CN110197128A - The recognition of face architecture design method planned as a whole based on edge calculations and cloud - Google Patents
The recognition of face architecture design method planned as a whole based on edge calculations and cloud Download PDFInfo
- Publication number
- CN110197128A CN110197128A CN201910378707.9A CN201910378707A CN110197128A CN 110197128 A CN110197128 A CN 110197128A CN 201910378707 A CN201910378707 A CN 201910378707A CN 110197128 A CN110197128 A CN 110197128A
- Authority
- CN
- China
- Prior art keywords
- model
- face
- feature
- recognition
- cloud
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—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
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
A kind of recognition of face architecture design method planned as a whole based on edge calculations and cloud of disclosure of the invention, it include: the model training step of fringe node, according to the concept of edge calculations, in the different corresponding lightweight deep learning networks of region point design, and partial model training is carried out；Data upload corresponding model result for the calculated result of multistage fringe node with model transmitting step；The data of upload and model are trained study, obtain a Global Face identification model by cloud model training step, the form planned as a whole using cloud；And the final feedback step of task.Edge calculations technology and cloud are planned as a whole form and are applied in the recognition of face task under big data by the present invention, the recognition of face framework that the response time is fast for constructing one, accuracy rate is high, it is handled by the way that basic task is placed on Data Frontend, reduces operating lag caused by data transmission；The partial model for integrating each fringe node improves whole face recognition accuracy rate.
Description
Technical field
The present invention relates to deep learning applied technical fields, and in particular to a kind of people planned as a whole based on edge calculations and cloud
Face identifies architecture design method.
Background technique
Video monitoring is universal in national big and medium-sized cities in recent years, and is widely used to Crime prevention and control system construction
In, and become the powerful technique means of public security organ's solving criminal cases.Especially case is robbed in social event, particularly serious case and two
In, the trail of evidence obtained in surveillance video plays a key effect for the quick detection of case.In order to grasp weight comprehensively
The action trail of point personnel, needs the key area in each counties and cities, the whole province, such as railway station, builds to bus station a large amount of portrait
Therefore bayonet can all generate a large amount of image data daily.Due to the increase of the super molal quantity magnitude of portrait big data, acquired
To face monitor video data can not all be aggregated into the total platform in cloud, and carry out by way of cloud computing processing point
Analysis.This is because the transmission of data needs huge broadband to occupy, and Cloud Server can be because huge data be analyzed in processing
Time delay and real-time response can not be made to emergency event.In order to the recognition of face subsystem in conjunction with each key area
Monitored results, there is an urgent need to a kind of new architectures for integrating regional model and result to construct Intelligent human-face discriminance analysis system.
In recent years, the range of artificial intelligence field referred state key construction.This imply that artificial intelligence and phase
The combination for closing industry is inexorable trend that China develop towards intelligent direction, to push industry towards in terms of intelligent, automation
Development is of great significance.It is most importantly directed to different industry tasks in artificial intelligence field, designs corresponding depth
Practise network model.The basic characteristics of deep learning network are that models fitting ability is strong, contain much information and precision height, be can satisfy not
Different demands in the same industry.A large amount of portrait monitoring data can provide good basis for deep learning network training, but
Bring the training problem of data transmission and Deep model.How it to be directed to above-mentioned problem, designs corresponding reasonable deep learning
Area data and partial model are planned as a whole to cloud, and then the sound of recognition of face subsystem in region may be implemented by the network architecture
The raising of speed is answered, and the raising of the face recognition accuracy rate in cloud may be implemented.
Summary of the invention
The purpose of the present invention is to solve drawbacks described above in the prior art, provide a kind of based on edge calculations and cloud
The recognition of face architecture design method of pool utilizes local data carry out office in each fringe node first with edge calculations
Portion's model training and Model checking；Pass through a small amount of shape for differentiating result and partial model to cloud, planning as a whole using cloud of transmission again
Formula, the training higher human face recognition model of accuracy.
The purpose of the present invention can be reached by adopting the following technical scheme that:
A kind of recognition of face architecture design method planned as a whole based on edge calculations and cloud, comprising: the model of fringe node
Training step, data and model transmitting step, cloud model training step and the final feedback step of task；The mould of fringe node
Type training step is mainly the concept according to edge calculations, in the different corresponding lightweight deep learning nets of region point design
Network, and carry out partial model training；Data and model transmitting step are uploaded primarily directed to the calculated result of multistage fringe node
Corresponding model result；Cloud model training step is mainly the form for utilizing cloud to plan as a whole, by the data of upload and model into
Row training study, obtains a Global Face identification model.
Specifically, operating procedure is as follows:
S1, according to mission requirements, the fringe node in design focal point region, mainly arrangement intelligent front end equipment, including take the photograph
As head and lightweight service device；
S2, in fringe node region, arrange video image acquisition front end, mutually tied with the lightweight service device in fringe node
It closes, fringe node human face recognition model is trained, and basic task result is fed back；
S3, data and model transmitting step: small in order to occupy broadband needed for the transmission of data, the content of transmission is benefit
The basic model finished with a small amount of training data of marginal point each in step S2 and the training of each fringe node；
S4, cloud model training step: data acquired in the method using step S3 and basic model are united with cloud
The form raised, constructs the deep learning network model in a cloud, and is trained；
The final feedback step of S5, task: test facial image is carried out model survey by the model trained according to step S2
Examination obtains task basic result, then the data for needing further to analyze is uploaded to cloud；Cloud utilizes the data of upload
Step S4 trains the model finished further to analyze, and task result is issued to each fringe node, to fringe node knot
Fruit is modified.
Further, the fringe node selection in step S1: due to being directed to public safety problem, so fringe node
Layout area be mainly the big each key area of mobility of people, including be not limited to cell entrance, security protection inside community
Point, traffic intersection camera, railway station point etc..
Further, the fringe node in step S1 is to make the processing to data that can be positioned as close to data
Acquisition position reduces the time loss of data transmission and improves the processing capability in real time of data, therefore, the lightweight arranged
Server should have certain data-handling capacity, can using local data base to the human face recognition model in fringe node into
Row training.
Further, the process of step S2 is as follows:
S21, human face data is acquired using video monitoring intelligent front end, is mentioned by Face detection algorithm and face
Algorithm is taken, corresponding face image data x is obtainedi；
S22, the fringe node for assuming setting are n, and the training data of each node is xi, corresponding lightweight local number
It is according to libraryThen by obtaining n human face recognition model after training, wherein the human face recognition model of i-th of fringe node is
Li=Locali(xi), i=1,2,3 ..., n (1)
Wherein, Locali() indicates the feature extraction submodel of i-th of fringe node, matchi() indicates i-th of side
The Matching Model of edge node, LiIndicate by the data characteristics extracted after model, SiIt is human face recognition model processing result；
S23, the recognition result of each fringe node is subjected to task feedback in corresponding fringe node.
Further, in step S3, the data of required upload include the feature extraction submodel Local=that training finishes
[Local1,Local2,...,Localn] and a small amount of training sample X=[x '1,x′2,...,x′n]；And required is upper
It passes channel and only needs common communication bandwidth and normal network transmission speed.
Further, in step S4, the model and training sample uploaded using each fringe node is planned as a whole using cloud
Form, construct a global deep learning network model.By carrying out Combined Treatment to the model of upload, mould is then utilized
The method of type optimization, optimizes overall model, improves the accuracy rate of recognition of face.
The structured design process of network is as follows in step S4:
S41, model Combined Treatment: by carrying out linear combination to the fringe node human face recognition model of all uploads, into
And realize the feature extraction functions to the sample data of input.Since fringe node human face recognition model is to utilize a large amount of local numbers
According to trained as a result, structure and the parameter of partial model are held essentially constant, to each partial model when therefore combining beyond the clouds
The face characteristic being calculated is spliced, and the input feature vector F of model optimization is obtained:
Loc′i=Locali(x′i), i=1,2 .., n
F=[Loc '1,Loc′2,...,Loc′n] (3)
S42, model optimization processing:
Firstly, one autocoder of building, autocoder includes feature coding layer, hidden layer and feature reconstruction layer,
And first autocoder is trained；Secondly, the feature coding layer of the autocoder finished in a upper training it
A new feature coding layer is added afterwards, and new feature reconstruction layer is added in corresponding feature reconstruction layer；Again, to stacking and
At network model be trained, and repeat previous action；Then, by successive ignition training, stack is finally constituted certainly
Dynamic encoder model；Finally, the feature reconstruction network in the stack autocoder model finished will be trained to give up, in feature
Classifier is added after coding network, realizes the classification to input feature vector F.
Further, the autocoder includes feature coding layer, hidden layer and feature reconstruction layer, reconstructs feature
F' is obtained by following autocoder model:
H=σf(wfF+bf) (4)
F '=σr(wrH+br) (5)
Wherein, H is hidden layer output, σf,σrIt is the activation primitive of feature coding layer and the activation letter of feature reconstruction layer respectively
Number, wf,bfAnd wr,brIt is feature coding layer parameter and feature reconstruction layer parameter respectively.
Further, the loss function of the autocoder is constrained with formula (6),
By continuing to optimize loss function, finally stabilised autocoder parameter is obtained.
Further, the sorter network of input feature vector F can use and Softmax layers are added after feature coding network
Carry out sort operation.
The present invention has the following advantages and effects with respect to the prior art:
The form that the technology of edge calculations and cloud are planned as a whole is applied in the recognition of face task under big data by the present invention,
The recognition of face framework that the response time is fast for constructing one, accuracy rate is high；It, can be by basic task with the method for edge calculations
It is placed on Data Frontend to be handled, reduces operating lag caused by data transmission；The information planned as a whole using cloud can be integrated each
The partial model of a fringe node improves whole face recognition accuracy rate.
Detailed description of the invention
Fig. 1 is the schematic diagram for the recognition of face framework planned as a whole in the present invention based on edge calculations and cloud；
Fig. 2 is the recognition of face network frame schematic diagram of edge calculations point in the present invention；
Fig. 3 is the network frame schematic diagram of autocoder in the present invention；
Fig. 4 is stack autocoder training pattern schematic diagram in the present invention；
Fig. 5 is recognition of face disaggregated model block schematic illustration in cloud in the present invention；
Fig. 6 is that multistage fringe node calculates the recognition of face training configuration diagram planned as a whole with cloud in the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Embodiment
The present embodiment relates generally to following a few class technologies: 1) edge calculations technology in the technology of network architecture design: utilizing
Edge calculations technology responds task close to Data Frontend, reduces operating lag；2) the recognition of face framework that cloud is planned as a whole:
The partial model for combining multiple fringe nodes is trained to the joint model by way of constructing autocoder, is improved and is known
Other accuracy rate.
The network architecture design is based on TensorFlow frame and Pycharm develops environment: TensorFlow frame is base
In the development frame of python language, can conveniently and efficiently build reasonable deep learning network, at the same have well across
Platform interaction capabilities.TensorFlow provides numerous encapsulation functions in deep learning framework and all kinds of image processing functions connect
Mouthful, including the relevant image processing function of OpenCV.TensorFlow frame be able to use simultaneously GPU model is trained and
Verifying, improves the efficiency of calculating.
Pycharm under windows platform or Linux platform develops environment development environment (IDE), is current depth
Degree study one of network design and the first choice of exploitation.Pycharm for client provide new template, design tool and test and
Debugging tool, while the interface for calling directly remote server can be provided for client.
A kind of recognition of face construction design method planned as a whole based on edge calculations and cloud is provided in the present embodiment, for appointing
Business demand disposes multiple fringe nodes, and each fringe node includes Data Frontend and lightweight service device；Firstly, each edge section
Point is trained fringe node human face recognition model using local data；Secondly, to multiple fringe node human face recognition models
It is transmitted with a small amount of data, is pooled to cloud server；Again, using cloud plan as a whole form, by the data of upload with
Model is trained study, obtains a Global Face identification model；Finally, when carrying out detection identification to human face data, edge
Node is analyzed using model, carries out simple feedback to task, and cloud uploads node using Global Face identification model
Suspicious data analyzed, and return result to each fringe node.Such as the overview flow chart that Fig. 1 is this method.Specifically
Steps are as follows:
Step 1: being directed to each fringe node, network used is relatively simple deep learning network, whole network packet
Containing 4 convolutional layers, 3 pond layers and 1 classification layer.Firstly, facial image to be normalized to the figure of 3*64*64 size
Then piece passes through 4 convolutional layers and 3 pond layers respectively, obtains the face characteristic information of higher-dimension, finally, classifying by 1
Layer, obtains the recognition result of face.Specifically as shown in Fig. 2, Fig. 2 is the recognition of face network frame of edge calculations point.
Step 2: needing to construct autocoder during model optimization.Each autocoder includes that feature is compiled
Code layer, hidden layer and feature reconstruction layer.Specifically as shown in figure 3, Fig. 3 is the basic framework of autocoder.
Step 3: during model optimization, the building of stack autocoder training pattern through the following steps that
Complete: firstly, one autocoder of building, autocoder includes feature coding layer, hidden layer and feature reconstruction layer, and
First autocoder is trained；Secondly, after the feature coding layer for the autocoder that a upper training finishes
A new feature coding layer is added, and new feature reconstruction layer is added in corresponding feature reconstruction layer；Again, to stacking
Network model be trained, and repeat previous action；Then, by successive ignition training, it is automatic to finally constitute stack
Encoder model；Finally, the feature reconstruction network in the stack autocoder model finished will be trained to give up, compiled in feature
Classifier is added after code network, realizes the classification to input feature vector F.Specifically as shown in figure 4, Fig. 4 is stack autocoding
Device training pattern.
Step 4: the recognition of face output of final world model is will to train the stack autocoder model finished
Feature reconstruction network is given up, and Softmax classifier is added after feature coding network, realizes point to the face characteristic of input
Class.Specifically as shown in figure 5, Fig. 5 is cloud recognition of face disaggregated model frame.
Step 5: the fringe node model trained according to above step two, carries out model measurement for test facial image,
Acquisition task basic result, then the data for needing further to analyze are uploaded into cloud；Cloud utilizes step to the data of upload
Four train the world model finished further to analyze, and task result is issued to each fringe node, to fringe node knot
Fruit is modified.Specifically as shown in fig. 6, Fig. 6 is the recognition of face training framework of multistage fringe node calculating with cloud pool.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (8)
1. a kind of recognition of face architecture design method planned as a whole based on edge calculations and cloud, which is characterized in that the face
Identify architecture design method the following steps are included:
S1, according to mission requirements, the fringe node in design focal point region simultaneously arranges intelligent front end equipment, including video image acquisition
Front end and lightweight service device；
S2, fringe node human face recognition model is trained using local data in each fringe node；And to basic task
As a result it is fed back；
S3, each fringe node human face recognition model and basic task result data are transmitted, is pooled to cloud server
The data of upload and model are trained study by S4, the form planned as a whole using cloud, obtain a Global Face identification
Model；
S5, input human face data carry out detection identification, fringe node is analyzed using human face recognition model, to human face data into
Row simple feedback, while cloud is analyzed using the suspicious data that Global Face identification model uploads node, and by result
Feed back to each fringe node.
2. the recognition of face architecture design method according to claim 1 planned as a whole based on edge calculations and cloud, feature
It is, in the step S1, when being directed to public safety problem, the layout area of fringe node is big each of mobility of people
Security protection point, traffic intersection camera, railway station point inside a key area, including cell entrance, community.
3. the recognition of face architecture design method according to claim 1 planned as a whole based on edge calculations and cloud, feature
It is, the process of the step S2 is as follows:
S21, human face data is acquired using video monitoring intelligent front end, is calculated by Face detection algorithm and face extraction
Method obtains corresponding face image data xi；
S22, the fringe node for assuming setting are n, and the training data of the i-th node is xi, i=1,2 ..., n, corresponding light weight
Grade local data base beThen by can get n human face recognition model, the recognition of face mould of i-th of fringe node after training
Type is
Li=Locali(xi), i=1,2,3 ..., n (1)
Wherein, Locali() indicates the feature extraction submodel of i-th of fringe node, matchi() indicates i-th of edge section
The Matching Model of point, LiIndicate by the data characteristics extracted after model, SiIt is human face recognition model processing result；
S23, the recognition result of each fringe node is subjected to task feedback in corresponding fringe node.
4. the recognition of face architecture design method according to claim 3 planned as a whole based on edge calculations and cloud, feature
It is, in the step S3, the data of required upload include the feature extraction submodel Local=[Local that training finishes1,
Local2,...,Localn] and a small amount of training sample X=[x '1,x′2,...,x′n]；And required upload channel is only
Need common communication bandwidth and normal network transmission speed.
5. the recognition of face architecture design method according to claim 1 planned as a whole based on edge calculations and cloud, feature
It is, the step S4 process is as follows:
S41, model Combined Treatment, by carrying out linear combination, realization pair to the fringe node human face recognition model of all uploads
The feature extraction functions of the sample data of input, it is special to the face that each fringe node human face recognition model is calculated beyond the clouds
Sign is spliced, and the input feature vector of model optimization is obtained:
Loc′i=Locali(x′i), i=1,2 .., n
F=[Loc '1,Loc′2,...,Loc′n] (3)
S42, model optimization processing: firstly, constructing multiple autocoders, each autocoder includes feature coding layer and spy
Reconstruction of layer is levied, and first autocoder is trained；Secondly, the feature of the autocoder finished in a upper training
A new feature coding is added layer by layer after coding layer, and new feature reconstruction layer is added in corresponding feature reconstruction layer；Again
It is secondary, network model made of stacking is trained, and repeat previous action；Then, pass through successive ignition training, final structure
At stack autocoder model；Finally, the feature reconstruction network in the stack autocoder model that training is finished
Give up, classifier is added after feature coding network, realizes the classification to input feature vector F.
6. the recognition of face architecture design method according to claim 5 planned as a whole based on edge calculations and cloud, feature
It is, the autocoder includes input layer, hidden layer and output layer, and reconstruct feature F' is by following autocoding
Device model is obtained:
H=σf(wfF+bf) (4)
F'=σr(wrH+br) (5)
Wherein, H is hidden layer output, σf,σrIt is the activation primitive of feature coding network and the activation letter of feature reconstruction network respectively
Number, wf,bfAnd wr,brIt is feature coding network parameter and feature reconstruction network parameter respectively.
7. the recognition of face architecture design method according to claim 5 planned as a whole based on edge calculations and cloud, feature
It is, the loss function of the autocoder is constrained with formula (6),
By continuing to optimize loss function, finally stabilised autocoder parameter is obtained.
8. the recognition of face architecture design method according to claim 5 planned as a whole based on edge calculations and cloud, feature
It is, the classifier is the Softmax layer for carrying out sort operation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910378707.9A CN110197128A (en) | 2019-05-08 | 2019-05-08 | The recognition of face architecture design method planned as a whole based on edge calculations and cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910378707.9A CN110197128A (en) | 2019-05-08 | 2019-05-08 | The recognition of face architecture design method planned as a whole based on edge calculations and cloud |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110197128A true CN110197128A (en) | 2019-09-03 |
Family
ID=67752531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910378707.9A Pending CN110197128A (en) | 2019-05-08 | 2019-05-08 | The recognition of face architecture design method planned as a whole based on edge calculations and cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110197128A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569911A (en) * | 2019-09-11 | 2019-12-13 | 深圳绿米联创科技有限公司 | Image recognition method, device, system, electronic equipment and storage medium |
| CN110581898A (en) * | 2019-10-09 | 2019-12-17 | 成都康胜思科技有限公司 | internet of things data terminal system based on 5G and edge calculation |
| CN110719210A (en) * | 2019-12-05 | 2020-01-21 | 赣江新区智慧物联研究院有限公司 | Industrial equipment predictive maintenance method based on cloud edge cooperation |
| CN111030861A (en) * | 2019-12-11 | 2020-04-17 | 中移物联网有限公司 | Edge calculation distributed model training method, terminal and network side equipment |
| CN111199279A (en) * | 2019-10-30 | 2020-05-26 | 山东浪潮人工智能研究院有限公司 | Cloud edge calculation and artificial intelligence fusion method and device for police service industry |
| CN111212264A (en) * | 2019-12-27 | 2020-05-29 | 中移（杭州）信息技术有限公司 | Image processing method and device based on edge calculation and storage medium |
| CN111400040A (en) * | 2020-03-12 | 2020-07-10 | 重庆大学 | Industrial Internet system based on deep learning and edge calculation and working method |
| CN111461694A (en) * | 2020-03-17 | 2020-07-28 | 上海大学 | Dish identification and pricing system and method based on multi-level learning model |
| CN111950416A (en) * | 2020-07-31 | 2020-11-17 | 中国工商银行股份有限公司 | Face recognition method and system based on block chain |
| CN111985650A (en) * | 2020-07-10 | 2020-11-24 | 华中科技大学 | Activity recognition model and system considering both universality and individuation |
| CN112487082A (en) * | 2019-09-12 | 2021-03-12 | 华为技术有限公司 | Biological feature recognition method and related equipment |
| CN112633513A (en) * | 2019-10-09 | 2021-04-09 | 中国电信股份有限公司 | Data processing method, system, device and computer readable storage medium |
| CN112784718A (en) * | 2021-01-13 | 2021-05-11 | 上海电力大学 | Insulator state identification method based on edge calculation and deep learning |
| CN113128529A (en) * | 2019-12-30 | 2021-07-16 | 中移(成都)信息通信科技有限公司 | Livestock identity recognition model training method, recognition method, system and device |
| CN113515368A (en) * | 2020-08-23 | 2021-10-19 | 陈顺发 | Data integration method combining big data and edge calculation and storage medium |
| WO2021218384A1 (en) * | 2020-04-29 | 2021-11-04 | 华为技术有限公司 | Algorithm update method, system and device |
| WO2022111398A1 (en) * | 2020-11-26 | 2022-06-02 | 华为技术有限公司 | Data model training method and apparatus |
| US11600111B2 (en) | 2020-06-01 | 2023-03-07 | Paralaxiom Technologies Private Limited | System and method for face recognition |
| CN117081502A (en) * | 2023-10-16 | 2023-11-17 | 锦浪科技股份有限公司 | Photovoltaic direct current arc monitoring platform and control method thereof |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770263A (en) * | 2017-10-16 | 2018-03-06 | 电子科技大学 | A kind of internet-of-things terminal safety access method and system based on edge calculations |
| CN108012121A (en) * | 2017-12-14 | 2018-05-08 | 安徽大学 | A kind of edge calculations and the real-time video monitoring method and system of cloud computing fusion |
| CN108446680A (en) * | 2018-05-07 | 2018-08-24 | 西安电子科技大学 | A kind of method for secret protection in face authentication system based on edge calculations |
| CN109492536A (en) * | 2018-10-12 | 2019-03-19 | 大唐高鸿信息通信研究院（义乌）有限公司 | A kind of face identification method and system based on 5G framework |
| CN109543829A (en) * | 2018-10-15 | 2019-03-29 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Method and system for hybrid deployment of deep learning neural network on terminal and cloud |
| CN109613428A (en) * | 2018-12-12 | 2019-04-12 | 广州汇数信息科技有限公司 | It is a kind of can be as system and its application in motor device fault detection method |
-
2019
- 2019-05-08 CN CN201910378707.9A patent/CN110197128A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107770263A (en) * | 2017-10-16 | 2018-03-06 | 电子科技大学 | A kind of internet-of-things terminal safety access method and system based on edge calculations |
| CN108012121A (en) * | 2017-12-14 | 2018-05-08 | 安徽大学 | A kind of edge calculations and the real-time video monitoring method and system of cloud computing fusion |
| CN108446680A (en) * | 2018-05-07 | 2018-08-24 | 西安电子科技大学 | A kind of method for secret protection in face authentication system based on edge calculations |
| CN109492536A (en) * | 2018-10-12 | 2019-03-19 | 大唐高鸿信息通信研究院（义乌）有限公司 | A kind of face identification method and system based on 5G framework |
| CN109543829A (en) * | 2018-10-15 | 2019-03-29 | 华东计算技术研究所（中国电子科技集团公司第三十二研究所） | Method and system for hybrid deployment of deep learning neural network on terminal and cloud |
| CN109613428A (en) * | 2018-12-12 | 2019-04-12 | 广州汇数信息科技有限公司 | It is a kind of can be as system and its application in motor device fault detection method |
Non-Patent Citations (3)
| Title |
|---|
| CHUNHUI DING 等: "A Novel Two-stage Learning Pipeline for Deep Neural Networks", 《NEURAL PROCESS LETT》 * |
| NASIF MUSLIM 等: "Face recognition in the Edge Cloud", 《ASSOCIATION FOR COMPUTING MACHINERY》 * |
| 丁春辉: "基于深度学习的暴力检测及人脸识别方法研究", 《中国博士学位论文全文数据库 信息科技辑（月刊）》 * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569911B (en) * | 2019-09-11 | 2022-06-07 | 深圳绿米联创科技有限公司 | Image recognition method, device, system, electronic equipment and storage medium |
| CN110569911A (en) * | 2019-09-11 | 2019-12-13 | 深圳绿米联创科技有限公司 | Image recognition method, device, system, electronic equipment and storage medium |
| CN112487082A (en) * | 2019-09-12 | 2021-03-12 | 华为技术有限公司 | Biological feature recognition method and related equipment |
| CN112487082B (en) * | 2019-09-12 | 2024-04-12 | 华为云计算技术有限公司 | Biological feature recognition method and related equipment |
| CN110581898A (en) * | 2019-10-09 | 2019-12-17 | 成都康胜思科技有限公司 | internet of things data terminal system based on 5G and edge calculation |
| CN110581898B (en) * | 2019-10-09 | 2022-04-08 | 成都康胜思科技有限公司 | Internet of things data terminal system based on 5G and edge calculation |
| CN112633513A (en) * | 2019-10-09 | 2021-04-09 | 中国电信股份有限公司 | Data processing method, system, device and computer readable storage medium |
| CN111199279A (en) * | 2019-10-30 | 2020-05-26 | 山东浪潮人工智能研究院有限公司 | Cloud edge calculation and artificial intelligence fusion method and device for police service industry |
| CN110719210A (en) * | 2019-12-05 | 2020-01-21 | 赣江新区智慧物联研究院有限公司 | Industrial equipment predictive maintenance method based on cloud edge cooperation |
| CN110719210B (en) * | 2019-12-05 | 2020-08-28 | 赣江新区智慧物联研究院有限公司 | Industrial equipment predictive maintenance method based on cloud edge cooperation |
| CN111030861A (en) * | 2019-12-11 | 2020-04-17 | 中移物联网有限公司 | Edge calculation distributed model training method, terminal and network side equipment |
| CN111030861B (en) * | 2019-12-11 | 2022-05-31 | 中移物联网有限公司 | Edge calculation distributed model training method, terminal and network side equipment |
| CN111212264A (en) * | 2019-12-27 | 2020-05-29 | 中移（杭州）信息技术有限公司 | Image processing method and device based on edge calculation and storage medium |
| CN113128529A (en) * | 2019-12-30 | 2021-07-16 | 中移(成都)信息通信科技有限公司 | Livestock identity recognition model training method, recognition method, system and device |
| CN111400040A (en) * | 2020-03-12 | 2020-07-10 | 重庆大学 | Industrial Internet system based on deep learning and edge calculation and working method |
| CN111461694A (en) * | 2020-03-17 | 2020-07-28 | 上海大学 | Dish identification and pricing system and method based on multi-level learning model |
| CN111461694B (en) * | 2020-03-17 | 2023-07-18 | 上海大学 | Dish identification and pricing system and method based on multi-level learning model |
| WO2021218384A1 (en) * | 2020-04-29 | 2021-11-04 | 华为技术有限公司 | Algorithm update method, system and device |
| US11600111B2 (en) | 2020-06-01 | 2023-03-07 | Paralaxiom Technologies Private Limited | System and method for face recognition |
| CN111985650A (en) * | 2020-07-10 | 2020-11-24 | 华中科技大学 | Activity recognition model and system considering both universality and individuation |
| CN111950416B (en) * | 2020-07-31 | 2023-08-29 | 中国工商银行股份有限公司 | Face recognition method and system based on block chain |
| CN111950416A (en) * | 2020-07-31 | 2020-11-17 | 中国工商银行股份有限公司 | Face recognition method and system based on block chain |
| CN113515368A (en) * | 2020-08-23 | 2021-10-19 | 陈顺发 | Data integration method combining big data and edge calculation and storage medium |
| CN113515368B (en) * | 2020-08-23 | 2022-09-09 | 厦门吉快科技有限公司 | Data integration method combining big data and edge calculation and storage medium |
| WO2022111398A1 (en) * | 2020-11-26 | 2022-06-02 | 华为技术有限公司 | Data model training method and apparatus |
| CN112784718A (en) * | 2021-01-13 | 2021-05-11 | 上海电力大学 | Insulator state identification method based on edge calculation and deep learning |
| CN117081502A (en) * | 2023-10-16 | 2023-11-17 | 锦浪科技股份有限公司 | Photovoltaic direct current arc monitoring platform and control method thereof |
| CN117081502B (en) * | 2023-10-16 | 2024-02-06 | 锦浪科技股份有限公司 | Photovoltaic direct current arc monitoring platform and control method thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110197128A (en) | The recognition of face architecture design method planned as a whole based on edge calculations and cloud | |
| CN110879989B (en) | Ads-b signal target identification method based on small sample local machine learning model | |
| CN110769195B (en) | Intelligent monitoring and recognizing system for violation of regulations on power transmission line construction site | |
| CN110674772A (en) | Intelligent safety control auxiliary system and method for electric power operation site | |
| CN109255286B (en) | Unmanned aerial vehicle optical rapid detection and identification method based on deep learning network framework | |
| CN108921039A (en) | The forest fire detection method of depth convolution model based on more size convolution kernels | |
| CN110070530A (en) | A kind of powerline ice-covering detection method based on deep neural network | |
| CN107808139A (en) | A kind of real-time monitoring threat analysis method and system based on deep learning | |
| CN107085696A (en) | A kind of vehicle location and type identifier method based on bayonet socket image | |
| CN108270999A (en) | A kind of object detection method, image recognition server and system | |
| CN109935080A (en) | The monitoring system and method that a kind of vehicle flowrate on traffic route calculates in real time | |
| CN105574489A (en) | Layered stack based violent group behavior detection method | |
| CN108334831A (en) | A kind of monitoring image processing method, monitoring terminal and system | |
| CN106056069A (en) | Unmanned aerial vehicle image analysis-based forest land resource asset evaluation method and evaluation system | |
| CN108256447A (en) | A kind of unmanned plane video analysis method based on deep neural network | |
| CN109583499B (en) | Power transmission line background object classification system based on unsupervised SDAE network | |
| CN110503099A (en) | Information identifying method and relevant device based on deep learning | |
| CN115294519A (en) | Abnormal event detection and early warning method based on lightweight network | |
| Jacobsen et al. | Machine learning: paving the way for more efficient disaster relief | |
| Song et al. | PDD: Post-Disaster Dataset for Human Detection and Performance Evaluation | |
| CN104615987B (en) | A kind of the wreckage of an plane intelligent identification Method and system based on error-duration model neutral net | |
| CN116310979B (en) | Image identification method, risk management and control platform and method, and safety management and control platform | |
| Toby et al. | A survey of deep learning techniques based on drone images for the search and rescue of victims from collapsed structures | |
| Krishna et al. | Social distancing and face mask detection using deep learning | |
| CN117079342A (en) | Man-box interaction detection method and system based on two-stage target detection |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190903 |
|
| WD01 | Invention patent application deemed withdrawn after publication |