CN111985323A - Face recognition method and system based on deep convolutional neural network - Google Patents
Face recognition method and system based on deep convolutional neural network Download PDFInfo
- Publication number
- CN111985323A CN111985323A CN202010675730.7A CN202010675730A CN111985323A CN 111985323 A CN111985323 A CN 111985323A CN 202010675730 A CN202010675730 A CN 202010675730A CN 111985323 A CN111985323 A CN 111985323A
- Authority
- CN
- China
- Prior art keywords
- face
- recognition
- age
- cutting
- feature vector
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/168—Feature extraction; Face representation
- G06V40/171—Local features and components; Facial parts ; Occluding parts, e.g. glasses; Geometrical relationships
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06V40/178—Human faces, e.g. facial parts, sketches or expressions estimating age from face image; using age information for improving recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a face recognition method and a face recognition system based on a deep convolutional neural network, wherein the method comprises the following steps: the method comprises the steps of detecting and cutting, namely receiving a face photo, obtaining coordinates of key facial feature points, and determining a cutting boundary according to the coordinates of the key facial feature points based on a preset cutting rule; a feature extraction step, namely acquiring a face feature vector with a fixed channel number according to the cut picture based on a backbone network of a face depth recognition model; an identity recognition step, namely matching the face feature vector and acquiring identity recognition information corresponding to the face feature vector according to the similarity; and an age identification step, namely acquiring identification ages based on age identification models of a plurality of age classifications according to the face feature vectors. The invention can improve the recognition processing speed, improve the accuracy of face and age recognition, and save manpower and material resources by reusing the face recognition model.
Description
Technical Field
The invention relates to the technical field of face recognition, in particular to a face recognition method and system based on a deep convolutional neural network.
Background
Face recognition is a biometric technology for identity recognition based on facial feature information of a person. With the maturity of the technology and the improvement of social acceptance, face recognition is widely applied in various fields.
The face age recognition is used as a sub-system of the face recognition, the calculation of the age is usually placed on a mobile device, the calculation performance of hardware of the mobile device is considered, a non-deep learning mechanism is adopted, the accuracy rate is poor, meanwhile, in specific application, the age prediction can usually wait for the recognition result of the face based on the deep learning to return together, and the delay of the feedback result can be increased.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art. Therefore, the invention provides a face recognition method based on a deep convolutional neural network, which can improve the accuracy of face age recognition.
The invention also provides a face recognition system based on the deep convolutional neural network, which has the face recognition method based on the deep convolutional neural network.
The face recognition method based on the deep convolutional neural network comprises the following steps: the method comprises the steps of detecting and cutting, namely receiving a face photo, detecting facial features in the face photo, obtaining coordinates of facial key feature points, determining a cutting boundary according to the coordinates of the facial key feature points based on a preset cutting rule, and obtaining a cut picture, wherein the cut picture comprises a facial recognition area and an edge area surrounding the facial recognition area; a feature extraction step, namely acquiring a face feature vector with a fixed channel number according to the cut picture based on a backbone network of a face depth recognition model; an identity recognition step, namely matching the face feature vector and acquiring identity recognition information corresponding to the face feature vector according to the similarity; and an age identification step, namely acquiring identification ages based on age identification models of a plurality of age classifications according to the face feature vectors.
The face recognition method based on the deep convolutional neural network provided by the embodiment of the invention at least has the following beneficial effects: by cutting the face photo, the cutting picture with fixed specification can improve the recognition processing speed, and the retention of the edge area can improve the recognition accuracy of the face and the age; the face feature vectors processed in face recognition are obtained by multiplexing the face recognition model, the accuracy of age recognition is improved, a deep age recognition learning model does not need to be repeatedly constructed, and manpower and material resources are saved.
According to some embodiments of the invention, the detecting clipping step comprises: a key point extraction step, namely performing multi-scale change according to the face picture to construct an image pyramid, constructing a plurality of candidate frames in the image pyramid, filtering the candidate frames, screening out the candidate frames with a face area, generating a face recognition area, and obtaining coordinates of key feature points of the face; and a picture cutting step, wherein the face picture is cut according to a preset cutting rule according to the coordinates of the key feature points of the face, wherein the preset cutting rule comprises the size of the cut picture and the distance between the face recognition area and a cutting frame in the cut picture.
According to some embodiments of the present invention, the resolution size of the cropped picture is 112 × 112, wherein the face recognition region is 16 away from the cropping frame.
According to some embodiments of the invention, the age identifying step comprises: taking the face feature vector as input data, and obtaining the probability corresponding to each age classification of the age identification model according to the age identification model; and obtaining the identification age according to the age classification and the probability.
According to some embodiments of the present invention, the fixed channel number of the face feature vector is 512 dimensions.
According to some embodiments of the invention, further comprising: a face recognition model training step, based on an MS-Celeb-1M face data set, acquiring the face feature vector through the detecting and cutting step and the feature extraction step, and training the face depth recognition model; and an age recognition model training step, wherein the face feature vector is obtained through the detection cutting step and the feature extraction step based on a MORPH II face age data set and an FG-NET face age data set, and the age recognition model is trained.
According to the second aspect of the invention, the face recognition system based on the deep convolutional neural network comprises: the detection cutting module is used for receiving a face photo, detecting facial features in the face photo, obtaining coordinates of facial key feature points, determining a cutting boundary according to the coordinates of the facial key feature points based on a preset cutting rule, and obtaining a cut picture, wherein the cut picture comprises the facial recognition area and an edge area surrounding the facial recognition area; the feature extraction module is used for acquiring a face feature vector with a fixed channel number according to the cut picture based on a backbone network of a face depth recognition model; the identity recognition module is used for matching the face feature vector and acquiring identity recognition information corresponding to the face feature vector according to the similarity; and the age identification module is used for acquiring the identification age based on the age identification models of a plurality of age classifications according to the face feature vector.
The face recognition system based on the deep convolutional neural network provided by the embodiment of the invention at least has the following beneficial effects: by cutting the face photo, the cutting picture with fixed specification can improve the recognition processing speed, and the retention of the edge area can improve the recognition accuracy of the face and the age; the face feature vectors processed in face recognition are obtained by multiplexing the face recognition model, the accuracy of age recognition is improved, a deep age recognition learning model does not need to be repeatedly constructed, and manpower and material resources are saved.
According to some embodiments of the invention, the detection clipping module comprises: the key point extraction module is used for carrying out multi-scale change according to the face photos, constructing an image pyramid, constructing a plurality of candidate frames in the image pyramid, filtering the candidate frames, screening out the candidate frames with face areas, generating face recognition areas and obtaining coordinates of key feature points of the faces; and the picture cutting module is used for cutting the face picture according to the coordinates of the key feature points of the face and a preset cutting rule, wherein the preset cutting rule comprises the size of the cut picture and the distance between the face recognition area and a cutting frame in the cut picture.
According to some embodiments of the invention, further comprising: and the face feature vector management module is used for managing the face feature vectors generated based on the face depth recognition model.
According to some embodiments of the invention, further comprising: the face recognition training module is used for acquiring the face feature vector through the detection cutting module and the feature extraction module based on an MS-Celeb-1M face data set and training the face depth recognition model; and the age recognition training module is used for acquiring the face feature vector through the detection cutting module and the feature extraction module based on the MORPH II face age data set and the FG-NET face age data set, and training the age recognition model.
Additional aspects and advantages of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a schematic flow chart of a method according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart illustrating a step of detecting clipping in the method according to the embodiment of the present invention;
FIG. 3 is a schematic diagram of a truncation in a method according to an embodiment of the present invention;
FIG. 4 is one of the block schematic block diagrams of the system of an embodiment of the present invention;
FIG. 5 is a second block diagram of the system according to the second embodiment of the present invention;
fig. 6 is a schematic diagram illustrating a processing procedure of the system according to the embodiment of the present invention.
Reference numerals:
400. A face feature vector management module 500, a face recognition training module 600, and an age recognition training module 700;
a key point extraction module 110 and a picture clipping module 120.
Detailed Description
Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the accompanying drawings are illustrative only for the purpose of explaining the present invention, and are not to be construed as limiting the present invention.
In the description of the present invention, the meaning of a plurality of means is one or more, the meaning of a plurality of means is two or more, and more than, less than, more than, etc. are understood as excluding the present number, and more than, less than, etc. are understood as including the present number. If the first and second are described for the purpose of distinguishing technical features, they are not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
The noun explains:
deep convolutional neural network ResNet of 100 layers: a deep convolutional neural network structure.
MS-Celeb-1M: microsoft published face data sets.
MORPH II: a public face age data set.
FG-NET: a public face age data set.
MTCNN model: a face detection method.
Arcface: a loss function method for face recognition.
Referring to fig. 1, a method of an embodiment of the invention includes: and a step of detecting and cutting, namely receiving and processing the face photo, detecting face facial features in the photo, obtaining coordinates (including two eyes, a nose and a left mouth corner and a right mouth corner) of facial key feature points, and then determining a cutting boundary according to a sitting picture of the facial key feature points based on a preset cutting rule to obtain a cut picture. Referring to fig. 3, the clipped region includes an innermost face recognition region and an edge region; the border area surrounds the face recognition area, and the border area may contain information such as hair, beard and the like which is beneficial to identity recognition and age recognition. And then, extracting the features of the cut human face picture, and acquiring a human face feature vector with a fixed channel number based on a backbone network of a human face depth recognition model. The face feature vectors are used for identity recognition and age recognition, respectively. In the step of identity recognition, identity recognition information (such as name, gender, ethnic group and the like) is obtained according to the similarity by matching the face feature vectors; in the age identification step, the face feature vector is input, and an identification age is obtained based on an age identification model of a plurality of age classifications.
In the method of the embodiment of the present invention, the detecting and cutting step is described with reference to fig. 2, and includes: a key point extraction step, namely performing multi-scale change according to an original face photo to construct an image pyramid, constructing a plurality of candidate frames in the image pyramid, filtering the candidate frames step by step to screen out the candidate frames with a face region, generating a face recognition region and obtaining coordinates of key feature points of the face; and a picture cutting step, wherein the face picture is cut according to the coordinates of the key feature points of the face and a preset cutting rule, wherein the preset cutting rule comprises the size of the cut picture and the limit distance between a face recognition area and a cutting frame in the cut picture. The resolution size of the cropped picture is set to 112 × 112, with the limit distance of the face recognition area from the cropping frame being 16. The resolution can better extract general characteristics with discrimination, and the distance setting of the face recognition area and the cutting frame can keep the information which is beneficial to recognition in the background on one hand, and on the other hand, the face keeps a certain proportion in the cut picture, so that the recognition accuracy can be improved, and the recognition and judgment result can be prevented from being influenced.
In the embodiment of the invention, the age identification model is a multi-classification age identification model, each classification corresponds to one age classification and has a corresponding age value. The face feature vector is used as input data, and the probability corresponding to each age classification of the age identification model can be obtained through a neural convolution network of the age identification model; then, according to the age classification and the probability, the identification age is obtained, and the calculation method comprises the following steps: n ═ Σ (N)i*pi) Wherein N represents the recognition age, NiIndicates the age, p, corresponding to the ith age categoryiRepresenting the probability of the ith age category.
In one embodiment of the invention, the training process of the face depth recognition model in face identity recognition is as follows. The method comprises the steps of adopting a deep convolutional neural network ResNet with 100 layers as a main network, using a million face data set MS-Celeb-1M to train a face recognition model to obtain a face deep recognition model for face identity recognition, and enabling each face photo to generate a face feature vector with a fixed output channel number after passing through the main network of the face deep recognition model through parameter weight fixation with 100 layers. In the embodiment of the present invention, the face feature vector is 512 dimensions; it should be understood that in some embodiments of the present invention, the number of channels may also be 128 or 256, that is, the output face feature vector is 128 dimensions or 256 dimensions, etc. Firstly, extracting and cutting a face of a face data set based on an MTCNN model, inputting original face photo data, and outputting 5 key coordinate points (two eyes, a nose, a left mouth angle and a right mouth angle) corresponding to each face photo, wherein the processing process comprises the following steps: carrying out multi-scale change on an original face photo to generate an image pyramid, dividing the image pyramid into a plurality of candidate frames, carrying out frame regression and key point positioning on the candidate frames through a plurality of cascaded convolutional neural networks based on a frame regression and a locator of face key points, filtering step by step according to the possibility of existence of the face position to obtain a face recognition area, and finally outputting coordinates of the face key feature points. Finally, based on a preset clipping rule, acquiring a clipped picture including a face recognition area according to the coordinates of the face key feature points according to a certain resolution ratio, wherein the resolution of the clipped picture is set to be 112 multiplied by 112, and the limit distance (also called edge distance) between the face recognition area and a clipping frame is 16; a cropped picture with uniform size rules (112 × 112) is obtained, and the distance between the face recognition area and the cropping frame does not exceed 16. In other embodiments of the present invention, the graph resolution size and edge distance may be set to other values. And inputting the corresponding identity identification information of the cut human face picture into a 100-layer convolutional neural network ResNet, carrying out model training by using a loss function based on Arcface, iterating for 18 ten thousand times in total, reducing the learning rate by ten times when iterating for 10 ten thousand times, and reducing the learning rate by ten times again when iterating for 14 ten thousand times until learning is finished. And after the training is finished, removing the last full-connection layer, and only remaining the backbone network as a face depth recognition model.
In the embodiment of the invention, after the face depth recognition model is obtained, the training process of the age recognition model is as follows. Based on the face age data sets MORPH II and FG-NET, after the photos in the data sets are input into a backbone network of the face depth recognition model, a plurality of corresponding 512-dimensional face feature vectors can be generated; each facial feature vector may obtain a corresponding age label from the data set. The generated face feature vector is randomly divided into two parts according to a specific proportion, for example, 70% of the face feature vector is selected as a training set, a 100 multi-classification age model is trained through a two-layer network of a full-connection structure by adopting a function based on cross entropy through a loss function (the loss function is 100 age classifications are output and respectively correspond to 1 year to 100 years), after the training is finished, the rest 30% of photos are used as test data of the age model, the probability obtained by each age classification is multiplied by the corresponding age and summed to obtain the specific age corresponding to each photo, then the difference between the actual age and the actual age is calculated, and finally the average error of the age model is verified to be 3.9 years. According to the embodiment of the invention, the face depth recognition model is multiplexed to obtain the shallow age recognition model generated by the obtained face feature vector, so that a recognition result with higher precision can be obtained.
The system of the embodiment of the present invention, referring to fig. 4, includes: the detection cutting module 100 is configured to receive a face photo, detect facial features in the face photo, obtain coordinates of facial key feature points, determine a cutting boundary according to the coordinates of the facial key feature points based on a preset cutting rule, and obtain a cut picture, where the cut picture includes a facial recognition area and an edge area surrounding the facial recognition area; the feature extraction module 200 is configured to obtain a face feature vector with a fixed number of channels according to the cut picture based on a backbone network of the face depth recognition model; the identity recognition module 300 is configured to match the face feature vector and obtain identity recognition information corresponding to the face feature vector according to the similarity; and the age identification module 400 is used for obtaining the identification age based on the age identification models of a plurality of age classifications according to the face feature vector.
Referring to fig. 5, the system according to the embodiment of the present invention includes: a key point extraction module 110 and a picture cropping module 120. The key point extraction module is used for carrying out multi-scale change according to the face photos, constructing an image pyramid, constructing a plurality of candidate frames in the image pyramid, filtering the candidate frames, screening out the candidate frames with face areas, generating face recognition areas and obtaining coordinates of key feature points of the face; and the picture cutting module is used for cutting the face picture according to the coordinates of the key feature points of the face and a preset cutting rule, wherein the preset cutting rule comprises the size of the cut picture and the distance between a face recognition area and a cutting frame in the cut picture.
Referring to fig. 5, the system according to the embodiment of the present invention further includes: a face feature vector management module 500, a face recognition training module 600, and an age recognition training module 700. And the face feature vector management module 500 is configured to manage the face feature vectors generated based on the face depth recognition model. The face recognition training module 600 is used for acquiring face feature vectors through the detection cutting module and the feature extraction module based on the MS-Celeb-1M face data set, and training a face depth recognition model; and the age recognition training module 700 is used for acquiring a face feature vector through the detection cutting module and the feature extraction module based on the MORPH II face age data set and the FG-NET face age data set, and training an age recognition model. The face feature vectors generated by the face recognition training module 600 and the corresponding identification data are stored in the face identification template library through the face feature vector management module 500. The face feature vector and the corresponding age labeling information acquired by the feature extraction module 200 are stored in an age identification training database through the face feature vector management module 500, and can be used as training data of the age identification training module 700.
The system of the embodiment of the invention refers to fig. 6 for the data processing process of the face photo. Firstly, the original face picture is cut and normalized by the detection cutting module 100, and the feature extraction module 200 outputs face recognition feature vectors of fixed channels (such as 512 dimensions); the face recognition feature vectors are sent to the identity module 300 and the age module 400 for processing at the same time. The identity recognition module 300 matches the face recognition feature vector from the face identity recognition template library according to the similarity to obtain corresponding identity data. The age identification module 400 obtains an identification age through a shallow age identification model; the shallow age identification model is a multi-classification age identification model, and each classification corresponds to one age classification and has a corresponding age value. The face feature vector is used as input data, and the probability corresponding to each age classification of the age identification model can be obtained; then, according to the age classification and the probability, the identification age is obtained, and the calculation method comprises the following steps: n ═ Σ (N)i*pi) Wherein N represents the recognition age, NiIndicates the age, p, corresponding to the ith age categoryiRepresenting the probability of the ith age category.
Although specific embodiments have been described herein, those of ordinary skill in the art will recognize that many other modifications or alternative embodiments are equally within the scope of this disclosure. For example, any of the functions and/or processing capabilities described in connection with a particular device or component may be performed by any other device or component. In addition, while various illustrative implementations and architectures have been described in accordance with embodiments of the present disclosure, those of ordinary skill in the art will recognize that many other modifications of the illustrative implementations and architectures described herein are also within the scope of the present disclosure.
Certain aspects of the present disclosure are described above with reference to block diagrams and flowchart illustrations of systems, methods, apparatus and/or computer program products according to example embodiments. It will be understood that one or more blocks of the block diagrams and flowchart illustrations, and combinations of blocks in the block diagrams and flowchart illustrations, respectively, can be implemented by executing computer-executable program instructions. Also, according to some embodiments, some blocks of the block diagrams and flow diagrams may not necessarily be performed in the order shown, or may not necessarily be performed in their entirety. In addition, additional components and/or operations beyond those shown in the block diagrams and flow diagrams may be present in certain embodiments.
Accordingly, blocks of the block diagrams and flowchart illustrations support combinations of means for performing the specified functions, combinations of elements or steps for performing the specified functions and program instruction means for performing the specified functions. It will also be understood that each block of the block diagrams and flowchart illustrations, and combinations of blocks in the block diagrams and flowchart illustrations, can be implemented by special purpose hardware-based computer systems that perform the specified functions, elements or steps, or combinations of special purpose hardware and computer instructions.
Program modules, applications, etc. described herein may include one or more software components, including, for example, software objects, methods, data structures, etc. Each such software component may include computer-executable instructions that, in response to execution, cause at least a portion of the functionality described herein (e.g., one or more operations of the illustrative methods described herein) to be performed.
The software components may be encoded in any of a variety of programming languages. An illustrative programming language may be a low-level programming language, such as assembly language associated with a particular hardware architecture and/or operating system platform. Software components that include assembly language instructions may need to be converted by an assembler program into executable machine code prior to execution by a hardware architecture and/or platform. Another exemplary programming language may be a higher level programming language, which may be portable across a variety of architectures. Software components that include higher level programming languages may need to be converted to an intermediate representation by an interpreter or compiler before execution. Other examples of programming languages include, but are not limited to, a macro language, a shell or command language, a job control language, a scripting language, a database query or search language, or a report writing language. In one or more exemplary embodiments, a software component containing instructions of one of the above programming language examples may be executed directly by an operating system or other software component without first being converted to another form.
The software components may be stored as files or other data storage constructs. Software components of similar types or related functionality may be stored together, such as in a particular directory, folder, or library. Software components may be static (e.g., preset or fixed) or dynamic (e.g., created or modified at execution time).
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.
Claims (10)
1. A face recognition method based on a deep convolutional neural network is characterized by comprising the following steps:
the method comprises the steps of detecting and cutting, namely receiving a face photo, detecting facial features in the face photo, obtaining coordinates of facial key feature points, determining a cutting boundary according to the coordinates of the facial key feature points based on a preset cutting rule, and obtaining a cut picture, wherein the cut picture comprises a facial recognition area and an edge area surrounding the facial recognition area;
a feature extraction step, namely acquiring a face feature vector with a fixed channel number according to the cut picture based on a backbone network of a face depth recognition model;
an identity recognition step, namely matching the face feature vector and acquiring identity recognition information corresponding to the face feature vector according to the similarity;
and an age identification step, namely acquiring identification ages based on age identification models of a plurality of age classifications according to the face feature vectors.
2. The method for face recognition based on deep convolutional neural network of claim 1, wherein the step of detecting clipping comprises:
a key point extraction step, namely performing multi-scale change according to the face picture to construct an image pyramid, constructing a plurality of candidate frames in the image pyramid, filtering the candidate frames, screening out the candidate frames with a face area, generating a face recognition area, and obtaining coordinates of key feature points of the face;
and a picture cutting step, wherein the face picture is cut according to a preset cutting rule according to the coordinates of the key feature points of the face, wherein the preset cutting rule comprises the size of the cut picture and the distance between the face recognition area and a cutting frame in the cut picture.
3. The method of claim 2, wherein the cropped picture has a resolution of 112 x 112, and the face recognition region is located at a distance of 16 from the cropped border.
4. The face recognition method based on the deep convolutional neural network of claim 1, wherein the age recognition step comprises:
taking the face feature vector as input data, and obtaining the probability corresponding to each age classification of the age identification model according to the age identification model;
and obtaining the identification age according to the age classification and the probability.
5. The method according to claim 1, wherein the fixed number of channels of the face feature vector is 512 dimensions.
6. The face recognition method based on the deep convolutional neural network of claim 1, further comprising:
a face recognition model training step, based on an MS-Celeb-1M face data set, acquiring the face feature vector through the detecting and cutting step and the feature extraction step, and training the face depth recognition model;
and an age recognition model training step, wherein the face feature vector is obtained through the detection cutting step and the feature extraction step based on a MORPH II face age data set and an FG-NET face age data set, and the age recognition model is trained.
7. A face recognition system based on a deep convolutional neural network, using the method of any of claims 1 to 6, comprising:
the detection cutting module is used for receiving a face photo, detecting facial features in the face photo, obtaining coordinates of facial key feature points, determining a cutting boundary according to the coordinates of the facial key feature points based on a preset cutting rule, and obtaining a cut picture, wherein the cut picture comprises the facial recognition area and an edge area surrounding the facial recognition area;
the feature extraction module is used for acquiring a face feature vector with a fixed channel number according to the cut picture based on a backbone network of a face depth recognition model;
the identity recognition module is used for matching the face feature vector and acquiring identity recognition information corresponding to the face feature vector according to the similarity;
and the age identification module is used for acquiring the identification age based on the age identification models of a plurality of age classifications according to the face feature vector.
8. The deep convolutional neural network-based face recognition system of claim 7, wherein the detection clipping module comprises:
the key point extraction module is used for carrying out multi-scale change according to the face photos, constructing an image pyramid, constructing a plurality of candidate frames in the image pyramid, filtering the candidate frames, screening out the candidate frames with face areas, generating face recognition areas and obtaining coordinates of key feature points of the faces;
and the picture cutting module is used for cutting the face picture according to the coordinates of the key feature points of the face and a preset cutting rule, wherein the preset cutting rule comprises the size of the cut picture and the distance between the face recognition area and a cutting frame in the cut picture.
9. The deep convolutional neural network-based face recognition system of claim 7, further comprising:
and the face feature vector management module is used for managing the face feature vectors generated based on the face depth recognition model.
10. The deep convolutional neural network-based face recognition system of claim 7, further comprising:
the face recognition training module is used for acquiring the face feature vector through the detection cutting module and the feature extraction module based on an MS-Celeb-1M face data set and training the face depth recognition model;
and the age recognition training module is used for acquiring the face feature vector through the detection cutting module and the feature extraction module based on the MORPH II face age data set and the FG-NET face age data set, and training the age recognition model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010675730.7A CN111985323B (en) | 2020-07-14 | 2020-07-14 | Face recognition method and system based on deep convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010675730.7A CN111985323B (en) | 2020-07-14 | 2020-07-14 | Face recognition method and system based on deep convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111985323A true CN111985323A (en) | 2020-11-24 |
| CN111985323B CN111985323B (en) | 2021-10-22 |
Family
ID=73439486
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010675730.7A Active CN111985323B (en) | 2020-07-14 | 2020-07-14 | Face recognition method and system based on deep convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111985323B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364831A (en) * | 2020-11-30 | 2021-02-12 | 姜培生 | Face recognition method and online education system |
| CN112766167A (en) * | 2021-01-20 | 2021-05-07 | 宁夏计算机软件与技术服务有限公司 | Face recognition method based on face feature vector |
| CN113077265A (en) * | 2020-12-08 | 2021-07-06 | 泰州市朗嘉馨网络科技有限公司 | Live client credit management system |
| CN113239812A (en) * | 2021-05-17 | 2021-08-10 | 上海中通吉网络技术有限公司 | Face recognition method based on picture |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170161551A1 (en) * | 2015-05-29 | 2017-06-08 | Tencent Technology (Shenzhen) Company Limited | Face key point positioning method and terminal |
| CN109543606A (en) * | 2018-11-22 | 2019-03-29 | 中山大学 | A kind of face identification method that attention mechanism is added |
| CN110175504A (en) * | 2019-04-08 | 2019-08-27 | 杭州电子科技大学 | A kind of target detection and alignment schemes based on multitask concatenated convolutional network |
| CN110674748A (en) * | 2019-09-24 | 2020-01-10 | 腾讯科技（深圳）有限公司 | Image data processing method, image data processing device, computer equipment and readable storage medium |
| CN111274947A (en) * | 2020-01-19 | 2020-06-12 | 广州广电卓识智能科技有限公司 | Multi-task multi-thread face recognition method, system and storage medium |
-
2020
- 2020-07-14 CN CN202010675730.7A patent/CN111985323B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170161551A1 (en) * | 2015-05-29 | 2017-06-08 | Tencent Technology (Shenzhen) Company Limited | Face key point positioning method and terminal |
| CN109543606A (en) * | 2018-11-22 | 2019-03-29 | 中山大学 | A kind of face identification method that attention mechanism is added |
| CN110175504A (en) * | 2019-04-08 | 2019-08-27 | 杭州电子科技大学 | A kind of target detection and alignment schemes based on multitask concatenated convolutional network |
| CN110674748A (en) * | 2019-09-24 | 2020-01-10 | 腾讯科技（深圳）有限公司 | Image data processing method, image data processing device, computer equipment and readable storage medium |
| CN111274947A (en) * | 2020-01-19 | 2020-06-12 | 广州广电卓识智能科技有限公司 | Multi-task multi-thread face recognition method, system and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 谭莲芝: ""基于深度学习的人脸属性识别方法研究"", 《中国优秀硕士学位论文全文数据库（信息科技辑）》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112364831A (en) * | 2020-11-30 | 2021-02-12 | 姜培生 | Face recognition method and online education system |
| CN113077265A (en) * | 2020-12-08 | 2021-07-06 | 泰州市朗嘉馨网络科技有限公司 | Live client credit management system |
| CN113077265B (en) * | 2020-12-08 | 2021-11-30 | 鑫绪(上海)信息技术服务有限公司 | Live client credit management system |
| CN112766167A (en) * | 2021-01-20 | 2021-05-07 | 宁夏计算机软件与技术服务有限公司 | Face recognition method based on face feature vector |
| CN113239812A (en) * | 2021-05-17 | 2021-08-10 | 上海中通吉网络技术有限公司 | Face recognition method based on picture |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111985323B (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111985323B (en) | Face recognition method and system based on deep convolutional neural network | |
| US10163022B1 (en) | Method for learning text recognition, method for recognizing text using the same, and apparatus for learning text recognition, apparatus for recognizing text using the same | |
| CN109034069B (en) | Method and apparatus for generating information | |
| US20180260735A1 (en) | Training a hidden markov model | |
| CN112396049A (en) | Text error correction method and device, computer equipment and storage medium | |
| KR101996371B1 (en) | System and method for creating caption for image and computer program for the same | |
| CN114706966A (en) | Voice interaction method, device and equipment based on artificial intelligence and storage medium | |
| US20180005087A1 (en) | Pattern recognition device, pattern recognition method, and computer program product | |
| CN111292333B (en) | Method and apparatus for segmenting an image | |
| CN111783812A (en) | Method and device for identifying forbidden images and computer readable storage medium | |
| CN110717407A (en) | Human face recognition method, device and storage medium based on lip language password | |
| CN108762936B (en) | Distributed computing system and method based on artificial intelligence image recognition | |
| JP2013008093A (en) | Image recognition device, image recognition method and program | |
| KR20230062251A (en) | Apparatus and method for document classification based on texts of the document | |
| US20210044864A1 (en) | Method and apparatus for identifying video content based on biometric features of characters | |
| CN111368066A (en) | Method, device and computer readable storage medium for acquiring dialogue abstract | |
| CN113537187A (en) | Text recognition method and device, electronic equipment and readable storage medium | |
| CN117252947A (en) | Image processing method, image processing apparatus, computer, storage medium, and program product | |
| CN112183542A (en) | Text image-based recognition method, device, equipment and medium | |
| CN111783725A (en) | Face recognition method, face recognition device and storage medium | |
| CN109858355B (en) | Image processing method and related product | |
| CN112801994B (en) | Bone age assessment method and system | |
| CN114780757A (en) | Short media label extraction method and device, computer equipment and storage medium | |
| KR102476334B1 (en) | Diary generator using deep learning | |
| JP2022028912A (en) | Verification processing device, verification processing method, and recording medium storing verification processing program |
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 |