CN112733720A - Face recognition method based on firework algorithm improved convolutional neural network - Google Patents

Face recognition method based on firework algorithm improved convolutional neural network Download PDF

Info

Publication number
CN112733720A
CN112733720A CN202110034002.2A CN202110034002A CN112733720A CN 112733720 A CN112733720 A CN 112733720A CN 202110034002 A CN202110034002 A CN 202110034002A CN 112733720 A CN112733720 A CN 112733720A
Authority
CN
China
Prior art keywords
spark
neural network
convolutional neural
face
model
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
Application number
CN202110034002.2A
Other languages
Chinese (zh)
Inventor
李峰
潘文哲
董林
向文杰
周军
秦晓飞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Shanghai for Science and Technology
Original Assignee
University of Shanghai for Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Shanghai for Science and Technology filed Critical University of Shanghai for Science and Technology
Priority to CN202110034002.2A priority Critical patent/CN112733720A/en
Publication of CN112733720A publication Critical patent/CN112733720A/en
Pending legal-status Critical Current

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
    • G06V40/161—Detection; Localisation; Normalisation
    • G06V40/166—Detection; Localisation; Normalisation using acquisition arrangements
    • 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
    • G06N3/048—Activation functions
    • 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
    • G06N3/08—Learning methods
    • 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
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a face recognition method based on a firework algorithm improved convolutional neural network, which comprises the following steps of: establishing a face database, setting face attribute labels, and constructing a face label data set, wherein the face label data set comprises a training set, a verification set and a test set; constructing an initial convolutional neural network model based on a firework algorithm; training the initial convolutional neural network model based on the training set and the verification set to obtain a target convolutional neural network model, and testing the target convolutional neural network through the test set to obtain a convolutional neural network model; the method is simple and efficient, can realize face recognition only by acquiring a face picture, and is accurate in recognition.

Description

Face recognition method based on firework algorithm improved convolutional neural network
Technical Field
The invention belongs to the technical field of computer vision technology and face recognition, and particularly relates to a face recognition method based on a firework algorithm improved convolutional neural network.
Background
The technology is developed rapidly at present, and we have entered into the information era comprehensively, and the face recognition technology is a product brought by the development of technology, and is used in many scenes, such as train arrival, gate for entrance and exit of dormitory buildings, bank security inspection, network people finding, and the like. Although the face recognition technology starts late in China, the face recognition technology develops rapidly under the strong support of China. Many colleges and universities develop novel talents in scientific research and culture in the aspects of computer vision, mode recognition, artificial intelligence and deep learning, and obtain good results. In the information age, many people worry about the problem of information security, and in fact, many of the authentication methods commonly used in the early days have not been completely secure as the age evolves, because such information is easily counterfeited. At present, a general face recognition system capable of being practically applied to any background does not exist. How to further improve the face recognition effect in the use of face recognition system is the technical problem that this patent will solve.
Disclosure of Invention
In order to solve the problems, the method is based on a firework algorithm optimization convolutional neural network model, firstly basic features of an input face are extracted, then global information and identity information contained in a face recognition sub-model are merged into an attribute classification sub-model to help improve the effect of attribute classification, and a countermeasure network is introduced to further improve the accuracy of face recognition.
The invention provides a face recognition method based on a firework algorithm improved convolutional neural network, which comprises the following steps as shown in figures 1-3:
s1, establishing a face database, setting face attribute labels based on the face database, and constructing a face label data set, wherein the face label data set comprises a training set, a verification set and a test set;
s2, constructing an initial convolutional neural network model based on a firework algorithm, wherein the convolutional neural network model comprises a loss function, an activation function and a classifier;
s3, training the initial convolutional neural network model based on the training set and the verification set to obtain a target convolutional neural network model, and testing the target convolutional neural network through the test set to obtain a convolutional neural network model;
and S4, acquiring the face attribute characteristics and the face global characteristics of the face label data set through the convolutional neural network model based on the face label data set, and identifying a target face image based on the face attribute characteristics and the face global characteristics, wherein the target face image is a face image recorded into the face database.
Preferably, the ratio of the training set, the verification set and the test set is 2:2: 1.
Preferably, the number of iterations of the convolutional neural network model is 1000.
Preferably, the S2 includes the steps of:
s2.1, setting an initial population, a minimum iteration number, a maximum iteration number, a firework explosion radius, the number of sparks, a spark first boundary and a spark second boundary;
s2.2, based on the firework algorithm, obtaining spark decoding through the initial population, obtaining spark dimension based on the number of the spark decoding, and constructing a spark forward propagation model based on the spark decoding and the spark dimension;
s2.3, constructing an error offset model based on the spark forward propagation model, judging whether to update the spark position or not based on the minimum iteration times and the maximum iteration times through the error offset model, if not, stopping the algorithm, and if so, executing S2.4;
s2.4, establishing a spark position model and a spark quantity model according to the firework explosion radius, the spark first boundary and the spark second boundary by setting a fixed constant based on the spark position vector and the spark position fitness, wherein the spark quantity is calculated through the spark quantity model;
s2.5, constructing the initial convolutional neural network model based on the spark position model and the spark quantity model, wherein the spark position model corresponds to a weight of the initial convolutional neural network model, and the initial convolutional neural network model realizes network layer-by-layer mapping through the weight based on the activation function to obtain an output result.
Preferably, said S2.4 further comprises the steps of:
s2.4.1, constructing a spark mapping model exceeding the explosion radius of the fireworks according to the first spark boundary and the second spark boundary and through the explosion radius of the fireworks based on a Gaussian variation algorithm;
s2.4.2, constructing a candidate individual probability model based on the spark mapping model and the spark position fitness;
and S2.4.3, constructing the spark position model and the spark quantity model through the spark position vector and a fixed constant based on the candidate individual probability model.
Preferably, the S2.4.2 includes determining next individual spark population based on the minimum value of the spark position fitness, and constructing candidate individual probability models by roulette based on the remaining fitness of the spark position fitness.
Preferably, the step S2.5 further includes returning to the step S2.3 after completing the network layer-to-layer mapping once.
Preferably, the convolutional neural network model further comprises a countermeasure network model; the convolutional neural network model is used for improving the learning ability and the accuracy of the convolutional neural network model;
the construction method of the countermeasure network model comprises the following steps:
s3.1, based on the face picture, inputting the vector of the face picture after being compressed and dimensionality reduced through a principal component analysis method into a generator of the countermeasure network model, and generating a false face picture of the face picture
S3.2, the fake face picture and the face picture pass through a discriminator of the confrontation network model and obtain a feedback result of the discriminator based on the classifier;
and S3.3, adjusting the network parameters of the generator and the arbiter based on the feedback result of the arbiter, so as to achieve the purpose of minimizing the loss function of the generator and maximizing the loss of the arbiter until the loss function tends to be balanced.
The positive progress effects of the invention are as follows:
the invention obtains the model of face recognition through the above operations, at present, only the face picture collected through the camera is needed to be input into the trained model, the characteristic vector obtained in the model through the picture is compared with the model vector in the library, if within a certain degree of error, the recognition is successful.
Drawings
FIG. 1 is a flow chart of the firework algorithm-based optimized convolutional neural network for a face recognition system according to the present invention;
FIG. 2 is a flow chart of the countermeasure network of the present invention;
FIG. 3 is a schematic diagram of a convolutional neural network training method according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application. In the embodiment, the basic features of the input face are extracted based on the firework algorithm optimization convolutional neural network model, then the global information and the identity information contained in the face recognition sub-model are merged into the attribute classification sub-model to help improve the attribute classification effect, and a countermeasure network is introduced to further improve the face recognition accuracy. The method comprises the following specific steps:
1. 10000 faces are input, a face database is established, each picture has a corresponding identity label, 1 represents that the face picture includes the face attribute, and 0 represents that the face picture does not include, for example, in a male label, 1 represents that the sample is male, and 0 represents that the sample is female. The face identity label is a number between 1 and 10000, the two face identity labels are the same number, namely, the two face identity labels represent that the two face identity labels belong to the same identity, and if the numbers of the face identity labels are different, the two face identity labels do not belong to the same identity.
2. 4000 out of 10000 sheets are used as training sets, 4000 sheets are used as verification sets, and the rest 2000 sheets are used as test sets. The data in the training set and the validation set will be used to train the model, and the data in the test set will be used to evaluate the performance of the trained model.
3. Extracting attribute features and human face global features from the preprocessed human face image sample through a designed convolutional neural network model, constructing a convolutional neural network, selecting the conditional _ crosstalk as a loss function, the sigmoid as an activation function and the Softmax as a classifier, and iterating for 1000 times.
Step 1: setting an initial population N, wherein the range of the parameter of the convolutional neural network corresponds to the initial population N; setting the current iteration number iter, wherein the maximum iteration number is iter _ max; controlling the radius of explosion and generating the number of smoke and fire respectively
Figure BDA0002893446850000061
And M; and setting parameters a and b for controlling the firework boundary.
Step 2. the training process of the convolutional neural network can be divided into a forward propagation part and a backward propagation part, wherein the forward propagation part is mainly used for transmitting parameters layer by layer and extracting characteristicsObtaining a final output result, comparing the model output with a target value to obtain an error offset, wherein the forward propagation process can be expressed by the following formula, wherein XpAs an initial quantity, W represents the weight of each node, and F represents the activation function of the node:
y′＝Fn(…(F1(XpW(1))W(1))…W(n))
and decoding the sparks generated by the firework algorithm into parameters of each layer of the convolutional neural network, wherein the number of the parameters corresponds to the dimensionality of the sparks and is substituted into the formula to obtain the final output.
And 3, substituting the output result into the LossFunction to calculate an error (fitness value), wherein a represents the output result of the model and y is a classification label. When y is 1, it indicates a correct class of tag, and when y is 0, it indicates an erroneous tag.
Figure BDA0002893446850000071
Step 4, optimizing the firework population, and based on the step 3, for each firework individual XiThe explosion, displacement and variation operations are carried out, the formula for generating the next generation of sparks by the explosion of the fireworks in the fireworks algorithm is as follows, and the formula (1) represents that the sparks are away from the fireworks AiIs generated within a range of (1), the formula (2) represents that the number of sparks is Si：
Figure BDA0002893446850000072
Figure BDA0002893446850000073
In the formula Xi(i ═ 1, 2, 3..) denotes the position vector of each firework, f (X)i) Indicating the fitness value, y, of the fireworks at that locationmin＝min(f(Xi))，ymax＝max(f(Xi) ε is a constant.
And 5, after the fireworks explode to generate sparks, in order to ensure the diversity of the fireworks, a variation mechanism is also added into the fireworks algorithm, and a calculation method of Gaussian variation is represented by a formula (3), wherein Gaussian (-1,1) -N (0, 1):
Figure BDA0002893446850000074
the explosion space of fireworks is not infinite, and for sparks exceeding the search boundary, the fireworks algorithm sets a mapping rule as shown in formula (4), wherein
Figure BDA0002893446850000075
And
Figure BDA0002893446850000076
representing the scope of the algorithmic search on the dimension K of
Figure BDA0002893446850000077
Figure BDA0002893446850000081
Step 6, selecting the next-generation firework population, and regarding the firework individuals X subjected to the explosion and variation operations in the steps 4 and 5iCalculating each firework individual X by using the formula in the step 3iSelecting the optimal firework individual to form the next generation firework population by using the selection strategies of the formula (5) and the formula (6), wherein the specific selection strategy is as follows: selecting min (f (X) with minimum fitness valuei) ) individual XkDirectly using a wheel roulette mode for one firework population individual and the rest N-1 firework individuals to select X candidate individualsiThe probability of its selection is as follows:
Figure BDA0002893446850000082
wherein R is: (Xi) Showing individual fireworks XiWith other individuals XjThe sum of the distances of (a) is specifically represented by the following formula (6):
Figure BDA0002893446850000083
step 7, judging whether the error reaches the minimum value of the threshold range or the maximum iteration number, and stopping the algorithm if the error reaches the minimum value of the threshold range or the maximum iteration number; if not, executing the following formula to further improve the position of the firework to obtain the parameter value of the next round.
And 8, correspondingly forming the updated firework positions into weights of the network one by one to obtain a new round of training parameters, and continuing to complete layer-by-layer mapping of the network until a predicted result is output and returning to the step 3.
4. This patent introduces a confrontation network when setting up convolution neural network to this improves the learning ability and the accuracy of this model, and concrete step is as follows:
step 1: and (3) inputting the vector of the picture after being compressed and dimensionality reduced by a principal component analysis method into a generator in the countermeasure network to generate a false picture similar to a true picture.
Step 2: taking the false pictures and the true pictures as input data of a discriminator, and extracting data characteristics by the discriminator through a multilayer convolution network to classify the pictures;
and step 3: according to the feedback result of the discriminator, the generator and the discriminator adjust the network parameters thereof, so as to achieve the purpose of minimizing the loss function of the generator and simultaneously maximizing the loss of the discriminator until the model loss tends to be balanced.
5. And inputting the training set and the verification set into the network for training. And inputting the pictures of the test set into the trained network, and checking the training effect.
6. The model for face recognition is obtained through the above operations, at present, only a face picture acquired through a camera needs to be input into a trained model, a feature vector acquired in the model through the picture is compared with the model in the library, and if the feature vector is within a certain degree of error, the recognition is successful.
The above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the present invention in its spirit and scope. Are intended to be covered by the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (8)

1. A face recognition method based on a firework algorithm improved convolutional neural network is characterized by comprising the following steps:
s1, establishing a face database, setting face attribute labels based on the face database, and constructing a face label data set, wherein the face label data set comprises a training set, a verification set and a test set;
s2, constructing an initial convolutional neural network model based on a firework algorithm, wherein the convolutional neural network model comprises a loss function, an activation function and a classifier;
s3, training the initial convolutional neural network model based on the training set and the verification set to obtain a target convolutional neural network model, and testing the target convolutional neural network through the test set to obtain a convolutional neural network model;
and S4, acquiring the face attribute characteristics and the face global characteristics of the face label data set through the convolutional neural network model based on the face label data set, and identifying a target face image based on the face attribute characteristics and the face global characteristics, wherein the target face image is a face image recorded into the face database.
2. The face recognition method based on the firework algorithm improved convolutional neural network as claimed in claim 1,
the proportion of the training set, the verification set and the test set is 2:2: 1.
3. The face recognition method based on the firework algorithm improved convolutional neural network as claimed in claim 1,
the number of iterations of the convolutional neural network model is 1000.
4. The face recognition method based on the firework algorithm improved convolutional neural network as claimed in claim 1,
the S2 includes the steps of:
s2.1, setting an initial population, a minimum iteration number, a maximum iteration number, a firework explosion radius, the number of sparks, a spark first boundary and a spark second boundary;
s2.2, based on the firework algorithm, obtaining spark decoding through the initial population, obtaining spark dimension based on the number of the spark decoding, and constructing a spark forward propagation model based on the spark decoding and the spark dimension;
s2.3, constructing an error offset model based on the spark forward propagation model, judging whether to update the spark position or not based on the minimum iteration times and the maximum iteration times through the error offset model, if not, stopping the algorithm, and if so, executing S2.4;
s2.4, establishing a spark position model and a spark quantity model according to the firework explosion radius, the spark first boundary and the spark second boundary by setting a fixed constant based on the spark position vector and the spark position fitness, wherein the spark quantity is calculated through the spark quantity model;
s2.5, constructing the initial convolutional neural network model based on the spark position model and the spark quantity model, wherein the spark position model corresponds to a weight of the initial convolutional neural network model, and the initial convolutional neural network model realizes network layer-by-layer mapping through the weight based on the activation function to obtain an output result.
5. The face recognition method based on the firework algorithm improved convolutional neural network as claimed in claim 4,
the S2.4 further comprises the steps of:
s2.4.1, constructing a spark mapping model exceeding the explosion radius of the fireworks according to the first spark boundary and the second spark boundary and through the explosion radius of the fireworks based on a Gaussian variation algorithm;
s2.4.2, constructing a candidate individual probability model based on the spark mapping model and the spark position fitness;
and S2.4.3, constructing the spark position model and the spark quantity model through the spark position vector and a fixed constant based on the candidate individual probability model.
6. The method for face recognition based on the firework algorithm improved convolutional neural network as claimed in claim 5,
the S2.4.2 includes determining next spark population individuals based on a minimum of the spark position fitness, and constructing candidate individual probability models by roulette based on remaining fitness of the spark position fitness.
7. The face recognition method based on the firework algorithm improved convolutional neural network as claimed in claim 4,
and S2.5, returning to the step S2.3 after completing the layer-by-layer mapping of the network.
8. The face recognition method based on the firework algorithm improved convolutional neural network as claimed in claim 1,
the convolutional neural network model also comprises a confrontation network model; the convolutional neural network model is used for improving the learning ability and the accuracy of the convolutional neural network model;
the construction method of the countermeasure network model comprises the following steps:
s3.1, based on the face picture, inputting the vector of the face picture after being compressed and dimensionality reduced through a principal component analysis method into a generator of the countermeasure network model, and generating a false face picture of the face picture
S3.2, the fake face picture and the face picture pass through a discriminator of the confrontation network model and obtain a feedback result of the discriminator based on the classifier;
and S3.3, adjusting the network parameters of the generator and the arbiter based on the feedback result of the arbiter, so as to achieve the purpose of minimizing the loss function of the generator and maximizing the loss of the arbiter until the loss function tends to be balanced.
CN202110034002.2A 2021-01-12 2021-01-12 Face recognition method based on firework algorithm improved convolutional neural network Pending CN112733720A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110034002.2A CN112733720A (en) 2021-01-12 2021-01-12 Face recognition method based on firework algorithm improved convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110034002.2A CN112733720A (en) 2021-01-12 2021-01-12 Face recognition method based on firework algorithm improved convolutional neural network

Publications (1)

Publication Number Publication Date
CN112733720A true CN112733720A (en) 2021-04-30

Family

ID=75590542

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110034002.2A Pending CN112733720A (en) 2021-01-12 2021-01-12 Face recognition method based on firework algorithm improved convolutional neural network

Country Status (1)

Country Link
CN (1) CN112733720A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170147922A1 (en) * 2015-11-23 2017-05-25 Daniel Chonghwan LEE Filtering, smoothing, memetic algorithms, and feasible direction methods for estimating system state and unknown parameters of electromechanical motion devices
CN107169565A (en) * 2017-04-27 2017-09-15 西安工程大学 Yarn quality prediction method based on fireworks algorithm improvement BP neural network
CN111814706A (en) * 2020-07-14 2020-10-23 电子科技大学 Face recognition and attribute classification method based on multitask convolutional neural network
CN112149883A (en) * 2020-09-07 2020-12-29 南京邮电大学 Photovoltaic power prediction method based on FWA-BP neural network

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170147922A1 (en) * 2015-11-23 2017-05-25 Daniel Chonghwan LEE Filtering, smoothing, memetic algorithms, and feasible direction methods for estimating system state and unknown parameters of electromechanical motion devices
CN107169565A (en) * 2017-04-27 2017-09-15 西安工程大学 Yarn quality prediction method based on fireworks algorithm improvement BP neural network
CN111814706A (en) * 2020-07-14 2020-10-23 电子科技大学 Face recognition and attribute classification method based on multitask convolutional neural network
CN112149883A (en) * 2020-09-07 2020-12-29 南京邮电大学 Photovoltaic power prediction method based on FWA-BP neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
吴盼: "面向图像分类的深层网络优化算法研究", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》 *

Similar Documents

Publication Publication Date Title
CN111695415B (en) Image recognition method and related equipment
US11816183B2 (en) Methods and systems for mining minority-class data samples for training a neural network
CN108647583B (en) Face recognition algorithm training method based on multi-target learning
CN105975573A (en) KNN-based text classification method
CN110619059B (en) Building marking method based on transfer learning
CN110674996B (en) Urban traffic noise prediction method
US11763551B2 (en) Systems and methods for fine tuning image classification neural networks
CN113297936B (en) Volleyball group behavior identification method based on local graph convolution network
WO2020171904A1 (en) Human body part segmentation with real and synthetic images
CN114842343A (en) ViT-based aerial image identification method
CN111144462B (en) Unknown individual identification method and device for radar signals
CN109886206B (en) Three-dimensional object identification method and equipment
CN117636183A (en) Small sample remote sensing image classification method based on self-supervision pre-training
CN113162919A (en) Intrusion detection method based on network abnormal flow identification
CN117521012A (en) False information detection method based on multi-mode context hierarchical step alignment
CN116452904B (en) Image aesthetic quality determination method
CN112733720A (en) Face recognition method based on firework algorithm improved convolutional neural network
CN115936773A (en) Internet financial black product identification method and system
CN115543762A (en) Method and system for expanding SMART data of disk and electronic equipment
CN115034426A (en) Rolling load prediction method based on phase space reconstruction and multi-model fusion Stacking integrated learning mode
KR102418232B1 (en) Method for evaluating a voice similarity of singer and device thereof
CN115249513A (en) Neural network copy number variation detection method and system based on Adaboost integration idea
CN114330650A (en) Small sample characteristic analysis method and device based on evolutionary element learning model training
CN117079336B (en) Training method, device, equipment and storage medium for sample classification model
CN115346231A (en) Cross-modal pedestrian re-identification 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