CN113158773A - Training method and training device for living body detection model - Google Patents

Training method and training device for living body detection model Download PDF

Info

Publication number
CN113158773A
CN113158773A CN202110245292.5A CN202110245292A CN113158773A CN 113158773 A CN113158773 A CN 113158773A CN 202110245292 A CN202110245292 A CN 202110245292A CN 113158773 A CN113158773 A CN 113158773A
Authority
CN
China
Prior art keywords
image
network
living body
detected
face
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.)
Granted
Application number
CN202110245292.5A
Other languages
Chinese (zh)
Other versions
CN113158773B (en
Inventor
张兴
谢思敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
TP Link Technologies Co Ltd
Original Assignee
TP Link Technologies Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by TP Link Technologies Co Ltd filed Critical TP Link Technologies Co Ltd
Priority to CN202110245292.5A priority Critical patent/CN113158773B/en
Publication of CN113158773A publication Critical patent/CN113158773A/en
Application granted granted Critical
Publication of CN113158773B publication Critical patent/CN113158773B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06V40/40—Spoof detection, e.g. liveness detection
    • G06V40/45—Detection of the body part being alive
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic 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
    • 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
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Biology (AREA)
  • Human Computer Interaction (AREA)
  • Multimedia (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Image Analysis (AREA)

Abstract

The application is suitable for the technical field of image processing, and provides a training method and a training device of a living body detection model, wherein the training method comprises the following steps: acquiring an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; inputting the sample data set into the first branch network and the second branch network, calculating a joint loss function, and training the second convolutional network according to the joint loss function to obtain a trained second convolutional network; and taking the first convolution network and the trained second convolution network as a target living body detection model. In the process, the third convolutional network in the second branch network is only used for assisting in training the second convolutional network, so that the excellent detection precision of the target living body detection model is ensured, unnecessary calculation amount is reduced, and the living body detection efficiency is improved.

Description

Training method and training device for living body detection model
Technical Field
The application belongs to the technical field of image processing, and particularly relates to a training method and a training device for a living body detection model.
Background
Liveness detection is a method of determining the true physiological characteristics of a subject in some authentication scenarios. Common attack means such as photos, face changing, masks, sheltering and screen copying can be effectively resisted, so that a user is helped to discriminate fraudulent behaviors, and the benefit of the user is guaranteed.
The existing in-vivo detection technology is often matched with various models to realize in-vivo detection. However, the multiple models require more computing resources, resulting in less efficient in vivo testing. This is a technical problem which needs to be solved urgently.
Disclosure of Invention
In view of this, embodiments of the present application provide a training method for a living body detection model, a living body detection method, a training apparatus for a living body detection model, a living body detection apparatus, a terminal device, a living body detection device, and a computer readable storage medium, which can solve the technical problem that the efficiency of living body detection is low due to the fact that many computing resources are needed by various models.
A first aspect of an embodiment of the present application provides a training method for a living body detection model, where the training method includes:
acquiring an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network form a second branch network;
inputting the sample data set into the first branch network and the second branch network, calculating a joint loss function, and training the second convolutional network according to the joint loss function to obtain a trained second convolutional network;
and taking the first convolution network and the trained second convolution network as a target living body detection model.
A second aspect of an embodiment of the present application provides a method of in vivo detection, the method including:
acquiring an image to be detected;
inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used for jointly training the second convolutional network;
and determining whether the image to be detected is a living body face image or not according to the detection result.
A third aspect of an embodiment of the present application provides a training apparatus for a living body detection model, the training apparatus including:
the first acquisition unit is used for acquiring an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network form a second branch network;
a training unit, configured to input the sample data set into the first branch network and the second branch network, calculate a joint loss function, train the second convolutional network according to the joint loss function, and obtain a trained second convolutional network;
and the cutting unit is used for taking the first convolution network and the trained second convolution network as a target living body detection model.
A fourth aspect of an embodiment of the present application provides an apparatus for in-vivo detection, the apparatus including:
the second acquisition unit is used for acquiring an image to be detected;
the detection unit is used for inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used for jointly training the second convolutional network;
and the judging unit is used for determining whether the image to be detected is a living body face image or not according to the detection result.
A fifth aspect of embodiments of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the method according to the first aspect when executing the computer program.
A sixth aspect of the embodiments of the present application provides a living body detection apparatus, including a camera module, a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps of the method of the second aspect when executing the computer program.
A fourth aspect of embodiments of the present application provides a computer-readable storage medium, which stores a computer program that, when executed by a processor, implements the steps of the method of the first or second aspect.
Compared with the prior art, the embodiment of the application has the advantages that: the second convolutional network in the first branch network is jointly trained by the first branch network in the initial model and the second branch network. And taking the first convolution network and the trained second convolution network as a target living body detection model. In the process, the third convolutional network in the second branch network is only used for assisting in training the second convolutional network, so that the excellent detection precision of the target living body detection model is ensured, unnecessary calculation amount is reduced, and the living body detection efficiency is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or the related technical descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 shows a schematic flow diagram of a training method of a living body detection model provided herein;
FIG. 2 shows a detailed schematic flow chart of step 101 of a training method for a living body detection model provided in the present application;
FIG. 3 shows a specific schematic flow chart of step A2 in the training method of the living body detection model provided in the present application;
FIG. 4 shows a specific schematic flow chart of step A3 in the training method of the living body detection model provided in the present application;
FIG. 5 shows a specific schematic flow chart of step 102 of a method for training a living body detection model provided herein;
FIG. 6 shows a schematic diagram of an initial model provided herein;
FIG. 7 shows a schematic flow chart of a method of in vivo detection provided herein;
FIG. 8 is a flowchart illustrating steps 701 of a method for detecting a living body according to the present application
FIG. 9 shows a specific schematic flowchart of step 703 in a method for detecting a living body provided by the present application;
FIG. 10 shows a detailed schematic flow chart of another method of in vivo testing provided herein;
FIG. 11 is a schematic diagram of a training apparatus for a living body detection model provided herein;
FIG. 12 is a schematic view of an in vivo testing device provided herein;
fig. 13 is a schematic diagram of a terminal device according to an embodiment of the present invention;
FIG. 14 is a schematic view of a living body detecting apparatus according to an embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
In order to better understand the technical problems addressed by the present application, the above background will be further explained herein:
the existing in-vivo detection technology is often matched with various models to realize in-vivo detection. For example: and inputting the image to be recognized into a depth estimation model, and performing depth estimation through the depth estimation model to obtain a depth image. And identifying the depth image and the image to be identified based on the living body detection model to obtain a living body detection result.
Most living body detection equipment is limited in hardware condition, so that the time is consumed and the efficiency is low when various models are calculated, and real-time detection cannot be performed.
In view of the above, embodiments of the present application provide a training method for a living body detection model, a living body detection method, a training apparatus for a living body detection model, a living body detection apparatus, a terminal device, a living body detection device, and a computer readable storage medium, which can solve the above technical problems.
Wherein, the terminal device and the living body detecting device may be the same device or different devices.
Referring to fig. 1, fig. 1 shows a schematic flow chart of a training method of an activity detection model provided in the present application. As shown in fig. 1, the training method is applied to a terminal device, and includes the following steps:
step 101, obtaining an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network constitute a second branch network.
It is noted that the initial model and the sample data set are different according to different application scenarios. The present embodiment does not limit the application scenarios from step 101 to step 103. In order to better explain the technical solution of the present embodiment, the present embodiment takes an application scenario as a biopsy scenario as an example, and the technical solution of the present embodiment is explained. Those skilled in the art can obtain technical solutions of other application scenarios by analogy with the technical solution of the living body detection scenario, which is not described herein.
Since different network structures in the initial model bear different processing tasks, the present application divides the hierarchical structure in the initial model into a first convolutional network, a second convolutional network, and a third convolutional network. In a living body detection scene, a first convolution network is used for feature extraction, a second convolution network is used for reconstructing a Fourier spectrogram, and a third convolution network is used for detection classification. In order to better explain the technical solution of the present embodiment, the first convolution network, the second convolution network, and the third convolution network are taken as examples.
Taking a living body detection scene as an example, the sample data set includes a sample image, an original fourier spectrogram corresponding to the sample image, and annotation information corresponding to the sample image. The annotation information can be manually added information or model identification information, and is used for annotating the image type of the sample image, wherein the image type includes but is not limited to a living body face image and a non-living body face image. It will be appreciated that as the initial model has more branching networks, there is a concomitant increase in data types in the sample data set.
Wherein, for the sample image, the terminal device can obtain the sample image in the existing database. The terminal equipment can also acquire sample images through the living body detection equipment (preferably, because different cameras and application scenes have differences, the existing database has a certain difference with the images to be identified in the actual application scene, and the detection performance of the trained model is reduced in the actual application.
The sample data set may be raw data (raw data includes, but is not limited to, a first raw image and an initial fourier spectrum, etc.) or raw data after being preprocessed. In order to optimize the amount of calculation in the training process, the present application preprocesses the original data to obtain a sample data set, and the specific process is as shown in the following optional embodiments:
as an alternative embodiment of the present application, step 101 includes the following steps a1 to A3. Referring to fig. 2, fig. 2 is a schematic flow chart showing a step 101 of a training method of an activity detection model provided by the present application.
Step a1, a first raw image is acquired.
The first original image includes a plurality of live face images and a plurality of non-live face images (the first original image refers to an image that has not been preprocessed). Illustratively, the first original image may be as follows: the method comprises the steps of firstly, taking a living body face image, secondly, taking a black-and-white paper face image, thirdly, taking a static face displayed by a mobile phone, a tablet and a computer screen by a living body detection device, fourthly, taking a face in a video played by the mobile phone, the tablet and the computer screen by the living body detection device, and the like, wherein the secondly to the fourthly, all are non-living body face images.
Step a2, detecting a face region in the first original image, and acquiring the sample image corresponding to the face region.
The face detection method includes, but is not limited to, combination of one or more models such as single shot multi-box face detection (SSD) and target detection (yollo).
Because the data volume of the face area is large, in order to further optimize the calculated amount in the training process, the application provides the following optional embodiments:
as an alternative embodiment of the present application, step a2 includes the following steps a21 to a 24. Referring to fig. 3, fig. 3 is a specific schematic flowchart illustrating a step a2 in the training method of the living body detection model provided in the present application.
Step A21, obtaining a first face frame of the first original image through a face detection model; the first face frame is used for marking a face area.
And A22, enlarging the size of the first face frame to a second preset size to obtain the enlarged first face frame.
On one hand, in order to ensure that a complete face image is obtained and prevent a partial area of a face from overflowing out of the first face frame, on the other hand, effective information of the living body detection is not only contained in the first face frame but also contained in the first face frame and a background nearby the first face frame, so that the terminal device enlarges the first face frame to a second preset size to obtain the enlarged first face frame, and the accuracy of the living body detection is improved.
The second preset size may be a fixed size (that is, a fixed size value is preset), or the second preset size may be a size obtained according to a preset ratio, that is, the size of the first frame is multiplied by the preset ratio to obtain the second preset size.
Step A23, intercepting the area corresponding to the amplified first face frame in the first original image to obtain a second original image.
Step a24, reducing the size of the second original image to a third preset size to obtain the sample image.
In order to improve the model training efficiency, the terminal device reduces the size of the second original image to a third preset size, and reduces the calculation amount of the model, so as to improve the model training efficiency.
It is to be understood that the "reducing the size of the second original image to the third preset size" refers to reducing the entire second original image to the third preset size, and does not refer to cutting out an area of the third preset size in the second original image.
It is noted that as the size of the sample image is infinitely small, the detection accuracy of the trained second convolutional network is reduced. And when the size of the sample image is larger, the detection accuracy of the trained second convolution network is high, but the calculation amount is large. Therefore, a more appropriate third preset size can be drawn up according to the comprehensive requirements of the detection precision, the calculation efficiency and the like of the actual application scene.
Step A3, calculating the original Fourier spectrogram of the sample image according to the sample image.
In step a3, the fourier spectrogram of the sample image can be directly used as the original fourier spectrogram, or the original fourier spectrogram can be calculated by the following alternative embodiments:
as an alternative embodiment of the present application, step A3 includes the following steps a31 to a 33. Referring to fig. 4, fig. 4 is a specific schematic flowchart illustrating a step a3 in the training method of the living body detection model provided in the present application.
Step a31, calculating an initial fourier spectrum of the sample image.
The fourier spectrogram can reflect the difference of the live/non-live face images in the frequency domain (for example, the face of the electronic screen may have moire), thereby being helpful for classifying the live/non-live face.
Since the calculation of the fourier spectrogram is a conventional technique, it is not described herein again.
And step A32, performing normalization processing on the initial Fourier spectrum.
Because the same object has certain difference in the corresponding different images of the same object under the influence of different shooting environments or software and hardware limitations and other factors, the normalization processing is performed to avoid the difference.
Step A33, reducing the size of the normalized initial Fourier spectrum to a first preset size to obtain the original Fourier spectrum.
In order to improve the model training efficiency, the terminal device reduces the size of the initial fourier spectrogram to a first preset size, so that the calculation amount of the model is reduced, and the model training efficiency is improved.
It is to be understood that the "reducing the size of the initial fourier spectrogram to the first preset size" refers to reducing the entire initial fourier spectrogram to the first preset size, and does not refer to an area of the initial fourier spectrogram that is cut off by the first preset size.
Step 102, inputting the sample data set into the first branch network and the second branch network to obtain a joint loss function, and training the second convolutional network according to the joint loss function to obtain a trained second convolutional network.
The way of training the model to be trained includes the following two ways:
the method comprises the following steps: a student model having a simple network structure is taken as a second convolutional network, and a teacher model having a complex network structure is taken as a third convolutional network (both the second convolutional network and the third convolutional network are used for the living body detection). And respectively inputting the same sample images into a teacher model and a student model, and training the student model by adopting a learning distillation mode to obtain a trained second convolution network.
The method II comprises the following steps: as an alternative embodiment of the present application, step 102 includes the following steps B1 through B7. Referring to fig. 5, fig. 5 is a schematic flow chart showing a step 102 of the training method of the living body detection model provided in the present application.
And circularly executing the following steps B1 to B7 on each sample data set to obtain a trained second convolution network.
Step B1, inputting the sample image into the first convolution network, and obtaining the feature data output by the first convolution network.
And step B2, inputting the characteristic data into a second convolution network to obtain a prediction result output by the second convolution network.
And B3, calculating a first loss function between the prediction result and the annotation information.
And step B4, inputting the characteristic data into a third convolution network to obtain a predicted Fourier spectrogram output by the third convolution network.
Step B5, calculating a second loss function between the predicted fourier spectrogram and the original fourier spectrogram.
Step B6, using the first loss function and the second loss function as the joint loss function.
And B7, updating the parameters of the second convolution network according to the joint loss function.
In order to better explain the steps B1 to B7, the present embodiment explains the steps B1 to B7 with reference to the drawings. Referring to fig. 6, fig. 6 shows a schematic diagram of an initial model provided by the present application. It should be noted that the initial model in fig. 6 is only used as an example, and parameters such as a network structure and a feature map of the initial model are not limited at all. As shown in fig. 6, a dashed box 1 represents a first convolutional network, a dashed box 2 represents a second convolutional network, and a dashed box 3 represents a third convolutional network. The initial model comprises two branch networks, wherein the first branch network is formed by a dashed frame 1 and a dashed frame 2, and the second branch network is formed by a dashed frame 1 and a dashed frame 3. Preferably, the first convolutional network and the second convolutional network may preferably adopt a MINIFASnet network hierarchy, and the third convolutional network may preferably adopt an FTGenerator network hierarchy.
As shown in fig. 6, the sample image is input into the initial model. The sample image is passed through a first convolution network (dashed box 1) to obtain feature data output by the first convolution network. The feature data is processed by a second convolutional network (dashed box 2) to obtain a second convolutional network prediction result. And calculating a first Loss function, namely 'Softmax Loss' (the error type is cross entropy), of the prediction result corresponding to the labeling information corresponding to the sample image. The feature data is processed by a third convolution network (dashed box 3) to obtain a predicted fourier spectrogram. And calculating a second Loss function, namely 'FT Loss' (the error type is Mean Square Error (MSE)), corresponding to the original Fourier spectrogram and the predicted Fourier spectrogram. And taking the first loss function and the second loss function as a joint loss function. The parameters of the second initial convolutional network (dashed box 2) are updated according to the joint loss function. And circularly executing the steps B1 to B7 on each sample data set to obtain a trained second convolution network.
The trigger mechanism for completing training comprises the following two mechanisms: firstly, the sample data set is completely trained, secondly, the value of the joint loss function is not reduced (namely, the joint loss function is converged), and thirdly, the value of the first loss function of the first branch network is not reduced (namely, the first loss function is converged).
And 103, taking the first convolution network and the trained second convolution network as a target living body detection model.
In order to reduce the amount of calculation of the model, the third convolutional network in the initial model is cut, and the remaining first convolutional network and the trained second convolutional network (i.e., the first branch network) are used as the target living body detection model.
It will be appreciated that the third convolutional network is used only in the training phase to assist in improving the degree of detection of the second convolutional network. The target living body detection model can keep the detection precision under the condition of having smaller calculation amount.
In this embodiment, the second convolutional network in the first branch network is jointly trained by the first branch network in the initial model and the second branch network. And taking the first convolution network and the trained second convolution network as a target living body detection model. In the process, the third convolutional network in the second branch network is only used for assisting in training the second convolutional network, so that the excellent detection precision of the target living body detection model is ensured, unnecessary calculation amount is reduced, and the living body detection efficiency is improved.
Referring to fig. 7, fig. 7 is a schematic flow chart illustrating a method for in vivo detection provided by the present application. As shown in fig. 7, the method is applied to a living body detecting apparatus, and includes the steps of:
step 701, obtaining an image to be detected.
The image to be detected may be an original image (in order to better distinguish the original images corresponding to different images, the original image of the image to be detected is referred to as a third original image, and the third original image refers to an image acquired by the living body detection device or an image input to the living body detection device without being preprocessed). However, since the third original image often includes a large amount of redundant data (i.e. non-face regions), and the large amount of redundant data causes a large amount of calculation of the target living body detection model and low processing efficiency, in order to improve the processing efficiency of the model, the third original image may be preprocessed, specifically, as shown in the following optional embodiment:
as an alternative embodiment of the present application, step 701 includes the following steps 7011 to 7015. Referring to fig. 8, fig. 8 is a specific schematic flowchart illustrating a step 701 in a method for detecting a living body according to the present application.
Step 7011, a third original image is obtained.
Step 7012, a second face frame of the third original image is obtained through a face detection model; the second face frame is used for marking a face area.
The face detection method includes, but is not limited to, combination of one or more models such as single shot multi-box face detection (SSD) and target detection (yollo).
Step 7013, the size of the second face frame is enlarged to a fourth preset size, and the enlarged second face frame is obtained.
On one hand, in order to ensure that a complete face image is obtained and prevent a partial area of the second face from overflowing out of the second face frame, on the other hand, effective information of the living body detection is not only contained in the second face frame but also contained in the second face frame and a nearby background thereof, so that the terminal device expands the second face frame to a fourth preset size to obtain an enlarged second face area, and the accuracy of the living body detection is improved.
The fourth preset size may be a fixed size (that is, a fixed size value is preset), or the fourth preset size may be a size obtained according to a preset ratio, that is, the size of the second face frame is multiplied by the preset ratio to obtain the fourth preset size.
The fourth predetermined size may be the same as or different from the second predetermined size in the embodiment of fig. 3.
Step 7014, an area corresponding to the amplified second face frame is intercepted from the third original image, so as to obtain a fourth original image.
Step 7015, the size of the fourth original image is reduced to a fifth preset size, so as to obtain the image to be detected.
It is to be understood that the "reducing the size of the fourth original image to the fifth preset size" refers to reducing the entire fourth original image to the fifth preset size, and does not refer to cutting out the area of the fifth preset size in the fourth original image.
The fifth predetermined size may be the same as or different from the third predetermined size in the embodiment of fig. 3.
Step 702, inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used to jointly train the second convolutional network.
And 703, determining whether the image to be detected is a living body face image or not according to the detection result.
And the detection result is the confidence coefficient of the target image to be recognized and is used for representing the probability that the target image to be recognized is the human face image.
And if the image to be detected is single, determining whether the image to be detected is a living human face image or not by the living body detection equipment according to the threshold value and the detection result of the single image to be detected. And if the number of the images to be detected is multiple, the living body detection equipment determines whether the image to be detected is a living body face image or not according to the threshold value and the detection results of the multiple images to be detected. The "single image to be detected" means that only one image to be detected is needed in one in vivo detection, and the "multiple images to be detected" means that multiple images to be detected are needed in one in vivo detection. The two ways are shown in the following two alternative embodiments:
as an alternative embodiment of the present application, step 703 includes the following steps C1 to C2. Referring to fig. 9, fig. 9 is a specific schematic flowchart illustrating step 703 in a method for in vivo detection provided by the present application.
And step C1, if the detection result is larger than the threshold value, confirming that the image to be detected is a living body face image.
And step C2, if the detection result is not greater than the threshold value, confirming that the image to be detected is a non-living body face image.
As an alternative embodiment of the present application, steps 701 to 703 specifically include steps D1 to D4. Referring to fig. 10, fig. 10 is a specific schematic flow chart of another method for in-vivo detection provided in the present application.
And D1, acquiring a plurality of images to be detected.
The plurality of images to be detected can be images at any time in the shooting process.
Preferably, the plurality of target images to be recognized preferentially use k frame images (a plurality of images to be recognized) adjacent in time series.
Please refer to step 701 for the processing procedure of each image to be detected, which is not described herein again.
And D2, sequentially inputting the images to be detected into the target living body detection model to obtain the detection results corresponding to the images to be detected output by the target living body detection model.
And D3, if the detection results are all larger than the threshold value, confirming that the image to be detected is a living body face image.
And D4, if the detection results are not all larger than the threshold value, confirming that the image to be detected is a non-living body face image.
The number of the plurality of images to be recognized can be set according to the detection accuracy. The larger the number of images to be recognized, the higher the detection accuracy of the living body detection. The smaller the number of images to be recognized, the lower the detection accuracy of the living body detection, but the larger the amount of calculation. Therefore, the number of the images to be recognized can be set according to the detection precision and the comprehensive consideration of the calculated amount.
In order to more intuitively compare the detection accuracy between the steps C1 to C2 and the steps D3 to D4, the two are compared as follows: if the probability of success of the non-live attack in the steps C1 to C2 is p (p is less than 1), the probability of success of the non-live attack in the steps D3 to D4 is pk(k is the number of images to be recognized by the target). I.e., steps D3 through D4, may exponentially decrease the probability of success by a non-live attack.
In the present embodiment, the living body detection is performed using only the first branch network in the initial model. The second convolutional network in the first branch network is obtained by jointly training the first branch network through the first branch network and the second branch network. Therefore, the excellent detection precision is ensured, unnecessary calculation amount is reduced, and the living body detection efficiency is improved.
Fig. 11 shows a schematic diagram of an apparatus 11 for training a living body detection model, and fig. 11 shows a schematic diagram of an apparatus for training a living body detection model, and the apparatus for training a living body detection model shown in fig. 11 includes:
a first obtaining unit 111, configured to obtain an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network form a second branch network;
a training unit 112, configured to input the sample data set into the first branch network and the second branch network, calculate a joint loss function, train the second convolutional network according to the joint loss function, and obtain a trained second convolutional network;
and a clipping unit 113, configured to use the first convolutional network and the trained second convolutional network as a target living body detection model.
The application provides a training device of living body detection model, through the first branch network in the initial model with the second convolution network in the first branch network of second branch network joint training. And taking the first convolution network and the trained second convolution network as a target living body detection model. In the process, the third convolutional network in the second branch network is only used for assisting in training the second convolutional network, so that the excellent detection precision of the target living body detection model is ensured, unnecessary calculation amount is reduced, and the living body detection efficiency is improved.
The present application provides a living body detecting apparatus 12 applied to an access point device as shown in fig. 12, and referring to fig. 12, fig. 12 shows a schematic diagram of a living body detecting apparatus provided by the present application, and the living body detecting apparatus as shown in fig. 12 includes:
a second acquiring unit 121 for acquiring an image to be detected;
the detection unit 122 is configured to input the image to be detected into a target living body detection model, so as to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used for jointly training the second convolutional network;
and the judging unit 123 is configured to determine whether the image to be detected is a living body face image according to the detection result.
The living body detection device provided by the application only adopts the first branch network in the initial model to carry out the living body detection. The second convolutional network in the first branch network is obtained by jointly training the first branch network through the first branch network and the second branch network. Therefore, the excellent detection precision is ensured, unnecessary calculation amount is reduced, and the living body detection efficiency is improved.
Fig. 13 is a schematic diagram of a terminal device according to an embodiment of the present invention. As shown in fig. 13, a terminal device 13 of this embodiment includes: a processor 131, a memory 132, and a computer program 133, such as a training program for a living body detection model, stored in the memory 132 and executable on the processor 131. The processor 131, when executing the computer program 133, implements the steps of the above-described embodiments of a method for training a human detection model, such as the steps 101 to 103 shown in fig. 1. Alternatively, the processor 131, when executing the computer program 134, implements the functions of the units in the above-described device embodiments, such as the functions of the units 111 to 113 shown in fig. 11.
Illustratively, the computer program 133 may be divided into one or more units, which are stored in the memory 132 and executed by the processor 131 to accomplish the present invention. The one or more units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program 133 in the terminal device 13. For example, the computer program 133 may be divided into units with specific functions as follows:
the first acquisition unit is used for acquiring an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network form a second branch network;
a training unit, configured to input the sample data set into the first branch network and the second branch network, calculate a joint loss function, train the second convolutional network according to the joint loss function, and obtain a trained second convolutional network;
and the cutting unit is used for taking the first convolution network and the trained second convolution network as a target living body detection model.
The terminal device includes, but is not limited to, a processor 131 and a memory 132. Those skilled in the art will appreciate that fig. 13 is merely an example of one type of terminal device 13 and is not intended to limit one type of terminal device 13 and may include more or fewer components than shown, or some components may be combined, or different components, for example, the one type of terminal device may also include input-output devices, network access devices, buses, etc.
The Processor 131 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The storage 132 may be an internal storage unit of the terminal device 13, such as a hard disk or a memory of the terminal device 13. The memory 132 may also be an external storage device of the terminal device 13, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, provided on the terminal device 13. Further, the memory 132 may also include both an internal storage unit and an external storage device of the terminal device 13. The memory 132 is used for storing the computer programs and other programs and data required by the kind of terminal equipment. The memory 132 may also be used to temporarily store data that has been output or is to be output.
FIG. 14 is a schematic view of a living body detecting apparatus according to an embodiment of the present invention. As shown in fig. 14, a living body detecting apparatus 14 of this embodiment includes: a camera module 141, a processor 142, a memory 143, and a computer program 144, such as a biopsy program, stored in the memory 143 and executable on the processor 142. The processor 142, when executing the computer program 144, implements the steps of each of the above-described embodiments of the method for detecting a living body, such as steps 701 to 703 shown in fig. 7. Alternatively, the processor 142 implements the functions of the units in the above-mentioned device embodiments, for example, the functions of the units 121 to 123 shown in fig. 12, when executing the computer program 144.
Illustratively, the computer program 144 may be divided into one or more units, which are stored in the memory 143 and executed by the processor 142 to accomplish the present invention. The one or more elements may be a series of computer program instruction segments capable of performing certain functions that describe the execution of the computer program 144 in the one type of living being detection apparatus 14. For example, the computer program 144 may be divided into units with specific functions as follows:
the second acquisition unit is used for acquiring an image to be detected;
the detection unit is used for inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used for jointly training the second convolutional network;
and the judging unit is used for determining whether the image to be detected is a living body face image or not according to the detection result.
The living body detection equipment can be equipment with a face detection function, such as a card punch, a mobile terminal, a notebook computer, a tablet computer and the like. The living body detection device includes, but is not limited to, a camera module 141, a processor 142, and a memory 143. Those skilled in the art will appreciate that figure 14 is merely an example of one type of active detection device 14 and does not constitute a limitation of one type of active detection device 14 and may include more or fewer components than shown, or some components may be combined, or different components, for example, the one type of active detection device may also include input-output devices, network access devices, buses, etc.
The device module 141 is used for collecting image information.
The Processor 142 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic device, discrete hardware component, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 143 may be an internal storage unit of the living organism detection device 14, such as a hard disk or a memory of the living organism detection device 14. The memory 143 may also be an external storage device of the living body detection device 14, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, provided on the living body detection device 14. Further, the memory 143 may also include both internal and external memory units of the one type of living body detection device 14. The memory 143 is used to store the computer programs and other programs and data required by the one type of living organism detection apparatus. The memory 143 may also be used to temporarily store data that has been output or is to be output.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.
It should be noted that, for the information interaction, execution process, and other contents between the above-mentioned devices/units, the specific functions and technical effects thereof are based on the same concept as those of the embodiment of the method of the present application, and specific reference may be made to the part of the embodiment of the method, which is not described herein again.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-mentioned division of the functional units and modules is illustrated, and in practical applications, the above-mentioned function distribution may be performed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-mentioned functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored, and when the computer program is executed by a processor, the computer program implements the steps in the above-mentioned method embodiments.
The embodiments of the present application provide a computer program product, which when running on a mobile terminal, enables the mobile terminal to implement the steps in the above method embodiments when executed.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium and can implement the steps of the embodiments of the methods described above when the computer program is executed by a processor. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code to a photographing apparatus/biopsy device, a recording medium, computer Memory, Read-Only Memory (ROM), Random Access Memory (RAM), electrical carrier wave signals, telecommunication signals, and software distribution medium. Such as a usb-disk, a removable hard disk, a magnetic or optical disk, etc. In certain jurisdictions, computer-readable media may not be an electrical carrier signal or a telecommunications signal in accordance with legislative and patent practice.
In the above embodiments, the descriptions of the respective embodiments have respective emphasis, and reference may be made to the related descriptions of other embodiments for parts that are not described or illustrated in a certain embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus/network device and method may be implemented in other ways. For example, the above-described apparatus/network device embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implementing, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units.
It will be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It should also be understood that the term "and/or" as used in this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
As used in this specification and the appended claims, the term "if" may be interpreted contextually as "when", "upon" or "in response to" determining "or" in response to monitoring ". Similarly, the phrase "if it is determined" or "if [ a described condition or event ] is monitored" may be interpreted depending on the context to mean "upon determining" or "in response to determining" or "upon monitoring [ a described condition or event ]" or "in response to monitoring [ a described condition or event ]".
Furthermore, in the description of the present application and the appended claims, the terms "first," "second," "third," and the like are used for distinguishing between descriptions and not necessarily for describing or implying relative importance.
Reference throughout this specification to "one embodiment" or "some embodiments," or the like, means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the present application. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," or the like, in various places throughout this specification are not necessarily all referring to the same embodiment, but rather "one or more but not all embodiments" unless specifically stated otherwise. The terms "comprising," "including," "having," and variations thereof mean "including, but not limited to," unless expressly specified otherwise.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims (14)

1. A method of training a living body detection model, the method comprising:
acquiring an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network form a second branch network;
inputting the sample data set into the first branch network and the second branch network, calculating a joint loss function, and training the second convolutional network according to the joint loss function to obtain a trained second convolutional network;
and taking the first convolution network and the trained second convolution network as a target living body detection model.
2. The method of claim 1, wherein the sample data set comprises a sample image, an original fourier spectrogram corresponding to the sample image, and annotation information corresponding to the sample image; the marking information is used for marking the image type of the sample image; the image types comprise a living body face image and a non-living body face image;
the inputting the sample data into the first branch network and the second branch network, calculating a joint loss function, and training the second convolutional network according to the joint loss function to obtain a trained second convolutional network, including:
circularly executing the following steps by each sample data group to obtain the trained second convolution network;
the following steps are included:
inputting the sample image into the first convolution network to obtain feature data output by the first convolution network;
inputting the characteristic data into a second convolution network to obtain a prediction result output by the second convolution network;
calculating a first loss function between the prediction result and the labeling information;
inputting the characteristic data into a third convolution network to obtain a predicted Fourier spectrogram output by the third convolution network;
calculating a second loss function between the predicted fourier spectrogram and the original fourier spectrogram;
taking the first loss function and the second loss function as the joint loss function;
and updating the parameters of the second convolution network according to the joint loss function.
3. The method of claim 1, wherein the sample data set comprises a sample image, an original fourier spectrogram corresponding to the sample image, and annotation information corresponding to the sample image;
the obtaining of the initial model and the sample data set comprises;
acquiring a first original image;
detecting a face region in the first original image, and acquiring the sample image corresponding to the face region;
and calculating an original Fourier spectrogram of the sample image according to the sample image.
4. The method of claim 1, wherein computing the sample image raw fourier spectrogram from the sample image comprises:
calculating an initial Fourier spectrum of the sample image;
normalizing the initial Fourier spectrum;
and reducing the size of the normalized initial Fourier spectrum to a first preset size to obtain the original Fourier spectrum.
5. The method as claimed in claim 3, wherein said detecting a face region in the first original image and obtaining the sample image corresponding to the face region comprises:
obtaining a first face frame of the first original image through a face detection model; the first face frame is used for marking a face area;
the size of the first face frame is enlarged to a second preset size, and the enlarged first face frame is obtained;
intercepting an area corresponding to the amplified first face frame in the first original image to obtain a second original image;
and reducing the size of the second original image to a third preset size to obtain the sample image.
6. A method of in vivo testing, the method comprising:
acquiring an image to be detected;
inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used for jointly training the second convolutional network;
and determining whether the image to be detected is a living body face image or not according to the detection result.
7. The method of claim 6, wherein said acquiring the image to be detected comprises:
acquiring a third original image;
obtaining a second face frame of the third original image through a face detection model; the second face frame is used for marking a face area;
amplifying the size of the second face frame to a fourth preset size to obtain an amplified second face frame;
intercepting an area corresponding to the amplified second face frame in the third original image to obtain a fourth original image;
and reducing the size of the fourth original image to a fifth preset size to obtain the image to be detected.
8. The method of claim 6, wherein the determining whether the image to be detected is a living human face image according to the detection result comprises:
if the detection result is larger than the threshold value, confirming that the image to be detected is a living body face image;
and if the detection result is not greater than the threshold value, confirming that the image to be detected is a non-living body face image.
9. The method of claim 6, wherein said acquiring the image to be detected comprises:
acquiring a plurality of images to be detected;
inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model, wherein the detection result comprises the following steps:
sequentially inputting the images to be detected into the target living body detection model to obtain detection results corresponding to the images to be detected output by the target living body detection model;
determining whether the image to be detected is a living body face image according to the detection result, including:
if the detection results are all larger than the threshold value, confirming that the image to be detected is a living body face image;
and if the detection results are not all larger than the threshold value, confirming that the image to be detected is a non-living body face image.
10. A training apparatus for a living body detection model, the training apparatus comprising:
the first acquisition unit is used for acquiring an initial model and a sample data set; the initial model comprises a first convolution network, a second convolution network and a third convolution network; the first convolutional network and the second convolutional network form a first branch network; the first convolutional network and the third convolutional network form a second branch network;
a training unit, configured to input the sample data set into the first branch network and the second branch network, calculate a joint loss function, train the second convolutional network according to the joint loss function, and obtain a trained second convolutional network;
and the cutting unit is used for taking the first convolution network and the trained second convolution network as a target living body detection model.
11. An apparatus for in vivo testing, the apparatus comprising:
the second acquisition unit is used for acquiring an image to be detected;
the detection unit is used for inputting the image to be detected into a target living body detection model to obtain a detection result output by the target living body detection model; the detection result is used for representing the probability that the image to be detected is the living body face image; the target living body detection model is a first branch network in the initial model, and the first branch network comprises a first convolution network and a second convolution network; the initial model also comprises a second branch network; the first branch network and the second branch network are used for jointly training the second convolutional network;
and the judging unit is used for determining whether the image to be detected is a living body face image or not according to the detection result.
12. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1 to 5 when executing the computer program.
13. A living subject examination apparatus comprising a camera module, a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor realizes the steps of the method according to any one of claims 6 to 9 when executing the computer program.
14. A computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 5 or 6 to 9.
CN202110245292.5A 2021-03-05 2021-03-05 Training method and training device for living body detection model Active CN113158773B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110245292.5A CN113158773B (en) 2021-03-05 2021-03-05 Training method and training device for living body detection model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110245292.5A CN113158773B (en) 2021-03-05 2021-03-05 Training method and training device for living body detection model

Publications (2)

Publication Number Publication Date
CN113158773A true CN113158773A (en) 2021-07-23
CN113158773B CN113158773B (en) 2024-03-22

Family

ID=76884329

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110245292.5A Active CN113158773B (en) 2021-03-05 2021-03-05 Training method and training device for living body detection model

Country Status (1)

Country Link
CN (1) CN113158773B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114627534A (en) * 2022-03-15 2022-06-14 平安科技（深圳）有限公司 Living body discrimination method, electronic device, and storage medium
CN115147705A (en) * 2022-09-06 2022-10-04 平安银行股份有限公司 Face copying detection method and device, electronic equipment and storage medium

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB201520316D0 (en) * 2014-11-24 2015-12-30 Boeing Co Detecting and removing spoofing signals
CN107818313A (en) * 2017-11-20 2018-03-20 腾讯科技（深圳）有限公司 Vivo identification method, device, storage medium and computer equipment
CN109684993A (en) * 2018-12-21 2019-04-26 普联技术有限公司 A kind of face identification method based on nostril information, system and equipment
WO2019152983A2 (en) * 2018-02-05 2019-08-08 Board Of Trustees Of Michigan State University System and apparatus for face anti-spoofing via auxiliary supervision
CN110610237A (en) * 2019-09-17 2019-12-24 普联技术有限公司 Quantitative training method and device of model and storage medium
WO2020063475A1 (en) * 2018-09-25 2020-04-02 清华大学 6d attitude estimation network training method and apparatus based on deep learning iterative matching
CN111382646A (en) * 2018-12-29 2020-07-07 Tcl集团股份有限公司 Living body identification method, storage medium and terminal equipment
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
CN111860078A (en) * 2019-04-30 2020-10-30 北京眼神智能科技有限公司 Face silence living body detection method and device, readable storage medium and equipment
CN111914758A (en) * 2020-08-04 2020-11-10 成都奥快科技有限公司 Face in-vivo detection method and device based on convolutional neural network
CN112329696A (en) * 2020-11-18 2021-02-05 携程计算机技术（上海）有限公司 Face living body detection method, system, equipment and storage medium

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB201520316D0 (en) * 2014-11-24 2015-12-30 Boeing Co Detecting and removing spoofing signals
CN107818313A (en) * 2017-11-20 2018-03-20 腾讯科技（深圳）有限公司 Vivo identification method, device, storage medium and computer equipment
WO2019152983A2 (en) * 2018-02-05 2019-08-08 Board Of Trustees Of Michigan State University System and apparatus for face anti-spoofing via auxiliary supervision
WO2020063475A1 (en) * 2018-09-25 2020-04-02 清华大学 6d attitude estimation network training method and apparatus based on deep learning iterative matching
CN109684993A (en) * 2018-12-21 2019-04-26 普联技术有限公司 A kind of face identification method based on nostril information, system and equipment
CN111382646A (en) * 2018-12-29 2020-07-07 Tcl集团股份有限公司 Living body identification method, storage medium and terminal equipment
CN111860078A (en) * 2019-04-30 2020-10-30 北京眼神智能科技有限公司 Face silence living body detection method and device, readable storage medium and equipment
CN110610237A (en) * 2019-09-17 2019-12-24 普联技术有限公司 Quantitative training method and device of model and storage medium
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
CN111914758A (en) * 2020-08-04 2020-11-10 成都奥快科技有限公司 Face in-vivo detection method and device based on convolutional neural network
CN112329696A (en) * 2020-11-18 2021-02-05 携程计算机技术（上海）有限公司 Face living body detection method, system, equipment and storage medium

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
WENYUN SUN 等: "Face Spoofing Detection Based on Local Ternary Label Supervision in Fully Convolutional Networks", 《IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY》, vol. 15, 3 April 2020 (2020-04-03), pages 3181, XP011784552, DOI: 10.1109/TIFS.2020.2985530 *
杨成: "基于卷积神经网络的人脸活体检测方法研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》, no. 2021, 15 February 2021 (2021-02-15), pages 138 - 1806 *
陈浩楠: "基于光照一致化和上下文感知的人脸活体检测算法研究", 《中国博士学位论文全文数据库 (信息科技辑)》, no. 2021, 15 January 2021 (2021-01-15), pages 138 - 255 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114627534A (en) * 2022-03-15 2022-06-14 平安科技（深圳）有限公司 Living body discrimination method, electronic device, and storage medium
CN115147705A (en) * 2022-09-06 2022-10-04 平安银行股份有限公司 Face copying detection method and device, electronic equipment and storage medium
CN115147705B (en) * 2022-09-06 2023-02-03 平安银行股份有限公司 Face copying detection method and device, electronic equipment and storage medium

Also Published As

Publication number Publication date
CN113158773B (en) 2024-03-22

Similar Documents

Publication Publication Date Title
US10984556B2 (en) Method and apparatus for calibrating relative parameters of collector, device and storage medium
US20200110965A1 (en) Method and apparatus for generating vehicle damage information
US20210004587A1 (en) Image detection method, apparatus, device and storage medium
CN109376256B (en) Image searching method and device
CN109977832B (en) Image processing method, device and storage medium
CN110941978B (en) Face clustering method and device for unidentified personnel and storage medium
CN110349161B (en) Image segmentation method, image segmentation device, electronic equipment and storage medium
CN112149615A (en) Face living body detection method, device, medium and electronic equipment
CN113158773B (en) Training method and training device for living body detection model
CN113255516A (en) Living body detection method and device and electronic equipment
CN108241855B (en) Image generation method and device
CN112149570A (en) Multi-person living body detection method and device, electronic equipment and storage medium
CN111325709A (en) Wireless capsule endoscope image detection system and detection method
CN109064464B (en) Method and device for detecting burrs of battery pole piece
CN113780469A (en) Training method, medium, device and computing equipment of image recognition model
CN112287905A (en) Vehicle damage identification method, device, equipment and storage medium
CN110349108B (en) Method, apparatus, electronic device, and storage medium for processing image
CN110210314B (en) Face detection method, device, computer equipment and storage medium
CN111369557A (en) Image processing method, image processing device, computing equipment and storage medium
CN109165572B (en) Method and apparatus for generating information
CN111898529B (en) Face detection method and device, electronic equipment and computer readable medium
US20220122341A1 (en) Target detection method and apparatus, electronic device, and computer storage medium
CN113705666B (en) Split network training method, use method, device, equipment and storage medium
CN114067145A (en) Passive optical splitter detection method, device, equipment and medium
CN112989869B (en) Optimization method, device, equipment and storage medium of face quality detection model

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