767.3A CN110070049A (en) | 2019-04-23 | 2019-04-23 | Facial image recognition method and device, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110070049A true CN110070049A (en) | 2019-07-30 |
Family
ID=67368588
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910330767.3A Pending CN110070049A (en) | 2019-04-23 | 2019-04-23 | Facial image recognition method and device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110070049A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110619602A (en) * | 2019-09-25 | 2019-12-27 | 北京字节跳动网络技术有限公司 | Image generation method and device, electronic equipment and storage medium |
CN111368644A (en) * | 2020-02-14 | 2020-07-03 | 深圳市商汤科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
CN112559784A (en) * | 2020-11-02 | 2021-03-26 | 浙江智慧视频安防创新中心有限公司 | Image classification method and system based on incremental learning |
CN112597979A (en) * | 2021-03-03 | 2021-04-02 | 之江实验室 | Face recognition method for updating cosine included angle loss function parameters in real time |
CN113449668A (en) * | 2021-07-08 | 2021-09-28 | 杭州迅蚁网络科技有限公司 | Target angle identification method and device of flight device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109165566A (en) * | 2018-08-01 | 2019-01-08 | 中国计量大学 | A kind of recognition of face convolutional neural networks training method based on novel loss function |
CN109522872A (en) * | 2018-12-04 | 2019-03-26 | 西安电子科技大学 | A kind of face identification method, device, computer equipment and storage medium |
US20190114747A1 (en) * | 2016-04-07 | 2019-04-18 | Carmel Haifa University Economic Corporation Ltd. | Image dehazing and restoration |
-
2019
- 2019-04-23 CN CN201910330767.3A patent/CN110070049A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190114747A1 (en) * | 2016-04-07 | 2019-04-18 | Carmel Haifa University Economic Corporation Ltd. | Image dehazing and restoration |
CN109165566A (en) * | 2018-08-01 | 2019-01-08 | 中国计量大学 | A kind of recognition of face convolutional neural networks training method based on novel loss function |
CN109522872A (en) * | 2018-12-04 | 2019-03-26 | 西安电子科技大学 | A kind of face identification method, device, computer equipment and storage medium |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110619602A (en) * | 2019-09-25 | 2019-12-27 | 北京字节跳动网络技术有限公司 | Image generation method and device, electronic equipment and storage medium |
CN110619602B (en) * | 2019-09-25 | 2024-01-09 | 北京字节跳动网络技术有限公司 | Image generation method and device, electronic equipment and storage medium |
CN111368644A (en) * | 2020-02-14 | 2020-07-03 | 深圳市商汤科技有限公司 | Image processing method, image processing device, electronic equipment and storage medium |
CN111368644B (en) * | 2020-02-14 | 2024-01-05 | 深圳市商汤科技有限公司 | Image processing method, device, electronic equipment and storage medium |
CN112559784A (en) * | 2020-11-02 | 2021-03-26 | 浙江智慧视频安防创新中心有限公司 | Image classification method and system based on incremental learning |
CN112559784B (en) * | 2020-11-02 | 2023-07-04 | 浙江智慧视频安防创新中心有限公司 | Image classification method and system based on incremental learning |
CN112597979A (en) * | 2021-03-03 | 2021-04-02 | 之江实验室 | Face recognition method for updating cosine included angle loss function parameters in real time |
CN113449668A (en) * | 2021-07-08 | 2021-09-28 | 杭州迅蚁网络科技有限公司 | Target angle identification method and device of flight device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110070049A (en) | Facial image recognition method and device, electronic equipment and storage medium | |
CN110348537A (en) | Image processing method and device, electronic equipment and storage medium | |
CN110210535A (en) | Neural network training method and device and image processing method and device | |
CN110363150A (en) | Data-updating method and device, electronic equipment and storage medium | |
CN109800744A (en) | Image clustering method and device, electronic equipment and storage medium | |
CN109800737A (en) | Face recognition method and device, electronic equipment and storage medium | |
CN109740516A (en) | A kind of user identification method, device, electronic equipment and storage medium | |
CN109089133A (en) | Method for processing video frequency and device, electronic equipment and storage medium | |
CN109614613A (en) | The descriptive statement localization method and device of image, electronic equipment and storage medium | |
CN109871883A (en) | Neural network training method and device, electronic equipment and storage medium | |
CN109658572A (en) | Image processing method and device, electronic equipment and storage medium | |
CN110503023A (en) | Biopsy method and device, electronic equipment and storage medium | |
CN108764069A (en) | Biopsy method and device | |
CN110458102A (en) | A kind of facial image recognition method and device, electronic equipment and storage medium | |
CN110472091A (en) | Image processing method and device, electronic equipment and storage medium | |
CN109816764A (en) | Image generating method and device, electronic equipment and storage medium | |
CN109801270A (en) | Anchor point determines method and device, electronic equipment and storage medium | |
CN110298310A (en) | Image processing method and device, electronic equipment and storage medium | |
JP7074877B2 (en) | Network optimization methods and equipment, image processing methods and equipment, storage media and computer programs | |
CN108921117A (en) | Image processing method and device, electronic equipment and storage medium | |
CN109934275A (en) | Image processing method and device, electronic equipment and storage medium | |
CN108985176A (en) | image generating method and device | |
CN109711546A (en) | Neural network training method and device, electronic equipment and storage medium | |
CN110502651A (en) | Image processing method and device, electronic equipment and storage medium | |
CN110287671A (en) | Verification method and device, electronic equipment and storage medium |
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 |