CN114998964B - Novel license quality detection method - Google Patents
Novel license quality detection method Download PDFInfo
- Publication number
- CN114998964B CN114998964B CN202210621561.8A CN202210621561A CN114998964B CN 114998964 B CN114998964 B CN 114998964B CN 202210621561 A CN202210621561 A CN 202210621561A CN 114998964 B CN114998964 B CN 114998964B
- Authority
- CN
- China
- Prior art keywords
- neural network
- data set
- feature map
- certificate
- fine
- 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.)
- Active
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
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention relates to a novel license quality detection method, which comprises the following steps: constructing a face photograph data set, and processing the face photograph data set to obtain a processed face photograph data set; inputting the processed face image data set into a bilinear fine-grained artificial neural network for feature extraction to obtain a feature map, and inputting the feature map into a classifier for output; and training the bilinear fine-grained artificial neural network based on the characteristic diagram output by the classifier to obtain a neural network model for detecting certificate defect, and detecting the certificate defect quality through the trained neural network model. The method has the advantages of high detection speed and high identification accuracy, and the photo does not need to be uploaded in the detection process and is completely carried out locally, so that the cost is reduced, and the working efficiency can be effectively improved.
Description
Technical Field
The invention relates to the technical field of image detection, in particular to a novel license quality detection method.
Background
The collection of certificate photos is an important link in the life of residents. With the rapid development of the internet of things and big data technology in recent years, self-service and online certificate photo taking (such as identity cards, outbound certificates and other related services) services are successively provided, the work efficiency of related departments is greatly improved, and the time of the masses is saved.
However, in the process of taking a self-service certificate photo, the acquisition of a face portrait has a big problem at present: due to many factors such as unskilled operation, lighting environment, acquisition system and equipment, the acquired photos often do not conform to national standards (such as head deviation, eye closing, head raising, ornament wearing and the like) and cannot be used as standard certificate photos, and the self-service certificate photos are not timely fed back in the self-service handling process, so that the handling process is interrupted, reworked or on-site workers are called for help, manpower and time are wasted, and the significance of self-service certificate photo acquisition is lost.
To solve the problem, common defects (unqualified reasons) in the certificate photo need to be accurately identified and timely fed back to guide a collector to correct the wrong certificate photo collection mode. The method is characterized in that the preset unqualified reasons of multiple common self-service shot certificates are identified from a specific certificate, and the problem is essentially the Fine-Grained classification (Fine-Grained classification) of face images. The problem of classifying fine-grained images is that subclasses under a large class are identified, and compared with the classification of the large class of images, the problem of classifying fine-grained images is difficult in that the difference degree between subclass objects is much lower than that between the large classes, for example: the figure accouterment, the state of a chignon, the dressing standard or the head posture and the facial expression of the portrait in the certificate photo, and the like. In addition, there are many uncertain factors such as attitude, illumination, background interference, etc. when the fine-grained degree of the image is classified, and these factors will cause great interference to the small difference between the sub-objects, increasing the difficulty of classification. In the self-service certificate photo acquisition system, requirements on the wearing, chignon, ornaments and the like of a shot person are met, high requirements on the head posture, facial expression and the like of the shot person are also met, a fine-grained classification algorithm can evaluate whether a photo meets the system requirements, and if the photo has defects, the defect type can be identified and fed back to the shot person.
The main method for classifying and detecting the fine granularity of the certificate license defect at present comprises the following steps:
(1) Fine-grained classification using general purpose Neural networks (DCNN), has the disadvantage of difficulty in capturing discriminative local details.
(2) Based on the positioning-recognition method, a part with discrimination is found first, and then feature extraction and classification are carried out. The method based on positioning-recognition uses the process of classifying the portrait into fine granularity for reference, the research is relatively sufficient, strong supervision learning is mostly adopted, a large amount of manpower is needed to label the key area of the image, in addition, a large amount of irrelevant areas can be generated by utilizing the bottom-up area generation method, and the speed of the algorithm can be influenced to a great extent.
(3) The high-order coding method based on the convolution characteristics carries out high-order conversion on the CNN characteristics and then carries out classification, and mainly comprises Fisher Vector, kernel fusion and the like. The high-order coding method improves the expression capability of the characteristics by performing high-order synthesis on the CNN characteristics, and is a mainstream method at the present stage.
In recent years, research based on a positioning-recognition method gradually shifts to weak supervised learning, and a positioning sub-network is constructed by methods such as an attention mechanism and channel clustering, so that a distinctive region is discovered. However, the biggest problem of the method is that multiple steps of calculation are needed, the speed is slow, and the requirement on the real-time performance of the system is not met.
In the self-service certificate photo collection system, the common certificate photo collection quality evaluation method is an online manual auditing method, namely, after the certificate photo collection is finished, the certificate photo is uploaded to an internal system and is manually audited by a background auditor, and after the audit is finished, the result is fed back to a certificate photo clerk. Disadvantages of this approach include: the efficiency is low, the auditing standard is difficult to unify, and potential safety hazards exist.
Therefore, a new method for detecting the quality of the certificate becomes a hot issue of attention for those skilled in the art.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a novel license quality detection method.
In order to achieve the purpose, the invention provides the following scheme:
a novel license quality detection method comprises the following steps:
constructing a face photograph data set, and processing the face photograph data set to obtain a processed face photograph data set;
inputting the processed face photograph data set into a bilinear fine-grained artificial neural network for feature extraction to obtain a feature map, and inputting the feature map into a classifier for output;
and training the bilinear fine-grained artificial neural network based on the characteristic diagram output by the classifier to obtain a neural network model for detecting certificate defect, and detecting the certificate defect quality through the trained neural network model.
Preferably, the face photo data set is constructed based on a face image collected in the self-service certificate photo integrated transaction machine.
Preferably, the face photograph data set is subjected to data set enhancement processing for increasing the number of data sets and improving the precision of a training model, and the data set enhancement processing method comprises rotating, mirroring, clipping and changing the contrast.
Preferably, the bilinear fine-grained artificial neural network comprises two convolutional neural networks, the two convolutional neural networks are respectively a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used for acquiring a low-level feature map, and the second convolutional neural network is used for acquiring a detail feature map.
Preferably, the first convolutional neural network employs VGGNet, and the second convolutional neural network employs DenseNet.
Preferably, in the DenseNet, the size of the feature map is matched between every two adjacent Dense blocks by using Batch +1 × 1Conv +2 × 2 AvgPool.
Preferably, the face photograph data set is respectively input into the first convolution neural network and the second convolution neural network to respectively obtain a low-level feature map and a detail feature map, then the low-level feature map and the detail feature map are fused, the outer product of the low-level feature map and the detail feature map is calculated and then input into a full connection layer and a Softmax classifier, and fine-grained classification of the face image is carried out for identifying unqualified reasons for photographing the photograph.
Preferably, all images in the face photograph data set are divided into a training set, a testing set and a verification set, based on a python3.6+ Tensorflow2.0 programming environment, the training set is input into a bilinear fine-grained artificial neural network for detecting certificate cheating, a neural network model for detecting the certificate cheating is trained, test set verification is carried out based on the testing set and the verification set, and certificate photograph quality detection is carried out through the trained bilinear fine-grained artificial neural network.
The invention has the beneficial effects that:
the invention can detect and feed back in real time, detect the images which do not meet the requirements, classify the defect types at a fine granularity, and remind the user to change the posture so as to meet the national standard. The method has the advantages of high detection speed and high identification accuracy, and the photo does not need to be uploaded in the detection process and is completely carried out locally, so that the cost is reduced, and the working efficiency can be effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of a method of an embodiment of the present invention;
FIG. 2 is a schematic diagram of a bilinear neural network structure for detecting certificate fraud in an embodiment of the present invention;
fig. 3 is a schematic diagram of a VGGNet network structure according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a DenseNet network structure according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In order to make the aforementioned objects, features and advantages of the present invention more comprehensible, the present invention is described in detail with reference to the accompanying drawings and the detailed description thereof.
Referring to the attached figure 1, the invention discloses a novel license quality detection method, which is based on a bilinear artificial neural network model, namely, a network detects local area detection and positioning of a face image and extracts global features; and the other network is responsible for extracting a local high-level feature map of the face image. The two networks are coordinated with each other to complete the most important task in the classification process of the certificate cheating fine-grained images: and detecting a local area and extracting features. The detection method has great application value in the fields of self-service identification photo systems, remote Face living body detection (Face Anti-panning) of mobile platforms and the like.
Constructing a face photograph data set, and processing the face photograph data set to obtain a processed face photograph data set;
at present, human Face data sets commonly used in the field of artificial intelligence are more common, such as AR Face, LFW +, wire Face and the like, but due to factors such as privacy and the like, the human Face data sets used for a certificate photo system are almost not available, and the image collection is difficult. The method is based on a self-service certificate photo transaction all-in-one machine, a large number of face images are collected in the long-term use process, and a face certificate photo data set is constructed. In the certificate cheating image in the embodiment, four groups of defects are eye closing, head raising, mouth opening and strabismus respectively.
In addition, the data needs to be marked manually, and the workload is large. Therefore, before deep learning calculation, data set enhancement (data augmentation) processing is performed on limited data to avoid the problem of model accuracy reduction caused by insufficient data set sample number, and besides, the data set enhancement can avoid the overfitting problem in the training process to a certain extent. The data set enhancement method mainly comprises the means of rotating, mirroring, clipping, changing contrast and the like. After the data set is enhanced, about 1000 samples can be increased to about 5000 samples, and the condition for training the neural network model is initially met.
Inputting the processed face photograph data set into a neural network for feature extraction to obtain a feature map, and inputting the feature map into a classifier for output;
the bilinear fine-grained classification network used in the method is to input an image into two Convolutional Neural Networks (CNN), calculate an outer product of an output feature graph and then enter a Full connection layer (Full connection layer) to obtain a score of each category. The two convolutional neural networks are respectively a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is a network with a shallow layer to obtain a low-layer characteristic diagram of the image, and an object-level classification result is obtained; the second convolutional neural network uses a high-level complex network, mainly extracts the detail features of the image, then fuses the feature maps extracted twice, and inputs the feature maps into a full connection layer and a Softmax classifier to obtain a final classification result. FIG. 2 is a schematic diagram of a bilinear neural network structure for detecting certificate cheating, and feature maps of two convolutional neural networks are fused and then input into a classifier for output.
The digital image and the characteristic diagram are stored and calculated in a matrix form, so that the outer product of the characteristic diagram is the outer product of the matrix. Matrix outer product has many operation forms, the most common outer product form is matrix multiplication A × B, but the two characteristic diagrams of the bilinear neural network herein hardly satisfy the condition of matrix multiplication, namely the number of rows of A is equal to the number of columns of B. Therefore, in this embodiment, a kronecker product of a matrix is used as a fusion method of high-level and low-level feature maps, and if a and B are feature maps of two arbitrary dimensions, the kronecker product is defined as:
in this embodiment, VGGNet and DenseNet are used to form the required bilinear network. The generalization performance is good, the hierarchy is shallow, the migration to other image recognition items is easy, the structure is simple, and as shown in fig. 3, networks of VGG16 and VGG19 are commonly used.
ResNet (residual error network) in a deep neural network is frequently used, the network level can reach up to 152 levels, the method is characterized in that a feed-forward back propagation algorithm is adopted, although the method has no advantage in the representation mode of the model, the method can allow the neural network model to have deeper levels without gradient dispersion and gradient explosion, and the ResNet is widely used. In this embodiment, densnet, which occurs later than ResNet, is used, and compared with Residual Block of ResNet, densnet uses density Block, in each density Block, there is a direct connection between any two layers, that is, the input of each layer of the network is the union of the outputs of all the previous layers, and the feature map learned by the layer is directly transmitted to all the subsequent layers as input. Through dense connection, the problem of gradient disappearance is relieved, feature propagation is enhanced, feature multiplexing is encouraged, and parameter quantity is greatly reduced.
In dealing with the problem of mismatch in the number or size of the signatures, resNet expands the number of signatures with zero padding or with 1 × 1Conv (convolution), while DenseNet matches the size of the signatures between two Dense blocks using Batch +1 × 1Conv +2 × 2AvgPool as the Transition layer. This makes full use of the learned feature map without adding unnecessary extrinsic noise using zero padding. The DenseNet network structure is shown in fig. 4.
In conclusion, the two network structures are calculated in parallel, the outputs of the two networks are effectively fused and then output through the Softmax classifier, and the problem of fine-grained classification of the certificate photo defect data set can be effectively solved.
And training the neural network based on the characteristic diagram output by the classifier to obtain a neural network model for detecting certificate cheating diseases, and detecting the certificate photo quality through the trained neural network model. 60% (about 3000) of all images in the identification photograph dataset were taken as training set, 20% as verification set, and finally 20% as test set. Based on a Python3.6+ Tensorflow2.0 programming environment, all images of a training set are input into the bilinear fine-grained artificial neural network for certificate cheating detection, and a neural network model for certificate cheating detection can be trained.
The method can detect more than ten irregular actions such as eye closing, head lowering, head raising, head tilting, mouth opening, strabismus, glasses wearing, ear shielding and the like which are common in the use process of the self-service certificate photo processing all-in-one machine. The highest model training accuracy rate in the experiment can reach more than 85%, the highest test accuracy rate can reach more than 78%, the time for testing a single image is about 0.5s, and the daily requirements of the self-service certificate photo transaction all-in-one machine are met.
When a user transacts the certificate photo and needs to acquire a face image, the system can perform real-time detection and real-time feedback, detect the image which does not meet the requirement, classify the types of the defects in a fine-grained manner, and remind the user to change the posture so as to meet the national standard. The method has the advantages of high detection speed and high identification accuracy, and the photo is not required to be uploaded in the detection process and is completely carried out locally, so that the cost is reduced, and the working efficiency of the self-service certificate photo transaction all-in-one machine can be effectively improved.
The above-described embodiments are merely illustrative of the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements of the technical solutions of the present invention can be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are within the scope of the present invention defined by the claims.
Claims (6)
1. A novel license quality detection method is characterized by comprising the following steps:
constructing a face photograph data set, and processing the face photograph data set to obtain a processed face photograph data set;
inputting the processed face image data set into a bilinear fine-grained artificial neural network for feature extraction to obtain a feature map, and inputting the feature map into a classifier for output;
training the bilinear fine-grained artificial neural network based on the feature map output by the classifier to obtain a neural network model for detecting certificate defects, and detecting the certificate defect quality through the trained neural network model;
the bilinear fine-grained artificial neural network comprises two convolutional neural networks, wherein the two convolutional neural networks are respectively a first convolutional neural network and a second convolutional neural network, the first convolutional neural network is used for acquiring a low-layer characteristic diagram, and the second convolutional neural network is used for acquiring a detailed characteristic diagram;
the first convolutional neural network employs VGGNet, and the second convolutional neural network employs DenseNet.
2. The novel license quality detection method according to claim 1, characterized in that the face photograph data set is constructed based on a face image collected in a self-service license transaction all-in-one machine.
3. The method for detecting the quality of the license according to claim 2, characterized in that the face photograph data set is subjected to data set enhancement processing for increasing the number of data sets and improving the precision of a training model, and the data set enhancement processing method comprises rotating, mirroring, clipping and changing the contrast.
4. The novel license quality detection method according to claim 1, wherein in the DenseNet, the size of the feature map is matched between every two adjacent DenseBlock by using Batch +1 x 1Conv +2 x 2 Avgpool.
5. The novel license quality detection method according to claim 1, characterized in that the face photograph data sets are respectively input into the first convolutional neural network and the second convolutional neural network to respectively obtain a low-level feature map and a detail feature map, then the low-level feature map and the detail feature map are fused, the outer product of the low-level feature map and the detail feature map is calculated and then input into a full-link layer and a Softmax classifier, and fine-grained classification of face images is performed for identifying the unqualified reasons for shooting the licenses.
6. The novel certificate quality detection method according to claim 1, wherein all images in the face photograph data set are divided into a training set, a test set and a verification set, the training set is input into a bilinear fine-grained artificial neural network for certificate cheating detection based on a python3.6+ Tensorflow2.0 programming environment, a neural network model for certificate cheating detection is trained, test set verification is performed based on the test set and the verification set, and certificate quality detection is performed through the trained bilinear fine-grained artificial neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210621561.8A CN114998964B (en) | 2022-06-02 | 2022-06-02 | Novel license quality detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210621561.8A CN114998964B (en) | 2022-06-02 | 2022-06-02 | Novel license quality detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114998964A CN114998964A (en) | 2022-09-02 |
| CN114998964B true CN114998964B (en) | 2023-04-18 |
Family
ID=83030667
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210621561.8A Active CN114998964B (en) | 2022-06-02 | 2022-06-02 | Novel license quality detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114998964B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096535A (en) * | 2016-06-07 | 2016-11-09 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of face verification method based on bilinearity associating CNN |
| CN111191587A (en) * | 2019-12-30 | 2020-05-22 | 兰州交通大学 | Pedestrian re-identification method and system |
| CN112686242A (en) * | 2020-12-29 | 2021-04-20 | 昆明理工大学 | Fine-grained image classification method based on multilayer focusing attention network |
| KR20210086078A (en) * | 2019-12-31 | 2021-07-08 | 한국전력공사 | Method for learning artificial neural network for detecting power line from input image |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10916001B2 (en) * | 2016-11-28 | 2021-02-09 | Adobe Inc. | Facilitating sketch to painting transformations |
| CN110348283A (en) * | 2019-05-13 | 2019-10-18 | 常熟理工学院 | Fine granularity model recognizing method based on the fusion of bilinear model multilayer feature |
| CN112785486A (en) * | 2019-11-07 | 2021-05-11 | 英特尔公司 | Adaptive deformable kernel prediction network for image denoising |
| CN110674800B (en) * | 2019-12-04 | 2020-05-19 | 图谱未来(南京)人工智能研究院有限公司 | Face living body detection method and device, electronic equipment and storage medium |
| CN113158913A (en) * | 2021-04-25 | 2021-07-23 | 安徽科大擎天科技有限公司 | Face mask wearing identification method, system and terminal |
| CN113392875B (en) * | 2021-05-20 | 2023-03-24 | 广东工业大学 | Method, system and equipment for classifying fine granularity of image |
| CN114119551A (en) * | 2021-11-26 | 2022-03-01 | 南京烽火星空通信发展有限公司 | Quantitative analysis method for human face image quality |
-
2022
- 2022-06-02 CN CN202210621561.8A patent/CN114998964B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106096535A (en) * | 2016-06-07 | 2016-11-09 | 广东顺德中山大学卡内基梅隆大学国际联合研究院 | A kind of face verification method based on bilinearity associating CNN |
| CN111191587A (en) * | 2019-12-30 | 2020-05-22 | 兰州交通大学 | Pedestrian re-identification method and system |
| KR20210086078A (en) * | 2019-12-31 | 2021-07-08 | 한국전력공사 | Method for learning artificial neural network for detecting power line from input image |
| CN112686242A (en) * | 2020-12-29 | 2021-04-20 | 昆明理工大学 | Fine-grained image classification method based on multilayer focusing attention network |
Non-Patent Citations (2)
| Title |
|---|
| Abhimanyu Dubey 等.Pairwise Confusion for Fine-Grained Visual Classification.《CVF》.2018,1-17. * |
| 赵心怡.基于二代身份证的人脸验证系统关键技术研究.《中国优秀硕士学位论文全文数据库信息科技辑》.2019,I138-434. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114998964A (en) | 2022-09-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yuan et al. | Fingerprint liveness detection using an improved CNN with image scale equalization | |
| CN105005765B (en) | A kind of facial expression recognizing method based on Gabor wavelet and gray level co-occurrence matrixes | |
| WO2019214557A1 (en) | Method and system for detecting face image generated by deep network | |
| CN108985929B (en) | Training method, business data classification processing method and device, and electronic equipment | |
| CN106951856A (en) | Bag extracting method of expressing one's feelings and device | |
| CN106991390A (en) | A kind of hand-held testimony of a witness Compare System and method based on deep learning | |
| CN109165674A (en) | A kind of certificate photo classification method based on multi-tag depth convolutional network | |
| CN108090830B (en) | Credit risk rating method and device based on facial portrait | |
| CN105512624A (en) | Smile face recognition method and device for human face image | |
| CN108961675A (en) | Fall detection method based on convolutional neural networks | |
| CN106803069A (en) | Crowd's level of happiness recognition methods based on deep learning | |
| CN106599800A (en) | Face micro-expression recognition method based on deep learning | |
| CN109344759A (en) | A kind of relatives' recognition methods based on angle loss neural network | |
| CN110084149A (en) | A kind of face verification method based on difficult sample four-tuple dynamic boundary loss function | |
| CN106530200A (en) | Deep-learning-model-based steganography image detection method and system | |
| CN106056059B (en) | The face identification method of multi-direction SLGS feature description and performance cloud Weighted Fusion | |
| CN112528764B (en) | Facial expression recognition method, system and device and readable storage medium | |
| CN109214298A (en) | A kind of Asia women face value Rating Model method based on depth convolutional network | |
| CN108108760A (en) | A kind of fast human face recognition | |
| CN101916369A (en) | Face recognition method based on kernel nearest subspace | |
| CN106709418A (en) | Face identification method based on scene photo and identification photo and identification apparatus thereof | |
| CN111476727B (en) | Video motion enhancement method for face-changing video detection | |
| CN110516526A (en) | A kind of small sample target identification method based on Feature prototype metric learning | |
| CN114913923A (en) | Cell type identification method aiming at open sequencing data of single cell chromatin | |
| CN110276299A (en) | Batch meter appearance fault picture identification model and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |