CN112163493A - Video false face detection method and electronic device - Google Patents
Video false face detection method and electronic device Download PDFInfo
- Publication number
- CN112163493A CN112163493A CN202010995945.7A CN202010995945A CN112163493A CN 112163493 A CN112163493 A CN 112163493A CN 202010995945 A CN202010995945 A CN 202010995945A CN 112163493 A CN112163493 A CN 112163493A
- Authority
- CN
- China
- Prior art keywords
- video
- dimensional residual
- face
- neural network
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- 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
- G06V40/40—Spoof detection, e.g. liveness detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Life Sciences & Earth Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a video false face detection method and an electronic device, comprising the following steps: carrying out face positioning on a video to be detected to obtain a face sequence; preprocessing a face sequence to obtain a video sampling frame sequence with specified size and length; inputting a video sampling frame sequence into a trained three-dimensional residual learning convolutional neural network, and judging whether the face in the video to be detected is a false face; the three-dimensional residual learning convolutional neural network comprises one or more convolutional layers and corresponding maximum pooling layers, a plurality of three-dimensional residual learning layers consisting of one or more three-dimensional residual learning modules, an average pooling layer and an output layer; the three-dimensional residual error learning module comprises a first branch circuit and a second branch circuit which are respectively connected with the input of the three-dimensional residual error learning module, and an operation layer for adding the output results of the two first branch circuits and the second branch circuit. The method and the device perform residual error learning on the video characteristics so as to solve the problem of model degradation possibly caused by too deep network and improve the performance of the model.
Description
Technical Field
The invention belongs to the field of internet, and particularly relates to a video false face detection method and an electronic device.
Background
In recent years, in the fields of computer vision, multimedia security and the like, with the help of a counterfeiting system of a deep learning technology, media data contents such as images, videos and the like are randomly operated and modified, and a deep counterfeiting (Deepfake) technology for replacing faces and tampering the attributes of the faces is continuously developed and evolved, so that great security risks are brought to individuals, the society and the country, and further the trust loss in the digital field is caused. Meanwhile, compared with a deep counterfeiting technology which develops and evolves rapidly by means of deep learning, the counterfeiting detection technology which is counterbalanced by the deep counterfeiting detection technology does not obtain a satisfactory result compared with research, and the existing detection means are not satisfactory in the aspects of accuracy, confidence coefficient and the like, which is caused by natural lag of defense in an attack and defense system on one hand, and on the other hand, the difficulty faced by the field of counterfeiting detection research is reflected greatly.
At present, detection research on false face videos at home and abroad can be divided into two types, wherein the first type is detection based on hidden features of generated image videos, and the second type is detection based on features extracted by a neural network.
The detection based on the hidden features of the generated image video is mainly to perform 'attack' on the defects or the features of the existing false face image video generation mode, for example, aiming at the fact that some methods for generating the false face image video leave 'artifacts' which should not exist near the face of the generated result, a specific network is trained to detect the 'artifacts', for example, aiming at the fact that the blinking frequency of people in the false face video generated by some generation methods is not consistent with the natural frequency, the blinking frame is judged based on a long-time memory network so as to be detected by the blinking frequency, and the recently published method for detecting the hidden features that the generated face has 'boundaries' in most face false replacement methods. The detection method is based on the defects and characteristics of the generated result of the existing generation method, can achieve more than 90% of detection accuracy, but has the problems of low mobility, unknown effect of coping with the unknown generation method and the like.
The detection method based on the neural network for feature extraction is a detection method with higher universality, and can be divided into a detection method based on two-dimensional image features and a detection method based on three-dimensional time sequence features according to different dimensions of extracted features, wherein the detection method reduces the video detection problem into the problem of video sampling frame detection when the detection method is applied to false face video detection. The two-dimensional feature extraction detection method based on image dimensionality mainly trains an image two-classification model through two-dimensional Convolutional Neural Networks (CNN) with various different structures, and classifies whether sampled video frames are false or not, and some work introduces an additional module on the basis, for example, the judgment accuracy and robustness of a specific counterfeiting mode are increased by introducing a counterstudy process; the video dimension-based time sequence feature extraction detection method includes the steps of directly training a classifier on a video data set by using a network related to extraction of time sequence features, and then detecting a video to be detected by using the trained classifier, for example, directly extracting three-dimensional time sequence features of the whole input video by using a Recurrent Neural Network (RNN) and a long-term and short-term memory network (LSTM) to train the classifier. Compared with a detection method based on hidden features, the detection method has better universality and mobility, the current related work can also reach 90% or even higher detection accuracy, but the detection method also has the problems of sensitivity to training data and the like.
The defects of the prior art are mainly as follows: 1. underutilization of the model to the time sequence characteristics of the video; 2. the network model is easy to have network degradation problem when the network layer number is too deep.
Disclosure of Invention
In order to solve the problems, the invention provides a video false face detection method and an electronic device, which are based on a three-dimensional residual learning convolutional neural network and solve the technical problems that the time sequence characteristics are not sufficiently utilized and the model is easy to have network degradation when false face video detection is carried out.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a video false face detection method comprises the following steps:
1) carrying out face positioning on a video to be detected to obtain a face sequence;
2) preprocessing a face sequence to obtain a video sampling frame sequence with specified size and length;
3) inputting a video sampling frame sequence into a trained three-dimensional residual learning convolutional neural network, and judging whether the face in the video to be detected is a false face;
the three-dimensional residual learning convolutional neural network comprises one or more convolutional layers and corresponding maximum pooling layers, a plurality of three-dimensional residual learning layers consisting of one or more three-dimensional residual learning modules, an average pooling layer and an output layer;
the three-dimensional residual error learning module comprises a first branch circuit and a second branch circuit which are respectively connected with the input of the three-dimensional residual error learning module, and an operation layer for adding the output results of the two first branch circuits and the second branch circuit;
the first branch comprises a convolution layer of a convolution kernel, a convolution layer of b convolution kernel and a convolution layer of convolution kernel of a convolution kernel; the second branch circuit comprises a direct connection branch circuit which takes the input of the second branch circuit as the output of the second branch circuit, and a and b are self-defined parameters.
Further, the pre-processing comprises: frame sampling and cropping.
Further, the method of frame sampling includes an equidistant sampling method.
Further, the trained three-dimensional residual error learning convolutional neural network is obtained through the following steps:
1) collecting a plurality of sample videos, and carrying out face positioning on each sample video to obtain a plurality of sample face sequences;
2) preprocessing a sample face sequence to obtain a plurality of sample video sampling frame sequences with specified sizes and lengths;
3) and (3) iteratively inputting the sample video sampling frame sequences into the three-dimensional residual error learning convolutional neural network, and training the three-dimensional residual error learning convolutional neural network to obtain the trained three-dimensional residual error learning convolutional neural network.
Further, the sequence of sample video sample frames includes a set of frames of a specified length.
Further, the loss value obtained through calculation is subjected to back propagation through a cross entropy loss function, the neuron weight of the three-dimensional residual error learning convolution neural network is updated, and the three-dimensional residual error learning convolution neural network is trained.
Further, the optimizer used to train the three-dimensional residual learning convolutional neural network comprises an Adam optimizer.
Further, the activation functions used by the convolutional layer and the max pooling layer include a ReLU function.
A storage medium having a computer program stored therein, wherein the computer program is arranged to perform the above-mentioned method when executed.
An electronic device comprising a memory having a computer program stored therein and a processor arranged to run the computer to perform the method as described above.
Compared with the prior art, the invention has the following positive effects:
1. the problem of false face video detection is solved by using a three-dimensional convolutional neural network, the space-time characteristics of the video are fused, and the detection performance is further improved.
2. A three-dimensional residual error learning module is innovatively provided for residual error learning of video features so as to solve the problem of model degradation possibly caused by too deep network and further improve the performance of the model.
Drawings
Fig. 1 is a system framework of the three-dimensional residual learning convolutional neural network for false face video detection.
Fig. 2 is a schematic structural diagram of a three-dimensional residual error learning module according to the present invention.
Fig. 3 is a schematic structural diagram of a three-dimensional residual learning convolutional neural network according to the present invention.
Fig. 4 is an exemplary application scenario of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, 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.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The system flow diagram of the invention is shown in fig. 1, and the method of the invention supports the designed three-dimensional residual error learning convolutional neural network.
Experience tells that the number of network layers has a crucial influence on performance, and deeper networks can often obtain better performance, but experiments show that when the network depth is increased to a certain degree, a network degradation phenomenon may occur, namely the accuracy of the network reaches a saturation state, and is further deepened or even reduced.
Aiming at the phenomena, the invention constructs a three-dimensional convolution neural network model introducing a three-dimensional residual error learning module, and solves the problem of network degradation by introducing the designed three-dimensional residual error learning module on the existing deep three-dimensional convolution neural network. As shown in fig. 2, the inputs to the module are routed through two branches: sequentially through convolutional layers of convolution kernels of size 1 x 1, 3 x 3, 1 x 1, and direct output of the original input. The two are processed by the following addition operation to obtain the final output of the module, wherein the activation functions are the ReLU activation functions (origin: He K, Zhang X, Ren S, et al. deep reactive learning [ M ] [ S.l.: s.n ],2016: 770-. The former branch is sequentially subjected to convolution kernel convolution layers with different sizes, so that the purpose of further feature extraction on input is realized, the original input is directly subjected to addition operation with the features obtained in the former step, and the purpose of obtaining final output is realized, so that the final output result of the network is the 'residual error' between the expected output and the input, and not only the features extracted from the input are realized; the degradation problem caused by the fact that the network is too deep is hoped to be solved by learning the residual error, and the network performance is further improved; the aim of the model is to output a detection result of whether a face video to be detected is generated by counterfeiting, namely, the problem of video classification is solved.
The three-dimensional residual error learning module and the whole network structure after the module is introduced are shown in fig. 3. After the video data to be detected is input, the video data firstly passes through a convolution layer taking 7 × 7 as a convolution kernel, then passes through a maximum pooling layer of 3 × 3, and then is connected with four layers of the introduced three-dimensional residual error learning modules, the number of the introduced modules and the number of output channels of each layer are shown, then the video data passes through an average pooling layer with the size of 1 × 2, and finally the final detection result probability is obtained through a softmax output layer. The introduced three-dimensional residual error learning module, the common feature extraction convolution layer and the activation function of the pooling layer are Relu functions.
The model has the advantages that: by introducing the residual error learning module fusing the space-time characteristics, the space-time information of the video is considered, the problem of model degradation when the network is too deep is effectively solved, and the detection performance of the model is further improved.
By adopting the technical scheme of the invention, the false face video detection is realized, and the following problems are solved. First, the timing feature utilization problem: on the task of false face video detection, besides low-dimensional image features, the model performance can be improved to a certain extent by detecting the time sequence features of the video; second, degradation problems when the network hierarchy is too deep: by introducing the three-dimensional residual error learning module, the problem of model degradation when the network is too deep is effectively solved.
The technical scheme of the invention provides a method and a device for false face detection based on a three-dimensional residual error learning convolutional neural network on the basis of video data. According to the three-dimensional residual learning convolutional neural network, one embodiment of the invention provides a training method of a false face video detection model, which mainly comprises the following steps:
1) and carrying out face positioning, frame sampling and cutting on the videos in the training set and the test set to obtain a video sampling frame sequence with specified size and length as model input. In the experiment, the clipping result is an image frame with the size of 80 × 3 including the human face, the sequence length is set to be two groups of 20 frames and 40 frames, and the frame sampling strategy is equidistant sampling.
2) Initializing model parameters, including setting initial learning rate, epoch, and batch size, etc. In the experiment, the learning rate is 1e-5, the number of iteration rounds is set to be 100, and the batch size is 8.
3) Inputting the training set video sampling frame sequence obtained in the step 1 into the initialized three-dimensional residual error learning convolution neural network for iterative training, wherein an optimizer used by the model is an Adam optimizer, the initial learning rate is 1e-5, the weight attenuation is 1e-6, the loss function uses a cross entropy loss function, the loss value is calculated in the iterative training, and the loss value is used for back propagation to update the network neuron weight so as to obtain the optimal network.
4) After the network training is finished, the video sampling frame sequence in the test set obtained in the step 1 can be input into the network to obtain the detection result.
It should be noted that, for simplicity of description, the above-mentioned method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required by the invention.
Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.
The utilization of the present invention in a particular scenario is described below in conjunction with fig. 4.
At present, the uploading events of false face videos on various social media websites are infinite, and deep learning means is urgently needed for effective discrimination and detection. The practical application of the invention in the video uploading detection of the social media website is shown in fig. 4, the detector after the training is finished is arranged at the uploading file inspection rear end of the social media website, after a user uploads a certain section of character video to the social media website, the website transmits the video to the corresponding inspection rear end to be used as the input of the model, and obtains the detection result of the model, and the result is used as the criterion whether the section of video file uploaded by the user passes the audit or not: if no abnormity occurs in the detection, the video of the user can be uploaded to the social network site normally through the audit; and if the video is detected to be a fake face video, the user behavior cannot be recorded through auditing.
In the above technical solutions of the present invention, portions not described in detail can be implemented by using the prior art.
The method has impressive performance on the task of false face video Detection, the accuracy of a model obtained by training on a DFDC data set (dropout Detection Change) can reach 92.65%, the model score value in a DFDC match is up to 0.421, and the world rank is 10%.
Claims (10)
1. A video false face detection method comprises the following steps:
1) carrying out face positioning on a video to be detected to obtain a face sequence;
2) preprocessing a face sequence to obtain a video sampling frame sequence with specified size and length;
3) inputting a video sampling frame sequence into a trained three-dimensional residual learning convolutional neural network, and judging whether the face in the video to be detected is a false face;
the three-dimensional residual learning convolutional neural network comprises one or more convolutional layers and corresponding maximum pooling layers, a plurality of three-dimensional residual learning layers consisting of one or more three-dimensional residual learning modules, an average pooling layer and an output layer;
the three-dimensional residual error learning module comprises a first branch circuit and a second branch circuit which are respectively connected with the input of the three-dimensional residual error learning module, and an operation layer for adding the output results of the two first branch circuits and the second branch circuit;
the first branch comprises a convolution layer of a convolution kernel, a convolution layer of b convolution kernel and a convolution layer of convolution kernel of a convolution kernel; the second branch circuit comprises a direct connection branch circuit which takes the input of the second branch circuit as the output of the second branch circuit, and a and b are self-defined parameters.
2. The method of claim 1, wherein the pre-processing comprises: frame sampling and cropping.
3. The method of claim 1, wherein the method of frame sampling comprises an equidistant sampling method.
4. The method of claim 1, wherein the trained three-dimensional residual learning convolutional neural network is obtained by:
1) collecting a plurality of sample videos, and carrying out face positioning on each sample video to obtain a plurality of sample face sequences;
2) preprocessing a sample face sequence to obtain a plurality of sample video sampling frame sequences with specified sizes and lengths;
3) and (3) iteratively inputting the sample video sampling frame sequences into the three-dimensional residual error learning convolutional neural network, and training the three-dimensional residual error learning convolutional neural network to obtain the trained three-dimensional residual error learning convolutional neural network.
5. The method of claim 4, wherein the sequence of sample video sample frames comprises a set of frames of a specified length.
6. The method of claim 4, wherein the three-dimensional residual learning convolutional neural network is trained by backpropagating the calculated loss values through a cross entropy loss function, updating neuron weights of the three-dimensional residual learning convolutional neural network.
7. The method of claim 4, wherein the optimizer used to train the three-dimensional residual learning convolutional neural network comprises an Adam optimizer.
8. The method of claim 1, wherein the activation function used by the convolutional layer and the max pooling layer comprises a ReLU function.
9. A storage medium having a computer program stored thereon, wherein the computer program is arranged to, when run, perform the method of any of claims 1-8.
10. An electronic device comprising a memory having a computer program stored therein and a processor arranged to run the computer program to perform the method according to any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010995945.7A CN112163493A (en) | 2020-09-21 | 2020-09-21 | Video false face detection method and electronic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010995945.7A CN112163493A (en) | 2020-09-21 | 2020-09-21 | Video false face detection method and electronic device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112163493A true CN112163493A (en) | 2021-01-01 |
Family
ID=73864279
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010995945.7A Pending CN112163493A (en) | 2020-09-21 | 2020-09-21 | Video false face detection method and electronic device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112163493A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112784755A (en) * | 2021-01-25 | 2021-05-11 | 瑞芯微电子股份有限公司 | Light face tracking method and storage device |
| CN113723196A (en) * | 2021-08-02 | 2021-11-30 | 中国科学院信息工程研究所 | Video false face detection method and device based on prediction learning |
| CN113989713A (en) * | 2021-10-28 | 2022-01-28 | 杭州中科睿鉴科技有限公司 | Depth forgery detection method based on video frame sequence prediction |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107247949A (en) * | 2017-08-02 | 2017-10-13 | 北京智慧眼科技股份有限公司 | Face identification method, device and electronic equipment based on deep learning |
| CN108288271A (en) * | 2018-02-06 | 2018-07-17 | 上海交通大学 | Image detecting system and method based on three-dimensional residual error network |
| CN109376704A (en) * | 2018-11-30 | 2019-02-22 | 高新兴科技集团股份有限公司 | A kind of human face in-vivo detection method |
| CN109583342A (en) * | 2018-11-21 | 2019-04-05 | 重庆邮电大学 | Human face in-vivo detection method based on transfer learning |
| CN110909693A (en) * | 2019-11-27 | 2020-03-24 | 深圳市华付信息技术有限公司 | 3D face living body detection method and device, computer equipment and storage medium |
| CN110991366A (en) * | 2019-12-09 | 2020-04-10 | 武汉科技大学 | Shipping monitoring event identification method and system based on three-dimensional residual error network |
| CN111382677A (en) * | 2020-02-25 | 2020-07-07 | 华南理工大学 | Human behavior identification method and system based on 3D attention residual error model |
| CN111444881A (en) * | 2020-04-13 | 2020-07-24 | 中国人民解放军国防科技大学 | Fake face video detection method and device |
-
2020
- 2020-09-21 CN CN202010995945.7A patent/CN112163493A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107247949A (en) * | 2017-08-02 | 2017-10-13 | 北京智慧眼科技股份有限公司 | Face identification method, device and electronic equipment based on deep learning |
| CN108288271A (en) * | 2018-02-06 | 2018-07-17 | 上海交通大学 | Image detecting system and method based on three-dimensional residual error network |
| CN109583342A (en) * | 2018-11-21 | 2019-04-05 | 重庆邮电大学 | Human face in-vivo detection method based on transfer learning |
| CN109376704A (en) * | 2018-11-30 | 2019-02-22 | 高新兴科技集团股份有限公司 | A kind of human face in-vivo detection method |
| CN110909693A (en) * | 2019-11-27 | 2020-03-24 | 深圳市华付信息技术有限公司 | 3D face living body detection method and device, computer equipment and storage medium |
| CN110991366A (en) * | 2019-12-09 | 2020-04-10 | 武汉科技大学 | Shipping monitoring event identification method and system based on three-dimensional residual error network |
| CN111382677A (en) * | 2020-02-25 | 2020-07-07 | 华南理工大学 | Human behavior identification method and system based on 3D attention residual error model |
| CN111444881A (en) * | 2020-04-13 | 2020-07-24 | 中国人民解放军国防科技大学 | Fake face video detection method and device |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112784755A (en) * | 2021-01-25 | 2021-05-11 | 瑞芯微电子股份有限公司 | Light face tracking method and storage device |
| CN113723196A (en) * | 2021-08-02 | 2021-11-30 | 中国科学院信息工程研究所 | Video false face detection method and device based on prediction learning |
| CN113723196B (en) * | 2021-08-02 | 2024-05-28 | 中国科学院信息工程研究所 | Video virtual dummy face detection method and device based on predictive learning |
| CN113989713A (en) * | 2021-10-28 | 2022-01-28 | 杭州中科睿鉴科技有限公司 | Depth forgery detection method based on video frame sequence prediction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112398779B (en) | Network traffic data analysis method and system | |
| CN111461089B (en) | Face detection method, and training method and device of face detection model | |
| Nguyen et al. | Automatic image filtering on social networks using deep learning and perceptual hashing during crises | |
| CN112163488B (en) | Video virtual dummy face detection method and electronic device | |
| CN110309304A (en) | A kind of file classification method, device, equipment and storage medium | |
| CN110351244A (en) | A kind of network inbreak detection method and system based on multireel product neural network fusion | |
| CN109005145A (en) | A kind of malice URL detection system and its method extracted based on automated characterization | |
| CN111738054B (en) | Behavior anomaly detection method based on space-time self-encoder network and space-time CNN | |
| CN106295501A (en) | The degree of depth based on lip movement study personal identification method | |
| CN108282460B (en) | Evidence chain generation method and device for network security event | |
| CN113094707B (en) | Lateral movement attack detection method and system based on heterogeneous graph network | |
| CN111241338B (en) | Depth feature fusion video copy detection method based on attention mechanism | |
| CN112163493A (en) | Video false face detection method and electronic device | |
| CN114422211B (en) | HTTP malicious traffic detection method and device based on graph attention network | |
| CN111177469A (en) | Face retrieval method and face retrieval device | |
| CN114330966A (en) | Risk prediction method, device, equipment and readable storage medium | |
| Li et al. | PhishBox: An approach for phishing validation and detection | |
| CN104915399A (en) | Recommended data processing method based on news headline and recommended data processing method system based on news headline | |
| Yuan et al. | A novel approach for malicious URL detection based on the joint model | |
| CN113628059A (en) | Associated user identification method and device based on multilayer graph attention network | |
| CN110958244A (en) | Method and device for detecting counterfeit domain name based on deep learning | |
| CN112163494A (en) | Video false face detection method and electronic device | |
| CN114422271A (en) | Data processing method, device, equipment and readable storage medium | |
| CN116723005A (en) | Method and system for tracking malicious code implicit information under polymorphic hiding | |
| Gong et al. | AGRAMPLIFIER: Defending federated learning against poisoning attacks through local update amplification |
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 |