CN110414350A - The face false-proof detection method of two-way convolutional neural networks based on attention model - Google Patents
The face false-proof detection method of two-way convolutional neural networks based on attention model Download PDFInfo
- Publication number
- CN110414350A CN110414350A CN201910560641.5A CN201910560641A CN110414350A CN 110414350 A CN110414350 A CN 110414350A CN 201910560641 A CN201910560641 A CN 201910560641A CN 110414350 A CN110414350 A CN 110414350A
- Authority
- CN
- China
- Prior art keywords
- face
- msr
- rgb
- feature vector
- false
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Evolutionary Computation (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Evolutionary Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Image Analysis (AREA)
- Collating Specific Patterns (AREA)
Abstract
The face false-proof detection method of the invention discloses a kind of two-way convolutional neural networks based on attention model, comprising: building training set, the training sample formed including RGB facial image and MSR facial image;Construct face anti-counterfeiting detection network, including the feature extraction unit for extracting RGB feature vector sum MSR feature vector, Fusion Features unit for being merged to RGB feature vector sum MSR feature vector, the tagsort unit for classifying to fusion feature vector;Face anti-counterfeiting detection network is trained using training set, obtains face anti-counterfeiting detection model;In application, be input in face anti-counterfeiting detection model after being pre-processed to RGB face picture to be measured, be computed output test result, i.e., true face or false face.This method can reduce influence and RGB picture abundant texture information of the illumination to detection using MSR picture simultaneously, keep testing result accurate and have certain generalization.
Description
Technical field
The invention belongs to biological identification false-proof fields, and in particular to a kind of two-way convolutional Neural net based on attention model
The face false-proof detection method of network.
Background technique
With the development of technology, the mankind use a variety of biological characteristics as the important documents of Verification System, such as fingerprint, people
Face, sound, pupil etc..No matter economically or social perspective, face are all one of most influential biological characteristics.
In addition, this technology is applied in numerous occasions due to the fast development of recognition of face and Face datection, it is big to secret
The access control system of occasion, the small login system to laptop, the even system for unlocking of mobile terminal and other biological are special
Sign is compared, and face authentication is increasingly becoming a kind of most-often used authentication mode.
Spoofing attack is a kind of attack for biological authentification system, it is by being presented legal biological characteristic to sensor
Version is forged, attempt authenticates biological authentification system for legitimate user illegal user, so that the illegal user be made to enter biology
Verification System.Since attacker can obtain the appearance feature of legitimate user from personal website or social network sites easily, very
The photos and videos that legitimate user can also closely be shot to attacker is compared with other biological feature, for face
Spoofing attack is easier to implement.
It is, in general, that face spoofing attack can be divided into 3 classes: photo attack, video attack and mask attack.Photo attack
Refer to that the photo of legitimate user is printed upon on paper or shows on the screen of the electronic device by attacker, is presented to biological identification system
A kind of attack of system sensor.Video attack is also referred to as Replay Attack, because this kind attack is by resetting legitimate user
Video is implemented.Mask attack then refers to that attacker puts on the 3D mask of legitimate user, and disguise oneself as legitimate user, attempts to enter
The attack of face identification system.
Since 21 century, research institution both domestic and external, which starts to detect face spoofing attack, has carried out a large amount of research,
Traditional method is broadly divided into three classes: the first kind is the method based on textural characteristics, mainly special using the texture manually extracted
Sign is to carry out face spoofing attack detection.Second class is the method based on motion information, mainly by sequence of frames of video to view
The blink of people, shakes the head in frequency, puts first-class motion information and is detected, so that judging whether is face spoofing attack.Third class is
Based on the method for rebuilding face three-dimensional information, sparse three-dimensional model mainly is reconstructed using face 2-dimentional photo, so that judgement is
True face or the attack of picture or video.
With the high speed development of modern processors computing capability and deep learning theory, the method for deep learning is had begun
It is applied in face spoofing attack detection.Feature is extracted relative to artificial, the method detection performance of neural network has greatly
It is promoted.Neural network may learn more judges face spoofing attack added with the feature of discrimination, but the side of deep learning
Method needs a large amount of training data to carry out model training, and powerful learning ability also allows the generalization of model to have certain reduction.
So the generalization of model is improved, to improve detection while how capable of farthest utilizing the learning ability of neural network
As a result accuracy.
Summary of the invention
The face anti-counterfeiting detection of the object of the present invention is to provide a kind of two-way convolutional neural networks based on attention model
Method, this method can reduce influence and RGB picture abundant texture information of the illumination to detection using MSR picture simultaneously, make
Testing result is accurate and has certain generalization.
For achieving the above object, the present invention the following technical schemes are provided:
A kind of face false-proof detection method of the two-way convolutional neural networks based on attention model, comprising the following steps:
Training set is constructed, the processing of illumination unification is carried out to RGB face picture using MSR algorithm, obtains MSR face figure
Piece, and face extraction is carried out to RGB face picture and MSR face picture, RGB facial image and MSR facial image are obtained, it should
RGB facial image and corresponding MSR facial image form a training sample, construct training set with this;
Face anti-counterfeiting detection network is constructed, including the feature extraction list for extracting RGB feature vector sum MSR feature vector
Member, the Fusion Features unit for being merged to RGB feature vector sum MSR feature vector, for fusion feature vector into
The tagsort unit of row classification;
Face anti-counterfeiting detection network is trained using training set, obtains face anti-counterfeiting detection model;
In application, being input in face anti-counterfeiting detection model, being computed after being pre-processed to RGB face picture to be measured
Output test result, i.e., true face or false face.
The face false-proof detection method has the beneficial effect that
The structure difference and most models of face anti-counterfeiting detection model in the present invention, using a kind of parallel two-way framework.
The characteristics of having merged the feature vector of RGB picture and the feature vector of MSR picture simultaneously, combined two kinds of face pictures, avoids
Influence of the illumination to testing result, while using the abundant details of RGB picture, improve detection effect.In addition, because to illumination into
It has gone unification, the generalization of model can be improved, model can be applied well under a variety of illumination.Finally, by this mould
Compared with the conventional method, accuracy rate has all well and good raising to the testing result of type.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to do simply to introduce, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art, can be with root under the premise of not making the creative labor
Other accompanying drawings are obtained according to these attached drawings.
Fig. 1 is the structural schematic diagram for the face anti-counterfeiting detection model that embodiment provides;
Fig. 2 is the structural schematic diagram of the feature extraction unit in Fig. 1;
Fig. 3 is the structural schematic diagram of Fusion Features unit in Fig. 1;
Fig. 4 is the structural schematic diagram of tagsort unit in Fig. 1;
Fig. 5 is the flow chart of the building that embodiment provides and training face anti-counterfeiting detection model;
Fig. 6 is the flow chart that face anti-counterfeiting detection is carried out using face anti-counterfeiting detection model that embodiment provides.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, with reference to the accompanying drawings and embodiments to this
Invention is described in further detail.It should be appreciated that the specific embodiments described herein are only used to explain the present invention,
And the scope of protection of the present invention is not limited.
In order to realize to face anti-counterfeiting detection, this example provides a kind of two-way convolutional neural networks based on attention model
Face false-proof detection method, specifically include face anti-counterfeiting detection model construction and using the face anti-counterfeiting detection model to face
Carry out two stages of anti-fake judgement.
Construct the face anti-counterfeiting detection model stage
Building face anti-counterfeiting detection model mainly includes the building of training set, the building of face anti-counterfeiting detection network and people
The training of face anti-counterfeiting detection network.As shown in figures 1-4, every partial content is gradually described below:
The training set of the building of training set, building includes two parts, and a part is RGB data library, and another part is pair
Image carries out the MSR database formed after MSR algorithm process.
CASIA-FASD is used in the present invention, the RGB face picture in REPLAY-ATTACK and OULU-NPU is as data
Source, firstly, the RGB face picture in data source forms RGB data library, then, using MSR algorithm to the RGB people in data source
Face picture carries out the processing of illumination unification, obtains MSR face picture, forms MSR database.
After obtaining RGB data library and MSR database, it is also necessary to complete RGB face picture and MSR face picture into
Row face extraction is aligned and is cut out to RGB face picture and MSR face picture using MTCNN algorithm that is, in face extraction
It cuts operation (144 × 144 sizes can be cropped to), obtains RGB facial image and MSR facial image, the RGB facial image and right
The MSR facial image answered forms a training sample, constructs training set with this.
MSR (Multi-Scale Retinex) algorithm is the algorithm for image enhancement based on Retinex theory.Irradiation is schemed
As assuming to be estimated as space smoothing image, original image is S (x, y), and reflected image is R (x, y), and luminance picture is L (x, y).
Retinex algorithm is to estimate the variation of illumination in image by calculating pixel and peripheral region under the action of weighted average,
Luminance picture is removed, only retains S (x, y) attribute, to achieve the effect that brightness unification.MSR algorithm is based on single scale
A kind of multi-Scale Retinex Algorithm for developing of Retinex algorithm, the advantages of absorbing single scale method, solve single ruler
The shortcomings that degree method, realizes better brightness unification effect.
The building of face anti-counterfeiting detection network, as shown in Figure 1, the face anti-counterfeiting detection network of building includes feature extraction list
Member, Fusion Features unit and tagsort unit.
Wherein, feature extraction unit is mainly used for extracting RGB feature vector to RGB facial image, mentions to MSR facial image
Take MSR feature vector.As shown in Figure 1, feature extraction unit includes two-way parallel organization, wherein be all the way RGB feature extractor,
Feature extraction is carried out to the RGB facial image of input, exports RGB feature vector, another way is MSR feature extractor, to input
MSR facial image carry out feature extraction, export MSR feature vector.RGB feature extractor and MSR feature extractor structure phase
Together, it is Mobile Net or ResNet18, can be obtained using the identical RGB feature extractor of structure with MSR feature extractor
The identical RGB feature vector sum MSR feature vector of intrinsic dimensionality.
Fusion Features unit is mainly used for merging the identical RGB feature vector sum MSR feature vector of dimension, such as schemes
Shown in 2, Fusion Features unit includes attention model, specifically inputs the identical RGB feature vector sum MSR feature vector of dimension
Into attention model, step-by-step is weighted fusion, obtains the constant fusion feature vector of single dimension.
Wherein, the RGB feature vector sum MSR feature vector of attention model obtained according to training set feature extraction carries out
Synchronous blending weight training.Weight vector is initialized first, and vector magnitude is (N, 2), and being worth is 0.5, and wherein N is feature vector
Dimension, in trained process, loss function backpropagation updates weighting parameter, when integrally training is completed, blending weight
It is optimal, to obtain optimal anti-fake model.
Tagsort unit is mainly used for classifying to fusion feature vector, output category result.As shown in Fig. 2, special
Levying taxon includes at least one layer of full connection stratum, is mainly used for carrying out fusion feature vector full attended operation, output one
The bivector that a adduction is 1, the corresponding serial number of biggish confidence level represent the prediction result of final output, and 0 represents true face,
1 represents false face, and the dimension of last full articulamentum is (N, 2), and wherein N is fused intrinsic dimensionality, and 2 indicate final output two
Class classification results (true face and false face).
The training of face anti-counterfeiting detection network
After constructing face anti-counterfeiting detection network, i.e., face anti-counterfeiting detection network is instructed using training sample
Practice, specifically, when training, training sample and corresponding label is input in face anti-counterfeiting detection network, the predicted value of output
Costing bio disturbance is carried out with label, the weight parameter of face anti-counterfeiting detection network is updated using backpropagation, at the end of training, really
Fixed network parameter and face anti-counterfeiting detection group of networks is at face anti-counterfeiting detection model.
In order to improve training effectiveness and prevent from training over-fitting, rotated using the human face region of random angles and random fixed
Size is cut, and human face region obtains the input size 144 × 144 by pretreatment, and it is anti-to be input to face as training sample
Training in puppet detection network.The loss function used when training is cross entropy loss function, and the optimizer of use is Adam optimization
Device, initial learning rate are set as 0.001.The mini-batch size that uses is 64 when training, i.e., is sent into 64 every time greatly
The small image for being 144 × 144 is trained, and after 300 batches of training, model parameter is saved.
Using face anti-counterfeiting detection model is carried out to face the anti-fake judgement stage
In application, RGB face picture is carried out as shown in figure 4, RGB face picture to be detected is pre-processed
After MSR algorithm process, MSR face picture is obtained, and complete RGB face picture and MSR face picture are aligned and are cut
Operation obtains RGB facial image and MSR facial image, the RGB facial image of acquisition and MSR facial image is input to training
In good face anti-counterfeiting detection model, the model parameter kept is loaded, propagates calculate forward, obtain RGB and MSR characteristics of image
Vector, enters back into and carries out Fusion Features in attention model, finally exports the prediction result of true and false face.
The result obtained by the model prediction of face anti-counterfeiting detection is compared with existing method, with relatively good accurate
Rate.
Technical solution of the present invention and beneficial effect is described in detail in above-described specific embodiment, Ying Li
Solution is not intended to restrict the invention the foregoing is merely presently most preferred embodiment of the invention, all in principle model of the invention
Interior done any modification, supplementary, and equivalent replacement etc. are enclosed, should all be included in the protection scope of the present invention.
Claims (8)
1. a kind of face false-proof detection method of the two-way convolutional neural networks based on attention model, comprising the following steps:
Training set is constructed, the processing of illumination unification is carried out to RGB face picture using MSR algorithm, obtains MSR face picture, and
Face extraction is carried out to RGB face picture and MSR face picture, obtains RGB facial image and MSR facial image, the RGB face
Image and corresponding MSR facial image form a training sample, construct training set with this;
Face anti-counterfeiting detection network is constructed, including the feature extraction unit for extracting RGB feature vector sum MSR feature vector,
Fusion Features unit for being merged to RGB feature vector sum MSR feature vector, for dividing fusion feature vector
The tagsort unit of class;
Face anti-counterfeiting detection network is trained using training set, obtains face anti-counterfeiting detection model;
In application, being input in face anti-counterfeiting detection model after being pre-processed to RGB face picture to be measured, being computed output
Testing result, i.e., true face or false face.
2. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in claim 1,
It is characterized in that, in face extraction, is aligned and is cut behaviour to RGB face picture and MSR face picture using MTCNN algorithm
Make, obtains RGB facial image and MSR facial image.
3. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in claim 1,
It is characterized in that, feature extraction unit includes two-way parallel organization, wherein being all the way RGB feature extractor, to the RGB face of input
Image carries out feature extraction, exports RGB feature vector, and another way is MSR feature extractor, to the MSR facial image of input into
Row feature extraction exports MSR feature vector.
4. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in right wants 3, special
Sign is that RGB feature extractor is identical as MSR feature extractor structure, is Mobile Net or ResNet18, utilizes structure phase
With RGB feature extractor and MSR feature extractor can obtain the identical RGB feature vector sum MSR feature of intrinsic dimensionality to
Amount.
5. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in right wants 1, special
Sign is that Fusion Features unit includes attention model, specifically that the identical RGB feature vector sum MSR feature vector of dimension is defeated
Enter into attention model, step-by-step is weighted fusion, obtains the constant fusion feature vector of single dimension.
6. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in right wants 1, special
Sign is that tagsort unit includes at least one layer of full connection stratum, is mainly used for carrying out fusion feature vector full connection behaviour
To make, exports the bivector that an adduction is 1, the corresponding serial number of biggish confidence level represents the prediction result of final output, and 0
True face is represented, 1 represents false face.
7. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in right wants 1, special
Sign is that when training, training sample and corresponding label are input in face anti-counterfeiting detection network, the predicted value and label of output
Costing bio disturbance is carried out, the weight parameter of face anti-counterfeiting detection network, at the end of training, determining net are updated using backpropagation
Network parameter and face anti-counterfeiting detection group of networks are at face anti-counterfeiting detection model.
8. the face false-proof detection method of the two-way convolutional neural networks based on attention model as described in right wants 1, special
Sign is that the loss function used when training is cross entropy loss function, and the optimizer of use is Adam optimizer, initial to learn
Rate is set as 0.001;The mini-batch size used when training is 144 × 144 for 64, i.e. 64 Zhang great little of each feeding
Image be trained, training 300 batches after, model parameter is saved.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910560641.5A CN110414350A (en) | 2019-06-26 | 2019-06-26 | The face false-proof detection method of two-way convolutional neural networks based on attention model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910560641.5A CN110414350A (en) | 2019-06-26 | 2019-06-26 | The face false-proof detection method of two-way convolutional neural networks based on attention model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110414350A true CN110414350A (en) | 2019-11-05 |
Family
ID=68359813
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910560641.5A Pending CN110414350A (en) | 2019-06-26 | 2019-06-26 | The face false-proof detection method of two-way convolutional neural networks based on attention model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110414350A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111144314A (en) * | 2019-12-27 | 2020-05-12 | 北京中科研究院 | Method for detecting tampered face video |
CN111191549A (en) * | 2019-12-23 | 2020-05-22 | 浙江大学 | Two-stage face anti-counterfeiting detection method |
CN111209855A (en) * | 2020-01-06 | 2020-05-29 | 电子科技大学 | Face image identification method based on two-channel dense convolution neural network with contour enhancement |
CN111368666A (en) * | 2020-02-25 | 2020-07-03 | 上海蠡图信息科技有限公司 | Living body detection method based on novel pooling and attention mechanism double-current network |
CN111639589A (en) * | 2020-05-28 | 2020-09-08 | 西北工业大学 | Video false face detection method based on counterstudy and similar color space |
CN111814574A (en) * | 2020-06-12 | 2020-10-23 | 浙江大学 | Face living body detection system, terminal and storage medium applying double-branch three-dimensional convolution model |
CN112734696A (en) * | 2020-12-24 | 2021-04-30 | 华南理工大学 | Face changing video tampering detection method and system based on multi-domain feature fusion |
CN112749686A (en) * | 2021-01-29 | 2021-05-04 | 腾讯科技(深圳)有限公司 | Image detection method, image detection device, computer equipment and storage medium |
CN112907530A (en) * | 2021-02-08 | 2021-06-04 | 南开大学 | Method and system for detecting disguised object based on grouped reverse attention |
CN113171082A (en) * | 2021-04-26 | 2021-07-27 | 中山大学孙逸仙纪念医院 | Method and device for evaluating injury of anterior cruciate ligament |
CN113705397A (en) * | 2021-08-16 | 2021-11-26 | 南京信息工程大学 | Face detection method based on dual-flow CNN structure fusion PRNU (vertical false positive) GAN (generic inverse) generation |
CN114202782A (en) * | 2020-09-02 | 2022-03-18 | 四川大学 | Face depth counterfeiting detection method based on traditional features and neural network |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2017100063A4 (en) * | 2016-12-01 | 2017-02-23 | Macau University Of Science And Technology | Omnidirectional Roughness Algorithm for Topographic Signature Analysis of Lunar Craters |
CN107194335A (en) * | 2017-05-12 | 2017-09-22 | 南京工程学院 | A kind of face identification method under complex illumination scene |
CN109858466A (en) * | 2019-03-01 | 2019-06-07 | 北京视甄智能科技有限公司 | A kind of face critical point detection method and device based on convolutional neural networks |
-
2019
- 2019-06-26 CN CN201910560641.5A patent/CN110414350A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
AU2017100063A4 (en) * | 2016-12-01 | 2017-02-23 | Macau University Of Science And Technology | Omnidirectional Roughness Algorithm for Topographic Signature Analysis of Lunar Craters |
CN107194335A (en) * | 2017-05-12 | 2017-09-22 | 南京工程学院 | A kind of face identification method under complex illumination scene |
CN109858466A (en) * | 2019-03-01 | 2019-06-07 | 北京视甄智能科技有限公司 | A kind of face critical point detection method and device based on convolutional neural networks |
Non-Patent Citations (1)
Title |
---|
HAONAN CHEN ET AL.: "Attention-Based Two-Stream Convolutional Networks for Face Spoofing Detection", 《IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111191549A (en) * | 2019-12-23 | 2020-05-22 | 浙江大学 | Two-stage face anti-counterfeiting detection method |
CN111144314A (en) * | 2019-12-27 | 2020-05-12 | 北京中科研究院 | Method for detecting tampered face video |
CN111209855A (en) * | 2020-01-06 | 2020-05-29 | 电子科技大学 | Face image identification method based on two-channel dense convolution neural network with contour enhancement |
CN111368666A (en) * | 2020-02-25 | 2020-07-03 | 上海蠡图信息科技有限公司 | Living body detection method based on novel pooling and attention mechanism double-current network |
CN111368666B (en) * | 2020-02-25 | 2023-08-18 | 上海蠡图信息科技有限公司 | Living body detection method based on novel pooling and attention mechanism double-flow network |
CN111639589A (en) * | 2020-05-28 | 2020-09-08 | 西北工业大学 | Video false face detection method based on counterstudy and similar color space |
CN111814574A (en) * | 2020-06-12 | 2020-10-23 | 浙江大学 | Face living body detection system, terminal and storage medium applying double-branch three-dimensional convolution model |
CN111814574B (en) * | 2020-06-12 | 2023-09-15 | 浙江大学 | Face living body detection system, terminal and storage medium applying double-branch three-dimensional convolution model |
CN114202782A (en) * | 2020-09-02 | 2022-03-18 | 四川大学 | Face depth counterfeiting detection method based on traditional features and neural network |
CN112734696B (en) * | 2020-12-24 | 2023-01-13 | 华南理工大学 | Face changing video tampering detection method and system based on multi-domain feature fusion |
CN112734696A (en) * | 2020-12-24 | 2021-04-30 | 华南理工大学 | Face changing video tampering detection method and system based on multi-domain feature fusion |
CN112749686A (en) * | 2021-01-29 | 2021-05-04 | 腾讯科技(深圳)有限公司 | Image detection method, image detection device, computer equipment and storage medium |
CN112907530A (en) * | 2021-02-08 | 2021-06-04 | 南开大学 | Method and system for detecting disguised object based on grouped reverse attention |
CN112907530B (en) * | 2021-02-08 | 2022-05-17 | 南开大学 | Method and system for detecting disguised object based on grouped reverse attention |
CN113171082A (en) * | 2021-04-26 | 2021-07-27 | 中山大学孙逸仙纪念医院 | Method and device for evaluating injury of anterior cruciate ligament |
CN113705397A (en) * | 2021-08-16 | 2021-11-26 | 南京信息工程大学 | Face detection method based on dual-flow CNN structure fusion PRNU (vertical false positive) GAN (generic inverse) generation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110414350A (en) | The face false-proof detection method of two-way convolutional neural networks based on attention model | |
Yuan et al. | Fingerprint liveness detection using an improved CNN with image scale equalization | |
Salama AbdELminaam et al. | A deep facial recognition system using computational intelligent algorithms | |
Yang et al. | Detecting fake images by identifying potential texture difference | |
Zhang et al. | Demeshnet: Blind face inpainting for deep meshface verification | |
CN110349136A (en) | A kind of tampered image detection method based on deep learning | |
CN111814620B (en) | Face image quality evaluation model establishment method, optimization method, medium and device | |
Chang et al. | Deepfake face image detection based on improved VGG convolutional neural network | |
CN101558431B (en) | Face authentication device | |
CN105956572A (en) | In vivo face detection method based on convolutional neural network | |
Ma et al. | A novel face presentation attack detection scheme based on multi-regional convolutional neural networks | |
CN105335719A (en) | Living body detection method and device | |
CN113537027B (en) | Face depth counterfeiting detection method and system based on face division | |
CN109344709A (en) | A kind of face generates the detection method of forgery image | |
CN112651333B (en) | Silence living body detection method, silence living body detection device, terminal equipment and storage medium | |
CN107609515A (en) | A kind of face alignment system and method for the double verification based on platform of soaring | |
Du | High-precision portrait classification based on mtcnn and its application on similarity judgement | |
Galbally et al. | Study on face identification technology for its implementation in the Schengen information system | |
Sedik et al. | An efficient cybersecurity framework for facial video forensics detection based on multimodal deep learning | |
CN111191549A (en) | Two-stage face anti-counterfeiting detection method | |
Yang et al. | Document Image Forgery Detection Based on Deep Learning Models | |
Fan et al. | Research on liveness detection algorithms based on deep learning | |
Dong | 3D face recognition neural network for digital human resource management | |
Kaur et al. | Improved Facial Biometric Authentication Using MobileNetV2 | |
Li et al. | [Retracted] Intelligent Face Recognition System Based on Universal Design Concept |
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: 20191105 |