CN110647864A - Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network - Google Patents
Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network Download PDFInfo
- Publication number
- CN110647864A CN110647864A CN201910945327.9A CN201910945327A CN110647864A CN 110647864 A CN110647864 A CN 110647864A CN 201910945327 A CN201910945327 A CN 201910945327A CN 110647864 A CN110647864 A CN 110647864A
- Authority
- CN
- China
- Prior art keywords
- feature vector
- feature
- countermeasure network
- person
- recognition method
- 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
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
A single multi-graph feature recognition method, device and medium based on a generated countermeasure network are provided, and the method comprises the following steps: obtaining a single human face data set, and extracting to obtain a characteristic vector set; constructing and generating a countermeasure network, and compressing the feature vector set into a first feature vector through a generator; extracting a second feature vector of the face picture to be recognized; and putting the two feature vectors into a discriminator to judge the probability with the fusion relation. The method reserves the commonality of the input features during prediction, and removes the difference which does not influence the recognition among the input features; all information in the input features is contained as much as possible; the performance of comparison is improved when a plurality of pictures are taken.
Description
Technical Field
The invention relates to the technical field of computer image processing, in particular to a single multi-graph feature recognition method, equipment and medium based on a generated countermeasure network.
Background
The face recognition technology has a high development prospect and economic benefit in the fields of public security investigation, access control systems, target tracking and other civil safety control systems.
Machine learning belongs to one of the branches of artificial intelligence and is at the core position. Machine learning enables a computer to learn, can simulate the learning behavior of human beings, establishes learning ability and realizes recognition and judgment. Machine learning uses algorithms to analyze mass data, finds out rules from the mass data, completes learning, and makes decisions and predictions on real events by using learned thinking models. Machine learning is an artificial intelligence method, deep learning is a technology for realizing machine learning, and a generative confrontation network is a classification in deep learning.
The Convolutional Neural Networks (CNN) is an identification algorithm based on an artificial Neural network and provided in combination with a deep learning theory. The convolutional neural network can extract high-level features and improve the expression capability of the features.
Generation of countermeasure Networks (GAN) was first proposed in the paper "Generation adaptive Networks" by IangGoodfellow et al in 2014 and introduced the field of deep learning. The GAN is a probability generation model, comprises a network model of a generator G and a network model of a discriminator D, is innovatively trained on two neural networks by using a confrontation training mechanism, and realizes parameter optimization by using a Stochastic Gradient Descent (SGD) method. GAN has prominent performances in the field of computer vision, such as image translation, image super-resolution, image restoration, and the like, and is also an important research technology for face recognition.
The prior art has low information utilization degree on a facial picture set, basically only uses single information and some simple statistical information, does not utilize more information, and has the problems of inaccurate and insufficient performance.
Patent application No. CN201910329770.3 relates to a face image recognition method and device, an electronic device and a storage medium, the method includes: obtaining a plurality of face images; obtaining a plurality of target objects to be identified according to a plurality of feature vectors obtained by correspondingly extracting features of the plurality of face images; obtaining a gradient parameter according to the plurality of feature vectors and the classification reference vector; and classifying the target objects to be recognized according to the gradient parameters to obtain a classification result, so that the recognition effect on the human face can be improved.
Disclosure of Invention
The invention aims to solve the existing problems and provides a single person multi-graph feature identification method, equipment and medium based on a generation countermeasure network.
In order to achieve the above object, the present invention adopts a technical solution comprising:
s1, obtaining a single human face data set, and extracting to obtain a characteristic vector set;
s2, constructing and generating a countermeasure network, and compressing the feature vector set into a first feature vector through a generator;
s3, extracting a second feature vector of the face picture to be recognized;
and S4, putting the two feature vectors into a discriminator, and judging the probability of fusion relation.
Further, in the step S4, it is determined that there is a fusion relationship, if the second feature vector belongs to the feature vector set, it is determined that there is a fusion relationship, that is, the two feature vectors are the same person; otherwise, it is not.
Further, in the discriminator in S4, the probability of having the fusion relationship is determined by calculating the cosine similarity between the first feature vector and the second feature vector.
Further, in the generation countermeasure network, the loss function is composed of two parts:
1. further, in S4, the cosine similarity is calculated by substituting the picture connection characteristics of the single person and the person to be recognized into A, B, and then the probability of being the same person is evaluated, the formula is as follows;
a threshold value threshold is cut out in the calculation, and when similarity > = threshold, the same person is considered, and when similarity < threshold, different persons are considered.
A face recognition system characterized by: the system comprises a storage unit, a processing unit and a processing unit, wherein the storage unit is used for storing a human face data set of a single person; generating a countermeasure network, comprising: the generator network inputs a feature vector set and outputs a first feature vector, namely, the generator compresses the information of the input features into one feature; and inputting the first characteristic vector and the second characteristic vector by the discriminator network, and judging the probability that the two characteristic vectors have a fused relation, namely calculating the cosine distance.
An electronic device, characterized in that: comprising a processor configured to perform any of the methods described above, and a memory for storing executable instructions for the processor.
A computer readable medium having computer program instructions stored thereon, characterized in that: the computer program instructions, when executed, perform any of the methods described above.
Compared with the prior art, the method of the invention comprises the characteristic distribution information of one person and multiple figures on the training set, the commonality of the input characteristics in the prediction process is kept, and the difference which does not influence the recognition among the input characteristics is removed; all information in the input features is contained as much as possible; the recognition performance is ensured, so that the comparison performance of a plurality of pictures can be improved.
The training process of the generator can enable the generator to learn the distribution of the features under the condition of one person with multiple graphs, namely the difference and the commonality between the features, and the training process is applied to a predicted scene, so that the commonality between the features can be kept, the more accurate features of the person can be obtained, and irrelevant information can be removed and identified.
The existence of the discriminator enables the generator to generate the features, and the features contain the information of all the input features as much as possible; and the existence of the identification loss function can enable the generator to generate the identification characteristics of a single person, and the identification performance is improved to a certain extent.
Drawings
Fig. 1 is a schematic diagram of the generation of a countermeasure network in accordance with the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more clearly apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The deep learning machine using NVIDIA DGX-1 in the embodiment comprises 8 Nvidia Tesla-V100 computing cards, each computing card has more than 210 hundred million transistors, and the core area is 815 square millimeters (other equivalent computing resources for providing computing power can also be adopted); and use a tensoflow training framework (or other deep learning training framework).
The problem that this embodiment was solved is, on the basis of having the face identification model, promote the performance that face set discerned and compares. First, an existing face recognition model is obtained and recorded as REC. The REC can be a multitask convolutional neural network (MTCNN), the MTCNN combines face region detection and face key point detection, and the overall structure can be divided into three-layer network structures of P-Net, R-Net and O-Net for face detection and key point positioning; or FaceNet, the method changes the characteristic into a point on the Euclidean plane directly by learning, then judges by comparing the distance between the points, and extracts the characteristic of face recognition. Any other existing visible light recognition model and method can be used for achieving the task.
The input of the REC is a face image, and the output is a feature vector of the face. For the two graphs p1 and p2, the probability that the two graphs are the same person is obtained by calculating the cosine distance between the feature vector REC (p1) and the feature vector REC (p 2).
In the scenario of this embodiment, two sets of pictures Set _ a and Set _ B are stored, where:
set _ a contains several faces of the same person Pa;
set _ B contains several faces of the same person Pb.
Whether Pa and Pb are the same person or not needs to be judged, the following steps are provided:
1. data preparation
Preparing a plurality of groups of annotation data to form an annotation set Anno; each group of data in the Anno is denoted as set _ ann, and comprises a plurality of face sets of the same person.
Preparing a plurality of groups of data to form a negative case set Q; each group of data in Q is recorded as Q, and is a face image of a person, so that the person corresponding to Q is not shown in Anno
The formal training set is divided into two training sets, X _0 and X _ 1. Each set of data for X _0 consists of one set _ ann and one graph for Q in D, and each set of data for X _1 consists of one set _ ann and one graph for set _ ann in D.
2. Algorithm design
To address this problem, a generative countermeasure network is designed. Referring to fig. 1, fig. 1 illustrates the generation of a competing network of the present invention,
processing a picture set from a person by using an REC to obtain a characteristic vector set FS _ A;
the generator G inputs a feature vector set and outputs a feature vector; the generator compresses the information of the input features into one feature
The discriminator D inputs the two feature vectors and judges the probability that the two feature vectors have a fused relationship. One of the feature vectors f is from the generator fusion, and the other feature vector fx is from any picture obtained by using REC feature extraction.
The judgment rule is as follows: if fx comes from FS _ A, D should be biased to consider that the two feature vectors have a fused relationship; if fx does not come from FS _ A, D should be biased to assume that the two feature vectors do not have a fused relationship.
Preferably, the discriminator D and the generator G are both multilayer perceptrons and adopt a multilayer transform structure; there are 2 layers, each layer has 1024 neurons, and a global pooling layer is connected.
The loss function of the network consists of two parts, one of GAN and one of identification correlation, where:
in the training, 4 neural networks are respectively trained through a stochastic gradient descent and a back propagation algorithm. When the training is converged, stopping the training and providing a mode for judging the neural network training convergence, wherein other modes for judging the training convergence can also be adopted, when the loss function value does not decrease any more in one day, the learning rate is reduced to one tenth of the previous mode to continue the training, and when the loss function cannot be continuously reduced for 5 times continuously, the training is considered to be converged.
Preferably, the present embodiment is based on calculation
To evaluate the probability that Pa and Pb are the same person. The formula is as follows;
cutting a threshold value threshold in the calculation, when similarity > = threshold, the people are considered to be the same person, and when similarity < threshold, the people are considered to be different persons; different threshold values are cut from the test set, different false alarm rates and different recall rates are obtained, and a threshold value needs to be selected according to the scene.
In the embodiment, the generated features are superior to direct selection, and feature representation can be improved, so that the performance of face recognition is improved. In the process of generating the confrontation network training, the training process of the generator can enable the generator to learn the distribution of the features under the condition of one person with multiple graphs, namely the difference and the commonality between the features, and the generator is applied to a predicted scene, so that the commonality between the features can be kept, the more accurate features of the person can be obtained, and irrelevant information can be removed and identified.
The existence of the discriminator enables the generator to generate the features, and the features contain the information of all the input features as much as possible; and the existence of the identification loss function can enable the generator to generate the identification characteristics of a single person, and the identification performance is improved to a certain extent.
In summary, the features thus generated have the following advantages: the method indirectly comprises the feature distribution information of one person and multiple graphs on the training set, retains the commonality of input features during prediction, and removes the difference without influencing recognition among the input features; all information in the input features is contained as much as possible; the recognition performance is ensured, so that the comparison performance of a plurality of pictures can be improved.
Based on the same technical concept, the embodiment further provides a face recognition system, which comprises a storage unit, a face recognition unit and a face recognition unit, wherein the storage unit is used for storing a face data set of a single person; generating a countermeasure network, wherein the generator network inputs a feature vector set and outputs a first feature vector, namely, the generator compresses the information of the input features into one feature; and inputting the first characteristic vector and the second characteristic vector by the discriminator network, and judging the probability that the two characteristic vectors have a fused relation, namely calculating the cosine distance.
Based on the same technical concept, the embodiment further provides an electronic device, which includes at least one processor and at least one memory for storing executable instructions of the processor, and in the embodiment of the present application, a specific connection medium between the processor and the memory is not limited, and the processor and the memory are connected through a bus; and the bus may be divided into an address bus, a data bus, a control bus, etc.
The processor is the control center of the electronic device,
the present embodiments may be implemented or performed with a general-purpose processor, such as a Central Processing Unit (CPU), a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc., that may interface various portions of an electronic device using various interfaces and lines, and perform any of the methods described above by executing or executing instructions stored in memory and by invoking data stored in memory.
The present embodiment also provides a computer readable medium storing a computer program executable by an electronic device, the computer program instructions when executed by a terminal device being processed to implement any of the methods described above. The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.
The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The embodiments of the present invention have been described in conjunction with the accompanying drawings and examples, which are given by way of illustration and not of limitation, and it will be apparent to those skilled in the art that various changes and modifications may be made as required within the scope of the appended claims.
Claims (7)
1. A single multi-graph feature recognition method based on a generated countermeasure network is characterized by comprising the following steps:
s1, obtaining a single human face data set, and extracting to obtain a characteristic vector set;
s2, constructing and generating a countermeasure network, and compressing the feature vector set into a first feature vector through a generator;
s3, extracting a second feature vector of the face picture to be recognized;
and S4, putting the two feature vectors into a discriminator, and judging the probability of fusion relation.
2. The single-person multi-graph feature recognition method based on generation of a countermeasure network as claimed in claim 1, wherein: s4, judging that the second feature vector has a fusion relation, if the second feature vector belongs to the feature vector set, the second feature vector has a fusion relation, namely the second feature vector and the second feature vector are the same person; otherwise, it is not.
3. The single-person multi-graph feature recognition method based on generation of a countermeasure network as claimed in claim 1 or 2, wherein: in S4, the discriminator determines the probability of having a fusion relationship by calculating the cosine similarity between the first feature vector and the second feature vector.
4. The single-person multi-graph feature recognition method based on generation of a countermeasure network as claimed in claim 1, wherein: in the generative countermeasure network, the loss function consists of two parts:
；
the single-person multi-graph feature recognition method based on generation of a countermeasure network as claimed in claim 1 or 2, wherein: in S4, calculating cosine similarity by substituting the picture connection characteristics of the single person and the person to be identified into A, B, and further evaluating the probability of the same person, wherein the formula is as follows;
a threshold value threshold is cut out in the calculation, and when similarity > = threshold, the same person is considered, and when similarity < threshold, different persons are considered.
5. A face recognition system characterized by: the system comprises a storage unit, a processing unit and a processing unit, wherein the storage unit is used for storing a human face data set of a single person;
generating a countermeasure network, comprising:
the generator network inputs a feature vector set and outputs a first feature vector, namely, the generator compresses the information of the input features into one feature;
and inputting the first characteristic vector and the second characteristic vector by the discriminator network, and judging the probability that the two characteristic vectors have a fused relation, namely calculating the cosine distance.
6. An electronic device, characterized in that: comprising a processor, and a memory for storing executable instructions for the processor, the processor configured to: performing the method of any one of claims 1-5.
7. A computer readable medium having computer program instructions stored thereon, characterized in that: the computer program instructions, when processed and executed, implement the method of any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910945327.9A CN110647864A (en) | 2019-09-30 | 2019-09-30 | Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910945327.9A CN110647864A (en) | 2019-09-30 | 2019-09-30 | Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110647864A true CN110647864A (en) | 2020-01-03 |
Family
ID=68993462
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910945327.9A Pending CN110647864A (en) | 2019-09-30 | 2019-09-30 | Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110647864A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112288032A (en) * | 2020-11-18 | 2021-01-29 | 上海依图网络科技有限公司 | Method and device for quantitative model training based on generation of confrontation network |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104517104A (en) * | 2015-01-09 | 2015-04-15 | 苏州科达科技股份有限公司 | Face recognition method and face recognition system based on monitoring scene |
| CN105991916A (en) * | 2015-02-05 | 2016-10-05 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN106599837A (en) * | 2016-12-13 | 2017-04-26 | 北京智慧眼科技股份有限公司 | Face identification method and device based on multi-image input |
| CN107958444A (en) * | 2017-12-28 | 2018-04-24 | 江西高创保安服务技术有限公司 | A kind of face super-resolution reconstruction method based on deep learning |
| CN107977932A (en) * | 2017-12-28 | 2018-05-01 | 北京工业大学 | It is a kind of based on can differentiate attribute constraint generation confrontation network face image super-resolution reconstruction method |
| CN108229330A (en) * | 2017-12-07 | 2018-06-29 | 深圳市商汤科技有限公司 | Face fusion recognition methods and device, electronic equipment and storage medium |
| CN108268845A (en) * | 2018-01-17 | 2018-07-10 | 深圳市唯特视科技有限公司 | A kind of dynamic translation system using generation confrontation network synthesis face video sequence |
| CN108319932A (en) * | 2018-03-12 | 2018-07-24 | 中山大学 | A kind of method and device for the more image faces alignment fighting network based on production |
| CN108446609A (en) * | 2018-03-02 | 2018-08-24 | 南京邮电大学 | A kind of multi-angle human facial expression recognition method based on generation confrontation network |
| CN109308725A (en) * | 2018-08-29 | 2019-02-05 | 华南理工大学 | A kind of system that expression interest figure in mobile terminal generates |
| CN109544656A (en) * | 2018-11-23 | 2019-03-29 | 南京信息工程大学 | A kind of compressed sensing image rebuilding method and system based on generation confrontation network |
| CN110222668A (en) * | 2019-06-17 | 2019-09-10 | 苏州大学 | Based on the multi-pose human facial expression recognition method for generating confrontation network |
| CN110288537A (en) * | 2019-05-20 | 2019-09-27 | 湖南大学 | Facial image complementing method based on the depth production confrontation network from attention |
-
2019
- 2019-09-30 CN CN201910945327.9A patent/CN110647864A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104517104A (en) * | 2015-01-09 | 2015-04-15 | 苏州科达科技股份有限公司 | Face recognition method and face recognition system based on monitoring scene |
| CN105991916A (en) * | 2015-02-05 | 2016-10-05 | 联想(北京)有限公司 | Information processing method and electronic equipment |
| CN106599837A (en) * | 2016-12-13 | 2017-04-26 | 北京智慧眼科技股份有限公司 | Face identification method and device based on multi-image input |
| CN108229330A (en) * | 2017-12-07 | 2018-06-29 | 深圳市商汤科技有限公司 | Face fusion recognition methods and device, electronic equipment and storage medium |
| CN107958444A (en) * | 2017-12-28 | 2018-04-24 | 江西高创保安服务技术有限公司 | A kind of face super-resolution reconstruction method based on deep learning |
| CN107977932A (en) * | 2017-12-28 | 2018-05-01 | 北京工业大学 | It is a kind of based on can differentiate attribute constraint generation confrontation network face image super-resolution reconstruction method |
| CN108268845A (en) * | 2018-01-17 | 2018-07-10 | 深圳市唯特视科技有限公司 | A kind of dynamic translation system using generation confrontation network synthesis face video sequence |
| CN108446609A (en) * | 2018-03-02 | 2018-08-24 | 南京邮电大学 | A kind of multi-angle human facial expression recognition method based on generation confrontation network |
| CN108319932A (en) * | 2018-03-12 | 2018-07-24 | 中山大学 | A kind of method and device for the more image faces alignment fighting network based on production |
| CN109308725A (en) * | 2018-08-29 | 2019-02-05 | 华南理工大学 | A kind of system that expression interest figure in mobile terminal generates |
| CN109544656A (en) * | 2018-11-23 | 2019-03-29 | 南京信息工程大学 | A kind of compressed sensing image rebuilding method and system based on generation confrontation network |
| CN110288537A (en) * | 2019-05-20 | 2019-09-27 | 湖南大学 | Facial image complementing method based on the depth production confrontation network from attention |
| CN110222668A (en) * | 2019-06-17 | 2019-09-10 | 苏州大学 | Based on the multi-pose human facial expression recognition method for generating confrontation network |
Non-Patent Citations (2)
| Title |
|---|
| YONGMING RAO ET AL.: "Learning Discriminative Aggregation Network for Video-based Face Recognition", 《2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
| ZHAOXIANG LIU ET AL.: "Fine-grained Attention-based Video Face Recognition", 《ARXIV:1905.01796V1 [CS.CV] 6 MAY 2019》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112288032A (en) * | 2020-11-18 | 2021-01-29 | 上海依图网络科技有限公司 | Method and device for quantitative model training based on generation of confrontation network |
| CN112288032B (en) * | 2020-11-18 | 2022-01-14 | 上海依图网络科技有限公司 | Method and device for quantitative model training based on generation of confrontation network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chen et al. | An edge traffic flow detection scheme based on deep learning in an intelligent transportation system | |
| US10417526B2 (en) | Object recognition method and device | |
| WO2019228317A1 (en) | Face recognition method and device, and computer readable medium | |
| WO2021258989A1 (en) | Facial anti-counterfeiting recognition method and apparatus, and device and storage medium | |
| CN110222573B (en) | Face recognition method, device, computer equipment and storage medium | |
| CN111695415A (en) | Construction method and identification method of image identification model and related equipment | |
| Tripathi et al. | Real time security framework for detecting abnormal events at ATM installations | |
| CN111428771B (en) | Video scene classification method and device and computer-readable storage medium | |
| CN110633698A (en) | Infrared picture identification method, equipment and medium based on loop generation countermeasure network | |
| Xia et al. | Face occlusion detection based on multi-task convolution neural network | |
| CN111368672A (en) | Construction method and device for genetic disease facial recognition model | |
| EP3779775B1 (en) | Media processing method and related apparatus | |
| WO2023179429A1 (en) | Video data processing method and apparatus, electronic device, and storage medium | |
| CN111160555A (en) | Processing method and device based on neural network and electronic equipment | |
| CN113705596A (en) | Image recognition method and device, computer equipment and storage medium | |
| CN111738074B (en) | Pedestrian attribute identification method, system and device based on weak supervision learning | |
| Vallimeena et al. | CNN algorithms for detection of human face attributes–a survey | |
| Fatima et al. | Global feature aggregation for accident anticipation | |
| CN111652181A (en) | Target tracking method and device and electronic equipment | |
| Shao et al. | COVAD: Content-oriented video anomaly detection using a self attention-based deep learning model | |
| Abou Loume et al. | Facial recognition in the opening of a door using deep learning and a cloud service | |
| Qin et al. | Application of video scene semantic recognition technology in smart video | |
| CN113657272A (en) | Micro-video classification method and system based on missing data completion | |
| CN110647864A (en) | Single multi-graph feature recognition method, equipment and medium based on generation countermeasure network | |
| CN110866609A (en) | Interpretation information acquisition method, device, server 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200103 |