CN113392742A - Abnormal action determination method and device, electronic equipment and storage medium - Google Patents
Abnormal action determination method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113392742A CN113392742A CN202110625830.3A CN202110625830A CN113392742A CN 113392742 A CN113392742 A CN 113392742A CN 202110625830 A CN202110625830 A CN 202110625830A CN 113392742 A CN113392742 A CN 113392742A
- Authority
- CN
- China
- Prior art keywords
- image
- abnormal
- target
- frame
- action
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 121
- 238000000034 method Methods 0.000 title claims abstract description 69
- 230000009471 action Effects 0.000 claims abstract description 79
- 230000008859 change Effects 0.000 claims abstract description 20
- 239000013598 vector Substances 0.000 claims description 73
- 238000013527 convolutional neural network Methods 0.000 claims description 38
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000012790 confirmation Methods 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 14
- 230000006870 function Effects 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 238000004590 computer program Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 210000000544 articulatio talocruralis Anatomy 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 210000000629 knee joint Anatomy 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
Images
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the application provides an abnormal action determining method, an abnormal action determining device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring a target video clip corresponding to a target action; acquiring human skeleton key points corresponding to each frame of image in a target video clip, and acquiring motion posture characteristics corresponding to the target video clip based on the human skeleton key points corresponding to each frame of image; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value. The method has the advantages that the change of the human body posture in the target video clip can be accurately described on the basis of the motion posture characteristics constructed on the basis of the key points of the human body skeleton in each frame of image in the target video clip, the confirmation of abnormal actions can be accurately and timely carried out by utilizing the motion posture characteristics and the abnormal recognition model, the method is simple and convenient to use, and the method is easy to widely apply.
Description
Technical Field
The present application relates to the field of computer vision technologies, and in particular, to a method and an apparatus for determining an abnormal motion, an electronic device, and a storage medium.
Background
When the body building is carried out through sports, if the movement is abnormal, the movement effect is easily influenced due to insufficient standards, and the body can be damaged, so that the determination of whether the movement is abnormal or not is particularly important in the movement.
When determining whether the action is abnormal, a professional teacher usually relies on manual experience to detect the abnormal action, however, the number of professional teachers is limited, abnormal actions in a large amount of actions cannot be analyzed in time, and the method relies on manual analysis and is difficult to be widely applied.
Disclosure of Invention
The embodiment of the application provides an abnormal action determining method and device, electronic equipment and a storage medium, and can effectively solve the problems that the abnormal action is not determined timely enough and is difficult to be widely applied.
According to a first aspect of embodiments of the present application, there is provided an abnormal motion determination method, including: acquiring a target video clip corresponding to a target action; acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-coder and a convolutional neural network; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value.
According to a second aspect of embodiments of the present application, there is provided an abnormal operation determination apparatus including: the acquisition module is used for acquiring a target video clip corresponding to the target action; the construction module is used for acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; the recognition module is used for inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, and the abnormal recognition model is a pre-trained variational self-encoder and a convolutional neural network; and the determining module is used for determining whether the target action in the video clip is an abnormal action or not according to the likelihood value.
According to a third aspect of embodiments of the present application, there is provided an electronic device comprising one or more processors; a memory; one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the method as applied to an electronic device, as described above.
According to a fourth aspect of the embodiments of the present application, there is provided a computer-readable storage medium having a program code stored therein, wherein the method described above is performed when the program code runs.
Acquiring a target video clip corresponding to a target action by adopting the abnormal action determining method provided by the embodiment of the application; acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-coder and a convolutional neural network; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value. The method has the advantages that the change of the human body posture in the target video clip can be accurately described on the basis of the motion posture characteristics constructed on the basis of the key points of the human body skeleton in each frame of image in the target video clip, the confirmation of abnormal actions can be accurately and timely carried out by utilizing the motion posture characteristics and the abnormal recognition model, the method is simple and convenient to use, and the method is easy to widely apply.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a flowchart of an abnormal operation determination method according to an embodiment of the present application;
fig. 2 is a flowchart of an abnormal operation determination method according to another embodiment of the present application;
FIG. 3 is a flow chart of a method for determining abnormal actions according to yet another embodiment of the present application;
FIG. 4 is a block diagram of an anomaly identification model provided in one embodiment of the present application;
FIG. 5 is a functional block diagram of an abnormal operation determination apparatus according to an embodiment of the present application;
FIG. 6 is a functional block diagram of an abnormal operation determination apparatus according to another embodiment of the present application;
fig. 7 is a block diagram of an electronic device for executing an abnormal operation determination method according to an embodiment of the present application.
Detailed Description
When the body building is carried out through sports, if the movement is abnormal, the movement effect is easily influenced due to insufficient standards, and the body can be damaged, so that the determination of whether the movement is abnormal or not is particularly important in the movement.
When determining whether the action is abnormal, a professional teacher usually relies on manual experience to detect the abnormal action, however, the number of professional teachers is limited, abnormal actions in a large amount of actions cannot be analyzed in time, and the method relies on manual analysis and is difficult to be widely applied.
In order to solve the above problem, an embodiment of the present application provides an abnormal motion determining method, which obtains a target video segment corresponding to a target motion; acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-coder and a convolutional neural network; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value. The method has the advantages that the change of the human body posture in the target video clip can be accurately described on the basis of the motion posture characteristics constructed on the basis of the key points of the human body skeleton in each frame of image in the target video clip, the confirmation of abnormal actions can be accurately and timely carried out by utilizing the motion posture characteristics and the abnormal recognition model, the method is simple and convenient to use, and the method is easy to widely apply.
The scheme in the embodiment of the present application may be implemented by using various computer languages, for example, object-oriented programming language Java and transliterated scripting language JavaScript, Python, and the like.
In order to make the technical solutions and advantages of the embodiments of the present application more apparent, the following further detailed description of the exemplary embodiments of the present application with reference to the accompanying drawings makes it clear that the described embodiments are only a part of the embodiments of the present application, and are not exhaustive of all embodiments. It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict.
Referring to fig. 1, an abnormal motion determination method provided in an embodiment of the present application may be applied to an electronic device, where the electronic device may be a smart phone, a computer, a server, or the like, and the method may specifically include the following steps.
The target action refers to a certain designated action, and the electronic device may acquire a target video segment corresponding to the target action to determine whether the target action in the target video segment is an abnormal action.
In some embodiments, the target video clip may be obtained by a user by using an image capture device to capture the target video clip, and uploading the target video clip to a designated application program, so that the electronic device may obtain the target video clip through the application program.
In some embodiments, the user may take a video to be recognized including the target motion by using the image acquisition device, so that the electronic device may acquire the video to be recognized. After the video to be identified is obtained, the electronic equipment can identify the human skeleton key points of each frame of image in the video to be identified; and extracting a target video clip corresponding to the target action from the video to be identified based on the human skeleton key point.
Specifically, each frame of image in the video to be identified is detected through a human skeleton key point detection algorithm, so that human skeleton key points corresponding to each frame of image and three-dimensional coordinates of the human skeleton key points can be obtained; constructing a posture vector of each frame of image based on the three-dimensional coordinates of the human skeleton key points; and clustering the attitude vectors by using a clustering algorithm to obtain a clustering result, and finally extracting a video segment corresponding to the target action from video data based on the clustering result.
As an implementation mode, each frame of image in a video to be identified is detected through a human skeleton key point detection algorithm, so that human skeleton key points corresponding to each frame of image and three-dimensional coordinates of the human skeleton key points can be obtained; and constructing a posture vector of each frame of image based on the three-dimensional coordinates of the human skeleton key points. After the pose vector corresponding to each frame of image is obtained, the image including the target motion may be designated as a target image, and the pose vector corresponding to the target image may be designated as a target vector. And sequentially calculating the similarity between each attitude vector and the target vector, taking the image corresponding to the attitude vector with the similarity larger than a preset value as a candidate image, finally extracting the candidate image and the target image from the video to be identified, and combining the candidate image and the target image into a video segment corresponding to the target action. The specific process of constructing the pose vector corresponding to each frame of image will be described in detail in the following embodiments.
After the target video clip is obtained, a human skeleton key point identification algorithm can be used for identifying human skeleton key points corresponding to each frame of image in the target video clip. After the human skeleton key points in each frame of image are identified, three-dimensional coordinates corresponding to the human skeleton key points can be acquired.
Based on the human skeleton key points corresponding to each frame of image, a posture vector corresponding to each frame of image can be constructed, wherein the posture vector comprises parameters of multiple dimensions and is used for describing the human posture in the image. The multiple dimensions may refer to different angles formed between different human skeleton key points, and the parameters of the multiple dimensions refer to cosine values corresponding to the multiple different angles. After the attitude vector corresponding to each frame of image is obtained, extracting the parameter of the same dimension in each attitude vector, and obtaining the time sequence vector corresponding to each dimension according to the frame number of the image used for constructing the attitude vector in the target video segment, wherein the time sequence vector is used for describing the change mode of the same dimension on the time sequence.
And each dimension corresponds to a time sequence vector, and then the time sequence vectors corresponding to each dimension are combined into a matrix, so that the motion attitude characteristics can be obtained. Each time sequence vector can describe a time-varying mode of one dimension, the matrix comprises time sequence vectors of multiple dimensions, and the multiple dimensions can describe human body gestures, so that the motion gesture features can describe varying modes of the human body gestures in the target video segment.
And step 130, inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-encoder and a convolutional neural network.
Generally, the standard pattern of the target action has a uniform specification, but the abnormal pattern of the target action is numerous, and if a supervised machine learning algorithm is adopted, all the abnormal patterns of the target action need to be simulated, which is obviously difficult to realize.
The variational self-encoder is an unsupervised machine learning mode, and a variational self-encoder and a convolutional neural network are used for learning a standard mode of target action to obtain the abnormal recognition model. When determining the abnormal motion, the motion posture feature may be input into an abnormality recognition model, and a likelihood value corresponding to the motion posture feature may be obtained. If the likelihood value is small, the mode of the target action in the target video segment corresponding to the motion attitude characteristic is indicated, and the difference between the mode of the target action and the standard mode of the target action is large, so that the target action can be regarded as abnormal action; if the likelihood value is large, the mode of the target action in the target video segment corresponding to the motion attitude feature is indicated, and the difference between the mode of the target action and the standard mode of the target action is small, so that the target action can be considered as the standard action.
And 140, determining whether the target action in the video clip is an abnormal action or not according to the likelihood value.
After the likelihood value is obtained, a preset likelihood value can be obtained, and whether the target action in the video clip is an abnormal action or not is determined according to the magnitude relation between the likelihood value and the preset likelihood value.
If the likelihood value is smaller than the preset likelihood value, determining the target action in the video clip as an abnormal action; if the likelihood value is greater than or equal to the preset likelihood value, the target action in the video clip can be determined to be a standard action.
According to the abnormal action determining method provided by the embodiment of the application, a target video clip corresponding to a target action is obtained; acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment on the basis of the human skeleton key points, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-encoder and a convolutional neural network; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value. The method has the advantages that the change of the human body posture in the target video clip can be accurately described on the basis of the motion posture characteristics constructed on the basis of the key points of the human body skeleton in each frame of image in the target video clip, the confirmation of abnormal actions can be accurately and timely carried out by utilizing the motion posture characteristics and the abnormal recognition model, the method is simple and convenient to use, and the method is easy to widely apply. And an abnormal recognition model is obtained through unsupervised training of the variational self-encoder and the convolutional neural network, the training process is simple and convenient, and the variational self-encoder can extract the characteristics in the motion attitude characteristics, so that the accuracy of abnormal motion recognition can be improved by using the abnormal recognition model.
Referring to fig. 2, another embodiment of the present application provides an abnormal motion determination method, which focuses on the process of obtaining the motion posture characteristic corresponding to the target video segment on the basis of the foregoing embodiment, and specifically, the method may include the following steps.
And step 220, identifying human skeleton key points corresponding to each frame of image in the target video clip.
After the target video segment is obtained, human skeleton key points in each frame of image can be identified by using a human skeleton key point detection algorithm, and after the human skeleton key points are identified, the three-dimensional coordinates of each human skeleton key point can be obtained.
And 230, constructing a posture vector corresponding to each frame of image based on the human skeleton key points corresponding to each frame of image, wherein the posture vector comprises parameters of multiple dimensions.
For each frame of image, after the human skeleton key points and the three-dimensional coordinates corresponding to the human skeleton key points are obtained, angles formed by different human skeleton key points can be calculated according to the three-dimensional coordinates corresponding to the human skeleton key points, and attitude vectors are obtained.
Specifically, when calculating the posture vector corresponding to each frame of image, the obtained human skeleton key points may be combined into a key point sequence according to a first preset rule, and each key point sequence includes 3 skeleton key points. For example, the obtained key points of the human skeleton are a left tiptoe, a left ankle joint and a left knee joint, and the key point sequence can be [ the left tiptoe, the left ankle joint and the left knee joint ].
Each key point sequence can calculate an angle, and when the angle is calculated according to the key point sequence, the calculation can be performed according to a preset calculation rule. For example, for any one of the key point sequences [ key point 1, key point 2, and key point 3], the calculation rule may be to calculate an angle between a line segment formed by the key points 1 and 2 and a line segment formed by the key points 2 and 3.
A plurality of angles can be formed among different human key points, the plurality of angles are a plurality of dimensions, and cosine values of the angles are parameters of the dimensions. Assuming that there are m angles formed between different key points of the human body, the corresponding attitude vector of each frame of image can be expressed as [ cos [ ]1，cos2，...，cosm]。
After the attitude vector corresponding to each frame of image is obtained, parameters of the same dimension in each attitude vector can be extracted, and a time sequence vector corresponding to each dimension is obtained according to the frame number of the image used for constructing the attitude vector in the target video segment.
As an embodiment, after obtaining the pose vector corresponding to each frame of image, the number of frames corresponding to the image may be marked in the pose vector, for example, the pose vector corresponding to the first frame of image may be marked as
The attitude vector corresponding to the nth frame can be recorded as
Extracting parameters of the same dimension in each attitude vector, wherein cos is the parameter of the first dimension if the parameter of the first dimension is extracted1The time sequence vector corresponding to the first dimension can be obtained as
The m-th dimension corresponds to a timing vector of
And step 250, taking a matrix formed by the time sequence vectors corresponding to each dimension as the motion attitude characteristic.
After the time sequence vectors corresponding to each dimension are obtained, the time sequence vectors can be combined into a matrix as the motion attitude feature. Taking m dimensions and n frames as an example of a target video segment, the motion posture features are as follows:
and 260, inputting the motion posture characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion posture characteristics, wherein the abnormal recognition model is a pre-trained variational self-encoder and a convolutional neural network.
And step 270, determining whether the target action in the video clip is an abnormal action according to the likelihood value.
The steps 260 to 270 refer to the corresponding parts of the previous embodiments, and are not described herein again.
The abnormal action determining method provided by the embodiment of the application identifies the key points of the human skeleton corresponding to each frame of image in the target video clip; constructing a posture vector corresponding to each frame of image based on the human skeleton key points, wherein the posture vector comprises parameters of multiple dimensions; extracting parameters of the same dimension in each attitude vector, and obtaining a time sequence vector corresponding to each dimension according to the frame number of an image used for constructing the attitude vector in the target video segment; taking a matrix formed by the time sequence vectors corresponding to each dimension as the motion attitude characteristic; and inputting the motion attitude characteristics into the abnormal recognition model, determining a likelihood value, and finally determining whether the target action is an abnormal action according to the likelihood value. The motion posture feature constructed based on the human skeleton key points in each frame of image in the target video segment can accurately describe the human posture change in the target video segment, so that when the motion posture feature is used for determining whether the target action is abnormal action, a more accurate result can be obtained, and the accuracy of determining the abnormal action is improved.
Referring to fig. 3, a further embodiment of the present application provides an abnormal operation determination method, which focuses on the process of obtaining an abnormal recognition model based on the foregoing embodiment, and specifically, the method may include the following steps.
The steps 310 to 320 can refer to the corresponding parts of the previous embodiments, and are not described herein again.
And acquiring a sample segment, wherein the sample segment is a video segment corresponding to a standard target action. When the sample segment is obtained, a standard target action performed by a professional can be shot to obtain a video segment corresponding to the standard target action.
And step 340, obtaining the sample motion posture characteristics corresponding to the sample segments.
After the sample segment is obtained, a sample motion gesture feature corresponding to the sample segment may be obtained. The motion posture characteristic of the sample can be obtained by identifying human skeleton key points corresponding to each frame of image in the sample fragment; constructing a posture vector corresponding to each frame of image based on the human skeleton key points, wherein the posture vector comprises parameters of multiple dimensions; extracting parameters of uniform dimensions in each attitude vector, and obtaining a time sequence vector corresponding to each dimension according to the number of frames of an image used for constructing the attitude vector in the sample segment; and taking a matrix formed by the time sequence vectors corresponding to each dimension as the motion attitude characteristic.
For a detailed description of the manner of obtaining the motion gesture features of the sample, reference may be made to the descriptions in step 220 to step 240 in the foregoing embodiments, which are not described herein again.
And 350, training the variational self-encoder and the convolutional neural network based on the sample motion posture characteristics to obtain the abnormal recognition model.
After the sample motion attitude feature is obtained, the sample motion attitude feature may be input to the convolutional neural network and the variational self-encoder, and the convolutional neural network and the variational self-encoder are trained to obtain the anomaly recognition model.
Referring to FIG. 4, a block diagram of an anomaly recognition model is shown. The anomaly identification model comprises a variational self-encoder and a convolutional neural network. The anomaly identification model sequentially comprises a first convolutional neural network, a coding network, a decoding network and a second convolutional neural network from input to output, wherein the coding network and the decoding network are the variational self-encoder.
After the sample motion attitude features are input into the first convolution neural network, first features extracted from the sample motion attitude features by the first convolution neural network can be obtained, then the first features are input into a coding network, and the coding network codes the first features into hidden variables. And inputting the hidden variable into a decoding network, decoding the hidden variable by the decoding network to obtain a second feature, and inputting the extracted second feature into a second convolution network to obtain a reconstructed motion attitude feature and a likelihood value.
Wherein the likelihood value is used for measuring the difference between the sample motion attitude characteristic and the reconstructed motion attitude characteristic. And adjusting parameters of the first convolutional neural network, the coding network, the decoding network and the second convolutional neural network according to the likelihood value until the likelihood value meets a preset condition to obtain the abnormal recognition model.
It is understood that the variational self-encoder is trained in order to make the decoding network and the coding network learn the correct mode of the target action, therefore, the difference between the reconstructed motion posture feature and the input sample motion posture feature should be small enough, that is, when the likelihood value is large enough, the training of the variational self-encoder is considered to be completed, and the anomaly recognition model is obtained.
And 360, inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-encoder and a convolutional neural network.
And step 370, determining whether the target action in the video clip is an abnormal action according to the likelihood value.
The steps 360 to 370 refer to the corresponding parts of the previous embodiments, and are not described herein again.
It should be noted that, steps 330 to 350 may be executed before step 360, before step 320, or before step 310, and may be selected according to actual needs, which is not limited herein.
According to the abnormal action determining method provided by the embodiment of the application, a sample segment is obtained, and the sample segment is a video segment corresponding to a standard target action; acquiring a sample motion attitude characteristic corresponding to the sample segment; and training a variational self-encoder based on the sample motion attitude characteristics to obtain the abnormal recognition model. The variational self-encoder and the convolutional neural network are trained on the basis of the video segment corresponding to the standard target action to obtain the abnormal recognition model, so that the abnormal recognition model can learn the standard mode of the target action, the training is convenient and fast, the abnormal action is confirmed on the basis of the abnormal recognition model, the use is simple and convenient, and the wide application is easy.
Referring to fig. 5, an abnormal operation determination apparatus 400 according to an embodiment of the present application is provided, where the abnormal operation determination apparatus 400 includes an obtaining module 410, a constructing module 420, an identifying module 430, and a determining module 440. The obtaining module 410 is configured to obtain a target video segment corresponding to a target action; the constructing module 420 is configured to acquire a human skeleton key point corresponding to each frame of image in the target video segment, and construct a motion posture feature corresponding to the target video segment based on the human skeleton key point corresponding to each frame of image, where the motion posture feature represents a change mode of a human posture in the target video segment; the recognition module 430 is configured to input the motion posture features into an anomaly recognition model, to obtain a likelihood value corresponding to the motion posture features, where the anomaly recognition model is a pre-trained variational self-encoder and a convolutional neural network; the determining module 440 is configured to determine whether a target action in the video segment is an abnormal action according to the likelihood value.
Further, the obtaining module 410 is further configured to obtain a video to be recognized that includes the target action; identifying human skeleton key points of each frame of image in the video to be identified; and extracting a target video clip corresponding to the target action from the video to be identified based on the human skeleton key point.
Further, referring to fig. 6, an abnormal motion determination apparatus 400 is provided in the present embodiment, where the building module 420 further includes an identification unit 421, a building unit 422, an extraction unit 423, and a composition unit 424. The identification unit 421 is configured to identify a human skeleton key point corresponding to each frame of image in the target video segment; the constructing unit 422 is configured to construct a pose vector corresponding to each frame of image based on the human skeleton key points corresponding to each frame of image, where the pose vector includes parameters of multiple dimensions; the extraction unit 423 is configured to extract parameters of the same dimension in each pose vector, and obtain a timing sequence vector corresponding to each dimension according to the number of frames of an image used for constructing the pose vector in the target video segment; the forming unit 424 is configured to use a matrix formed by the time sequence vectors corresponding to each dimension as the motion posture feature.
Further, the identifying module 430 is further configured to obtain a sample segment, where the sample segment is a video segment corresponding to a standard target action; acquiring a sample motion attitude characteristic corresponding to the sample segment; and training the variational self-encoder and the convolutional neural network based on the sample motion attitude characteristics to obtain the abnormal recognition model.
Further, the anomaly identification model sequentially includes, from input to output, a first convolutional neural network, a variational self-encoder, and a second convolutional neural network, where the variational self-encoder includes an encoding network and a decoding network, and the identification module 430 is further configured to input the sample motion attitude feature into the first convolutional neural network, and input into the encoding network after being processed by the first convolutional neural network to obtain an implicit variable; inputting the hidden variable into the decoding network, and inputting the processed hidden variable into a second convolutional neural network to obtain a reconstructed motion attitude feature and a likelihood value, wherein the likelihood value represents the difference between the reconstructed motion attitude feature and the input sample motion attitude feature; and adjusting parameters of the first convolutional neural network, the coding network, the decoding network and the second convolutional neural network until the likelihood value meets a preset condition, and obtaining the abnormal recognition model.
Further, the determining module 440 is further configured to determine a magnitude relationship between the likelihood value and a preset likelihood value; if the likelihood value is smaller than the preset likelihood value, determining a target action in the video clip as an abnormal action; and if the likelihood value is greater than or equal to the preset likelihood value, determining that the target action in the video clip is a standard action.
The abnormal action determining device provided by the embodiment of the application acquires a target video clip corresponding to a target action; acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-coder and a convolutional neural network; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value. The method has the advantages that the change of the human body posture in the target video clip can be accurately described on the basis of the motion posture characteristics constructed on the basis of the key points of the human body skeleton in each frame of image in the target video clip, the confirmation of abnormal actions can be accurately and timely carried out by utilizing the motion posture characteristics and the abnormal recognition model, the method is simple and convenient to use, and the method is easy to widely apply.
It should be noted that, as will be clear to those skilled in the art, for convenience and brevity of description, the specific working process of the above-described apparatus may refer to the corresponding process in the foregoing method embodiment, and is not described herein again.
Referring to fig. 7, an embodiment of the present application provides a block diagram of an electronic device, where the electronic device 500 includes a processor 510, a memory 520, and one or more applications, where the one or more applications are stored in the memory 520 and configured to be executed by the one or more processors 510, and the one or more programs are configured to perform the above-mentioned abnormal action confirmation method.
The electronic device 500 may be a terminal device capable of running an application, such as a smart phone or a tablet computer, or may be a server. The electronic device 500 in the present application may include one or more of the following components: a processor 510, a memory 520, and one or more applications, wherein the one or more applications may be stored in the memory 520 and configured to be executed by the one or more processors 510, the one or more programs configured to perform a method as described in the aforementioned method embodiments.
Processor 510 may include one or more processing cores. The processor 510 interfaces with various components throughout the electronic device 500 using various interfaces and circuitry to perform various functions of the electronic device 500 and process data by executing or executing instructions, programs, code sets, or instruction sets stored in the memory 520 and invoking data stored in the memory 520. Alternatively, the processor 510 may be implemented in hardware using at least one of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 510 may integrate one or more of a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), a modem, and the like. Wherein, the CPU mainly processes an operating system, a user interface, an application program and the like; the GPU is used for rendering and drawing display content; the modem is used to handle wireless communications. It is understood that the modem may not be integrated into the processor 510, but may be implemented by a communication chip.
The Memory 520 may include a Random Access Memory (RAM) or a Read-Only Memory (Read-Only Memory). The memory 520 may be used to store instructions, programs, code sets, or instruction sets. The memory 520 may include a stored program area and a stored data area, wherein the stored program area may store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing various method embodiments described below, and the like. The data storage area may also store data created during use by the electronic device 500 (e.g., phone books, audio-visual data, chat log data), and so forth.
The electronic equipment provided by the embodiment of the application acquires a target video clip corresponding to a target action; acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment; inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-coder and a convolutional neural network; and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value. The method has the advantages that the change of the human body posture in the target video clip can be accurately described on the basis of the motion posture characteristics constructed on the basis of the key points of the human body skeleton in each frame of image in the target video clip, the confirmation of abnormal actions can be accurately and timely carried out by utilizing the motion posture characteristics and the abnormal recognition model, the method is simple and convenient to use, and the method is easy to widely apply.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While the preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all alterations and modifications as fall within the scope of the application.
It will be apparent to those skilled in the art that various changes and modifications may be made in the present application without departing from the spirit and scope of the application. Thus, if such modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is intended to include such modifications and variations as well.
Claims (10)
1. An abnormal action determination method, characterized in that the method comprises:
acquiring a target video clip corresponding to a target action;
acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment;
inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, wherein the abnormal recognition model is a pre-trained variational self-coder and a convolutional neural network;
and determining whether the target action in the video clip is an abnormal action or not according to the likelihood value.
2. The method of claim 1, wherein the obtaining a target video segment corresponding to the target action comprises:
acquiring a video to be identified comprising the target action;
identifying human skeleton key points of each frame of image in the video to be identified;
and extracting a target video clip corresponding to the target action from the video to be identified based on the human skeleton key point.
3. The method of claim 1, wherein obtaining human skeleton key points corresponding to each frame of image in the target video segment, and constructing motion posture features corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image comprises:
identifying human skeleton key points corresponding to each frame of image in the target video clip;
constructing a posture vector corresponding to each frame of image based on the human skeleton key points corresponding to each frame of image, wherein the posture vector comprises parameters of multiple dimensions;
extracting parameters of the same dimension in each attitude vector, and obtaining a time sequence vector corresponding to each dimension according to the frame number of an image used for constructing the attitude vector in the target video segment;
and taking a matrix formed by the time sequence vectors corresponding to each dimension as the motion attitude characteristic.
4. The method of claim 1, wherein the anomaly identification model is obtained by:
acquiring a sample segment, wherein the sample segment is a video segment corresponding to a standard target action;
acquiring a sample motion attitude characteristic corresponding to the sample segment;
and training the variational self-encoder and the convolutional neural network based on the sample motion attitude characteristics to obtain the abnormal recognition model.
5. The method according to claim 4, wherein the anomaly recognition model comprises, from input to output, a first convolutional neural network, a variational self-encoder and a second convolutional neural network in sequence, the variational self-encoder comprises an encoding network and a decoding network, and the training of the variational self-encoder based on the sample motion posture features to obtain the anomaly recognition model comprises:
inputting the sample motion attitude characteristics into a first convolution neural network, and inputting the sample motion attitude characteristics into a coding network after the sample motion attitude characteristics are processed by the first convolution neural network to obtain an implicit variable;
inputting the hidden variable into the decoding network, and inputting the processed hidden variable into a second convolutional neural network to obtain a reconstructed motion attitude feature and a likelihood value, wherein the likelihood value represents the difference between the reconstructed motion attitude feature and the input sample motion attitude feature;
and adjusting parameters of the first convolutional neural network, the coding network, the decoding network and the second convolutional neural network until the likelihood value meets a preset condition, and obtaining the abnormal recognition model.
6. The method according to any one of claims 1-5, wherein said determining whether the target action in the video segment is an abnormal action according to the likelihood value comprises:
determining the magnitude relation between the likelihood value and a preset likelihood value;
if the likelihood value is smaller than the preset likelihood value, determining a target action in the video clip as an abnormal action;
and if the likelihood value is greater than or equal to the preset likelihood value, determining that the target action in the video clip is a standard action.
7. An abnormal operation determination apparatus, characterized in that the apparatus comprises:
the acquisition module is used for acquiring a target video clip corresponding to the target action;
the construction module is used for acquiring human skeleton key points corresponding to each frame of image in the target video segment, and acquiring motion posture characteristics corresponding to the target video segment based on the human skeleton key points corresponding to each frame of image, wherein the motion posture characteristics represent the change mode of human posture in the target video segment;
the recognition module is used for inputting the motion attitude characteristics into an abnormal recognition model to obtain a likelihood value corresponding to the motion attitude characteristics, and the abnormal recognition model is a pre-trained variational self-encoder and a convolutional neural network;
and the determining module is used for determining whether the target action in the video clip is an abnormal action or not according to the likelihood value.
8. The apparatus of claim 7, wherein the building module further comprises:
the identification unit is used for identifying human skeleton key points corresponding to each frame of image in the target video clip;
the construction unit is used for constructing a posture vector corresponding to each frame of image based on the human skeleton key points corresponding to each frame of image, and the posture vector comprises a plurality of dimensional parameters;
the extraction unit is used for extracting the same dimension parameter in each attitude vector and obtaining a time sequence vector corresponding to each dimension parameter according to the frame number of an image used for constructing the attitude vector in the target video segment;
and the composition unit is used for taking a matrix formed by the time sequence vectors corresponding to each dimension parameter as the motion attitude characteristic.
9. An electronic device, characterized in that the electronic device comprises:
one or more processors;
a memory electrically connected with the one or more processors;
one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications configured to perform the method of any of claims 1-6.
10. A computer-readable storage medium, having stored thereon program code that can be invoked by a processor to perform the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110625830.3A CN113392742A (en) | 2021-06-04 | 2021-06-04 | Abnormal action determination method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110625830.3A CN113392742A (en) | 2021-06-04 | 2021-06-04 | Abnormal action determination method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113392742A true CN113392742A (en) | 2021-09-14 |
Family
ID=77618408
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110625830.3A Pending CN113392742A (en) | 2021-06-04 | 2021-06-04 | Abnormal action determination method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113392742A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113688790A (en) * | 2021-09-22 | 2021-11-23 | 武汉工程大学 | Human body action early warning method and system based on image recognition |
| CN113743362A (en) * | 2021-09-17 | 2021-12-03 | 平安医疗健康管理股份有限公司 | Method for correcting training action in real time based on deep learning and related equipment thereof |
| CN114005183A (en) * | 2021-12-31 | 2022-02-01 | 北京瑞莱智慧科技有限公司 | Action recognition method, device, equipment and storage medium |
| CN114241375A (en) * | 2021-12-15 | 2022-03-25 | 首都体育学院 | Monitoring method used in movement process |
| CN114241601A (en) * | 2021-12-16 | 2022-03-25 | 北京数码视讯技术有限公司 | Soldier training posture detection method and device and electronic equipment |
| CN114343618A (en) * | 2021-12-20 | 2022-04-15 | 中科视语(北京)科技有限公司 | Training motion detection method and device |
| CN114565784A (en) * | 2022-03-15 | 2022-05-31 | 平安科技（深圳）有限公司 | Pedestrian abnormal behavior detection method and device based on clustering algorithm and storage medium |
| CN114663972A (en) * | 2021-11-05 | 2022-06-24 | 范书琪 | Target marking method and device based on motion difference |
| CN114764946A (en) * | 2021-09-18 | 2022-07-19 | 北京甲板智慧科技有限公司 | Action counting method and system based on time sequence standardization and intelligent terminal |
| CN114973097A (en) * | 2022-06-10 | 2022-08-30 | 广东电网有限责任公司 | Method, device, equipment and storage medium for recognizing abnormal behaviors in electric power machine room |
| CN115063891A (en) * | 2022-08-19 | 2022-09-16 | 山东新蓝海科技股份有限公司 | Human body abnormal sign data monitoring method |
| CN115100745A (en) * | 2022-07-05 | 2022-09-23 | 北京甲板智慧科技有限公司 | Swin transform model-based motion real-time counting method and system |
| CN115205750A (en) * | 2022-07-05 | 2022-10-18 | 北京甲板智慧科技有限公司 | Motion real-time counting method and system based on deep learning model |
| US20230137560A1 (en) * | 2021-11-04 | 2023-05-04 | Institute For Information Industry | Assistance system and method for guiding exercise postures in live broadcast |
| CN116246350A (en) * | 2023-05-11 | 2023-06-09 | 山东工程职业技术大学 | Motion monitoring method, device, equipment and storage medium based on motion capture |
| CN117392762A (en) * | 2023-12-13 | 2024-01-12 | 中国石油集团川庆钻探工程有限公司 | Characteristic behavior recognition method based on human body key point gesture coding |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111680562A (en) * | 2020-05-09 | 2020-09-18 | 北京中广上洋科技股份有限公司 | Human body posture identification method and device based on skeleton key points, storage medium and terminal |
| CN112016435A (en) * | 2020-08-25 | 2020-12-01 | 中移（杭州）信息技术有限公司 | Abnormal behavior detection method and device, electronic equipment and storage medium |
-
2021
- 2021-06-04 CN CN202110625830.3A patent/CN113392742A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111680562A (en) * | 2020-05-09 | 2020-09-18 | 北京中广上洋科技股份有限公司 | Human body posture identification method and device based on skeleton key points, storage medium and terminal |
| CN112016435A (en) * | 2020-08-25 | 2020-12-01 | 中移（杭州）信息技术有限公司 | Abnormal behavior detection method and device, electronic equipment and storage medium |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113743362A (en) * | 2021-09-17 | 2021-12-03 | 平安医疗健康管理股份有限公司 | Method for correcting training action in real time based on deep learning and related equipment thereof |
| CN114764946A (en) * | 2021-09-18 | 2022-07-19 | 北京甲板智慧科技有限公司 | Action counting method and system based on time sequence standardization and intelligent terminal |
| CN114764946B (en) * | 2021-09-18 | 2023-08-11 | 北京甲板智慧科技有限公司 | Action counting method and system based on time sequence standardization and intelligent terminal |
| CN113688790A (en) * | 2021-09-22 | 2021-11-23 | 武汉工程大学 | Human body action early warning method and system based on image recognition |
| US20230137560A1 (en) * | 2021-11-04 | 2023-05-04 | Institute For Information Industry | Assistance system and method for guiding exercise postures in live broadcast |
| CN114663972A (en) * | 2021-11-05 | 2022-06-24 | 范书琪 | Target marking method and device based on motion difference |
| CN114241375B (en) * | 2021-12-15 | 2024-08-20 | 首都体育学院 | Monitoring method for exercise process |
| CN114241375A (en) * | 2021-12-15 | 2022-03-25 | 首都体育学院 | Monitoring method used in movement process |
| CN114241601A (en) * | 2021-12-16 | 2022-03-25 | 北京数码视讯技术有限公司 | Soldier training posture detection method and device and electronic equipment |
| CN114343618A (en) * | 2021-12-20 | 2022-04-15 | 中科视语(北京)科技有限公司 | Training motion detection method and device |
| CN114005183A (en) * | 2021-12-31 | 2022-02-01 | 北京瑞莱智慧科技有限公司 | Action recognition method, device, equipment and storage medium |
| CN114565784A (en) * | 2022-03-15 | 2022-05-31 | 平安科技（深圳）有限公司 | Pedestrian abnormal behavior detection method and device based on clustering algorithm and storage medium |
| CN114973097A (en) * | 2022-06-10 | 2022-08-30 | 广东电网有限责任公司 | Method, device, equipment and storage medium for recognizing abnormal behaviors in electric power machine room |
| CN115100745A (en) * | 2022-07-05 | 2022-09-23 | 北京甲板智慧科技有限公司 | Swin transform model-based motion real-time counting method and system |
| CN115205750A (en) * | 2022-07-05 | 2022-10-18 | 北京甲板智慧科技有限公司 | Motion real-time counting method and system based on deep learning model |
| CN115063891A (en) * | 2022-08-19 | 2022-09-16 | 山东新蓝海科技股份有限公司 | Human body abnormal sign data monitoring method |
| CN116246350A (en) * | 2023-05-11 | 2023-06-09 | 山东工程职业技术大学 | Motion monitoring method, device, equipment and storage medium based on motion capture |
| CN117392762A (en) * | 2023-12-13 | 2024-01-12 | 中国石油集团川庆钻探工程有限公司 | Characteristic behavior recognition method based on human body key point gesture coding |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113392742A (en) | Abnormal action determination method and device, electronic equipment and storage medium | |
| CN109558832B (en) | Human body posture detection method, device, equipment and storage medium | |
| CN107784282B (en) | Object attribute identification method, device and system | |
| CN106897658B (en) | Method and device for identifying human face living body | |
| CN112329888B (en) | Image processing method, device, electronic equipment and storage medium | |
| CN107958230B (en) | Facial expression recognition method and device | |
| CN109145766A (en) | Model training method, device, recognition methods, electronic equipment and storage medium | |
| CN110363220B (en) | Behavior class detection method and device, electronic equipment and computer readable medium | |
| CN113239849B (en) | Body-building action quality assessment method, body-building action quality assessment system, terminal equipment and storage medium | |
| CN113392741A (en) | Video clip extraction method and device, electronic equipment and storage medium | |
| CN108900788B (en) | Video generation method, video generation device, electronic device, and storage medium | |
| CN108096833B (en) | Motion sensing game control method and device based on cascade neural network and computing equipment | |
| CN108875506B (en) | Face shape point tracking method, device and system and storage medium | |
| CN117523659A (en) | Skeleton-based multi-feature multi-stream real-time action recognition method, device and medium | |
| CN113781462A (en) | Human body disability detection method, device, equipment and storage medium | |
| CN113378852A (en) | Key point detection method and device, electronic equipment and storage medium | |
| CN113392743A (en) | Abnormal action detection method, abnormal action detection device, electronic equipment and computer storage medium | |
| CN116959097A (en) | Action recognition method, device, equipment and storage medium | |
| CN113808192B (en) | House pattern generation method, device, equipment and storage medium | |
| CN109886123B (en) | Method and terminal for identifying human body actions | |
| Hoang et al. | Poses classification in a Taekwondo lesson using skeleton data extracted from videos with shallow and deep learning architectures | |
| CN113505716A (en) | Training method of vein recognition model, and recognition method and device of vein image | |
| CN111985582B (en) | Knowledge point mastering degree evaluation method based on learning behaviors | |
| Vedalankar et al. | Live Yoga Pose Detection by using Image Processing and Deep Learning | |
| CN110111298A (en) | Intelligent house type size verification method, apparatus, equipment and readable storage medium storing program for executing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210914 |
|
| RJ01 | Rejection of invention patent application after publication |