CN110889393A - Human body posture estimation method and device - Google Patents
Human body posture estimation method and device Download PDFInfo
- Publication number
- CN110889393A CN110889393A CN201911259717.7A CN201911259717A CN110889393A CN 110889393 A CN110889393 A CN 110889393A CN 201911259717 A CN201911259717 A CN 201911259717A CN 110889393 A CN110889393 A CN 110889393A
- Authority
- CN
- China
- Prior art keywords
- human body
- information
- human
- posture estimation
- key points
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- 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
Abstract
The application provides a human body posture estimation method and equipment, which can use second information of human body key points obtained by a second human body posture estimation model to guide model training of a first human body posture estimation model, so that the human body key point knowledge learned by the second human body posture estimation model is transferred to the first human body posture estimation model in a knowledge distillation mode, the human body key point estimation precision of the first human body posture estimation model is improved, the faster test speed of the first posture estimation model is kept, and real-time human body posture estimation can be realized.
Description
Technical Field
The application relates to the field of image recognition, in particular to a human body posture estimation method and device.
Background
The human body posture estimation refers to the position of a human body key point given from a two-dimensional image, and can be used in various fields of human-computer interaction, visual monitoring, sports analysis, medical diagnosis, virtual reality, augmented reality and the like. Whether the single-person posture estimation or the multi-person posture estimation is carried out, the quality of the recognition result depends on the influence of the environment, such as human body shielding, illumination intensity, different scales and angles can cause interference on the recognition result of the key points of the human body. For multi-person pose estimation, the prior art can be roughly divided into two types: bottom-up body pose estimation and top-down body pose estimation. The method has the advantages that all human key points in an image are firstly detected by estimating the human posture from bottom to top, and then the human key points are combined into a plurality of independent human bodies, the speed of the method is high, but the precision is not high, and related papers such as Association recording, End-to-End learning for joint detection and grouping and the like published in Advances in New Information Processing Systems in 2017 by Newell A, Huang Z and the like. Top-down body pose estimation is based on first detecting all the bodies by the body detectors and then detecting the corresponding body key points for each body, which is a method with high precision but slow speed, and related papers such as Simple bases for human position estimation and tracking published by Xiao B, Wu H et al on european conference on Computer Vision (ECCV).
Disclosure of Invention
An object of the present application is to provide a method and an apparatus for estimating a body posture, which are used to solve the problem of low precision of the real-time bottom-up method for estimating a body posture in the prior art.
In order to achieve the above object, the present application provides a method for estimating a human body posture, wherein the method comprises:
constructing a first human body posture estimation model, wherein the first human body posture estimation model is trained by using second information of the human body key points output by a second human body posture estimation model, the first human body posture estimation model is constructed by using a bottom-up human body posture estimation method, and the second human body posture estimation model is constructed by using a top-down human body posture estimation method;
and inputting the human body image into the first human body posture estimation model, and acquiring human body key point first information and human body posture information in the human body image.
Further, constructing a first human posture estimation model, comprising:
acquiring a characteristic diagram of the human body image;
acquiring first information of key points of the human body in the human body image according to the characteristic diagram of the human body image;
according to the second information of the human key points output by the second human posture estimation model, correcting the first information of the human key points in the corresponding region;
and grouping the first information of the human body key points, and determining corresponding human body posture information according to a grouping result.
Further, acquiring a feature map of the human body image includes:
and performing feature extraction on the human body image through the convolution layer and the pooling layer to obtain a feature map of the human body image.
Further, adjusting the first information of the human body key points in the corresponding region according to the second information of the human body key points output by the second human body posture estimation model, including:
inputting the human body image into the second human body posture estimation model to obtain second information of the output human body key points;
determining the difference between the second information of the human body key points and the corresponding first information of the human body key points in the same region;
and correcting the first information of the human key points to minimize the difference, and acquiring the first information of the human key points with the minimum difference.
Further, correcting the first information of the human body key points to minimize the difference, and acquiring the first information of the human body key points with the minimum difference, includes:
acquiring a difference between the second information of the human key points in the same region and the corresponding first information of the human key points according to a preset loss function, and reducing the difference by a preset loss function optimization method through continuous model training;
and when the difference between the two meets a preset training stopping threshold value, stopping model training and acquiring the current first information of the human body key point.
Further, grouping the first information of the human body key points, and determining corresponding human body posture information according to a grouping result, including:
clustering the first information of the human key points according to the category of the first information of the human key points, and determining corresponding human posture information according to the first information of the human key points in the clustering classification.
Further, the constructing step of the second human posture estimation model includes:
acquiring a characteristic diagram of the human body image;
carrying out human body target detection on the characteristic diagram of the human body image through a convolutional neural network, and determining a single human body image in the human body image;
and carrying out single posture estimation on the single human body image to acquire second information of the human body key points in the single human body image.
Further, after determining a single human body image of the human body images, the method further includes:
and acquiring a plurality of single human body images from the human body image, and adjusting the plurality of single human body images into a uniform size.
Based on another aspect of the present application, the present application also provides an apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, cause the apparatus to perform the aforementioned human body posture estimation method.
The present application also provides a computer readable medium having computer readable instructions stored thereon, which can be executed by a processor to implement the aforementioned human body posture estimation method.
Compared with the prior art, the scheme provided by the application can be used for guiding the model training of the first human posture estimation model by using the second information of the human key points obtained by the second human posture estimation model, so that the human key point knowledge learned by the second human posture estimation model is transferred to the first human posture estimation model in a knowledge distillation mode, the human key point estimation precision of the first human posture estimation model is improved, the faster test speed of the first posture estimation model is kept, and the real-time human posture estimation can be realized.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
fig. 1 is a flowchart of a human body posture estimation method according to some embodiments of the present application.
Detailed Description
The present application is described in further detail below with reference to the attached figures.
In a typical configuration of the present application, the terminal and the network device each include one or more processors (CPUs), input/output interfaces, network interfaces, and memories.
The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.
Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory computer readable media (transient media), such as modulated data signals and carrier waves.
Fig. 1 illustrates a human body posture estimation method provided by some embodiments of the present application, where the method may specifically include the following steps:
step S101, constructing a first human body posture estimation model, wherein the first human body posture estimation model is guided to train by using second information of human body key points output by a second human body posture estimation model, the first human body posture estimation model is constructed by using a bottom-up human body posture estimation method, and the second human body posture estimation model is constructed by using a top-down human body posture estimation method;
step S102, inputting a human body image into the first human body posture estimation model, and acquiring human body key point first information and human body posture information in the human body image.
The method is particularly suitable for being used in the occasion of carrying out real-time human body posture estimation on the human body image containing multiple persons, and can transmit the second information of the human body key points learned by the second human body posture estimation model to the first human body posture estimation model, so as to guide the first human body posture estimation model to train to obtain more accurate first information of the human body key points.
In step S101, a first human body posture estimation model is first constructed. The first human body posture estimation model is constructed by using a bottom-up human body posture estimation method, the second human body posture estimation model is constructed by using a top-down human body posture estimation method, and the second human body posture estimation model is used for assisting training by using the second human body key point information output by the second human body posture estimation model in the training process of the first human body posture estimation model, so that the precision of the first human body posture estimation model is improved, and the first human body posture estimation model can output the first human body key point information with higher precision.
In some embodiments of the present application, constructing the first human posture estimation model may specifically include the following steps:
1) acquiring a characteristic diagram of the human body image; the human body images are a plurality of images used for training a first human body posture estimation model, each image is provided with a plurality of human bodies, the positions of key points of the human bodies and the human body postures are labeled in advance, the human body postures are connection graphs formed by connecting corresponding key points of the human bodies, the key points of the human bodies can be divided into different types, such as shoulders, elbows, hands and the like, specific key points of the human bodies comprise left shoulders, right shoulders, left elbows, right elbows, left hands, right hands and the like, and generally 17 key points of the human bodies are provided;
specifically, feature extraction is carried out on the human body image through a convolution layer and a pooling layer to obtain a feature map of the human body image, wherein the convolution layer and the pooling layer are processing steps in a convolutional neural network, and the feature image of the human body image is obtained by processing the human body image through convolution operation and pooling operation in the convolution layer;
2) acquiring first information of key points of the human body in the human body image according to the characteristic diagram of the human body image; the feature map obtained in the previous step is further passed through a plurality of connected convolutional layers, so as to obtain a heat map of the pose estimation of all human bodies including a plurality of channels, wherein each channel represents a category of a key point of a human body, for example, the number of channels may be 17, the human keypoints of the category for multiple humans in the initial human image have higher activations on the corresponding channel heat map, determining the positions of the human key points of the category according to the higher activated positions, calculating a loss function between the obtained positions of the human key points and the corresponding positions of the human key points which are labeled in advance, and continuously optimizing the gap between the two by an optimization method of a loss function such as a gradient descent method or the like, stopping model training when a preset training stopping condition is met, and including the current position of the human body key point in the first information of the human body key point;
3) according to the second information of the human key points output by the second human posture estimation model, correcting the first information of the human key points in the corresponding region; here, the second body posture estimation model transfers the acquired knowledge to the first body posture estimation model in a knowledge distillation manner, so that the first body posture estimation model can optimize its own output information according to the transferred knowledge. Knowledge distillation, i.e., extraction of dark knowledge, can induce training of student networks (student networks: streamlined, low complexity) by introducing soft targets associated with teacher networks (teacher networks: complex, but superior reasoning performance) as part of the overall loss, thereby enabling knowledge migration. In some embodiments of the present application, the knowledge of the second body pose estimation model is migrated to the first body pose estimation model with the second body pose estimation model as a teacher network and the first body pose estimation model as a student network.
Specifically, the method can comprise the following steps:
(a) inputting the human body image into the second human body posture estimation model to obtain second information of the output human body key points; here, the second body posture estimation model is a model constructed by a top-down body posture estimation method, and in some embodiments of the present application, the construction method of the second body posture estimation model may specifically include the following steps: acquiring a characteristic diagram of a human body image; determining a single human body image in the human body images according to the characteristic diagram of the human body images; and carrying out single posture estimation on the single human body image to acquire second information of the human body key points in the single human body image. Here, the input human body image and the human body image input to the first human body posture estimation model are the same image batch, and the frame of the single human body image and the human body key point of the single human body image are preset in the human body image and are used as targets in the model training process.
The second human body posture estimation model also uses the convolution layer and the pooling layer to obtain a characteristic diagram of the human body image, then carries out human body target detection on the characteristic diagram of the human body image through the convolution neural network, and then carries out single posture estimation on each detected human body image. The single-person posture estimation of each human body image can adopt the existing single-person posture estimation technology, so that the positions of a plurality of human body key points in each single-person human body image are obtained, and the position information is contained in the second information of the human body key points.
(b) Determining the difference between the second information of the human body key points and the corresponding first information of the human body key points in the same region; the accuracy of the second information of the human key points obtained according to the second human posture estimation model is higher, the accuracy of the first information of the human key points obtained according to the first human posture estimation model is lower, and a certain difference exists between the two information, specifically, the positions of the human key points obtained according to the second human posture estimation model and the positions of the corresponding human key points obtained according to the first human posture estimation model have a deviation;
(c) correcting the first information of the human key points to enable the difference to be minimum, and acquiring the first information of the human key points when the difference is minimum; specifically, the difference between the second information of the human key points in the same region and the corresponding first information of the human key points is obtained according to a preset loss function, and the difference between the second information of the human key points and the corresponding first information of the human key points is reduced through continuous model training by a preset loss function optimization method; and when the difference between the two meets a preset training stopping threshold value, stopping model training and acquiring the current first information of the human body key point. Specifically, the positions of the human key points obtained according to the second human posture estimation model are accurate, the first information of the human key points obtained according to the first human posture estimation model needs to be continuously trained with the second information of the human key points as a target to optimize, specifically, a loss function between the first information of the human key points and the corresponding second information of the human key points is established, the loss function is, for example, cross entropy of the first information and the second information of the human key points, values of the loss function are continuously reduced in continuous training through a loss function optimization method, and the loss function optimization method is, for example, a gradient descent method and the like. And if the value of the loss function is smaller than a preset training stopping threshold which can be set by a user according to the needs of the user, stopping the training of the first human body posture estimation model, and taking the current first information of the human body key points as the final first information of the human body key points.
4) Grouping the first information of the human body key points, and determining corresponding human body posture information according to a grouping result; the first information of the key points of the human body has different types of information, such as shoulders, elbows, hands, knees, feet and the like, and the first information of the key points of the human body can be grouped according to the types of information; specifically, the first information of the human body key points can be clustered according to the category of the first information of the human body key points, corresponding human body posture information is determined according to the first information of the human body key points in the clustering, and when the human body posture information is determined, the human body posture information can be determined according to a preset human body key point connection rule, for example, the human body key point with the category of elbow can only be connected with the human body key point with the category of shoulder or hand, and the like.
In step S102, a human body image is input into the first human body posture estimation model, and first information of human body key points and human body posture information in the human body image are acquired. Here, the human body image refers to an image including a plurality of human bodies for estimating the human body pose in an actual scene, and the image does not label key points of the human bodies in advance, but processes the image by using a trained first human body pose estimation model to obtain first information of the key points of the human bodies in the image and corresponding human body pose information.
Some embodiments of the present application also provide an apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, cause the apparatus to perform the aforementioned human body posture estimation method.
Some embodiments of the present application also provide a computer readable medium having stored thereon computer readable instructions executable by a processor to implement the aforementioned human body posture estimation method.
Compared with the prior art, the scheme provided by the application can be used for guiding the model training of the first human posture estimation model by using the second information of the human key points obtained by the second human posture estimation model, so that the human key point knowledge learned by the second human posture estimation model is transferred to the first human posture estimation model in a knowledge distillation mode, the human key point estimation precision of the first human posture estimation model is improved, the faster test speed of the first posture estimation model is kept, and the real-time human posture estimation can be realized.
It should be noted that the present application may be implemented in software and/or a combination of software and hardware, for example, implemented using Application Specific Integrated Circuits (ASICs), general purpose computers or any other similar hardware devices. In one embodiment, the software programs of the present application may be executed by a processor to implement the steps or functions described above. Likewise, the software programs (including associated data structures) of the present application may be stored in a computer readable recording medium, such as RAM memory, magnetic or optical drive or diskette and the like. Additionally, some of the steps or functions of the present application may be implemented in hardware, for example, as circuitry that cooperates with the processor to perform various steps or functions.
In addition, some of the present application may be implemented as a computer program product, such as computer program instructions, which when executed by a computer, may invoke or provide methods and/or techniques in accordance with the present application through the operation of the computer. Program instructions which invoke the methods of the present application may be stored on a fixed or removable recording medium and/or transmitted via a data stream on a broadcast or other signal-bearing medium and/or stored within a working memory of a computer device operating in accordance with the program instructions. An embodiment according to the present application comprises a device comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, trigger the device to perform a method and/or a solution according to the aforementioned embodiments of the present application.
It will be evident to those skilled in the art that the present application is not limited to the details of the foregoing illustrative embodiments, and that the present application may be embodied in other specific forms without departing from the spirit or essential attributes thereof. The present embodiments are therefore to be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference sign in a claim should not be construed as limiting the claim concerned. Furthermore, it is obvious that the word "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. A plurality of units or means recited in the apparatus claims may also be implemented by one unit or means in software or hardware. The terms first, second, etc. are used to denote names, but not any particular order.
Claims (10)
1. A human body posture estimation method, wherein the method comprises the following steps:
constructing a first human body posture estimation model, wherein the first human body posture estimation model is trained by using second information of the human body key points output by a second human body posture estimation model, the first human body posture estimation model is constructed by using a bottom-up human body posture estimation method, and the second human body posture estimation model is constructed by using a top-down human body posture estimation method;
and inputting the human body image into the first human body posture estimation model, and acquiring human body key point first information and human body posture information in the human body image.
2. The method of claim 1, wherein constructing a first human pose estimation model comprises:
acquiring a characteristic diagram of the human body image;
acquiring first information of key points of the human body in the human body image according to the characteristic diagram of the human body image;
according to the second information of the human key points output by the second human posture estimation model, correcting the first information of the human key points in the corresponding region;
and grouping the first information of the human body key points, and determining corresponding human body posture information according to a grouping result.
3. The method of claim 2, wherein obtaining the feature map of the human body image comprises:
and performing feature extraction on the human body image through the convolution layer and the pooling layer to obtain a feature map of the human body image.
4. The method according to claim 2, wherein modifying the human body key point first information in the corresponding region according to the human body key point second information output by the second human body posture estimation model comprises:
inputting the human body image into the second human body posture estimation model to obtain second information of the output human body key points;
determining the difference between the second information of the human body key points and the corresponding first information of the human body key points in the same region;
and correcting the first information of the human key points to minimize the difference, and acquiring the first information of the human key points with the minimum difference.
5. The method according to claim 4, wherein the modifying the human body key point first information to minimize the gap and obtaining the human body key point first information when the gap is minimized includes:
acquiring a difference between the second information of the human key points in the same region and the corresponding first information of the human key points according to a preset loss function, and reducing the difference by a preset loss function optimization method through continuous model training;
and when the difference between the two meets a preset training stopping threshold value, stopping model training and acquiring the current first information of the human body key point.
6. The method of claim 2, wherein grouping the human body key point first information and determining corresponding human body posture information according to a grouping result comprises:
clustering the first information of the human key points according to the category of the first information of the human key points, and determining corresponding human posture information according to the first information of the human key points in the clustering classification.
7. The method of claim 1, wherein the constructing of the second body pose estimation model comprises:
acquiring a characteristic diagram of the human body image;
carrying out human body target detection on the characteristic diagram of the human body image through a convolutional neural network, and determining a single human body image in the human body image;
and carrying out single posture estimation on the single human body image to acquire second information of the human body key points in the single human body image.
8. The method of claim 7, wherein after determining a single one of the human images, further comprising:
and acquiring a plurality of single human body images from the human body image, and adjusting the plurality of single human body images into a uniform size.
9. An apparatus comprising a memory for storing computer program instructions and a processor for executing the program instructions, wherein the computer program instructions, when executed by the processor, cause the apparatus to perform the method of any of claims 1 to 8.
10. A computer readable medium having computer readable instructions stored thereon which are executable by a processor to implement the method of any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911259717.7A CN110889393A (en) | 2019-12-10 | 2019-12-10 | Human body posture estimation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911259717.7A CN110889393A (en) | 2019-12-10 | 2019-12-10 | Human body posture estimation method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110889393A true CN110889393A (en) | 2020-03-17 |
Family
ID=69751307
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911259717.7A Pending CN110889393A (en) | 2019-12-10 | 2019-12-10 | Human body posture estimation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110889393A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111626105A (en) * | 2020-04-15 | 2020-09-04 | 南京旷云科技有限公司 | Attitude estimation method and device and electronic equipment |
| CN112800991A (en) * | 2021-02-02 | 2021-05-14 | 厦门美图之家科技有限公司 | Rapid human body posture estimation method and device with calculated quantity self-adaption capability and electronic equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108345869A (en) * | 2018-03-09 | 2018-07-31 | 南京理工大学 | Driver's gesture recognition method based on depth image and virtual data |
| CN109657631A (en) * | 2018-12-25 | 2019-04-19 | 上海智臻智能网络科技股份有限公司 | Human posture recognition method and device |
| CN109902628A (en) * | 2019-02-28 | 2019-06-18 | 广州大学 | A kind of seat Management System of Library of view-based access control model Internet of Things |
| CN109919245A (en) * | 2019-03-18 | 2019-06-21 | 北京市商汤科技开发有限公司 | Deep learning model training method and device, training equipment and storage medium |
| CN109948453A (en) * | 2019-02-25 | 2019-06-28 | 华中科技大学 | A kind of more people's Attitude estimation methods based on convolutional neural networks |
| CN110334743A (en) * | 2019-06-10 | 2019-10-15 | 浙江大学 | A kind of progressive transfer learning method based on the long memory network in short-term of convolution |
| CN110472481A (en) * | 2019-07-01 | 2019-11-19 | 华南师范大学 | A kind of sleeping position detection method, device and equipment |
| CN110503077A (en) * | 2019-08-29 | 2019-11-26 | 郑州大学 | A kind of real-time body's action-analysing method of view-based access control model |
-
2019
- 2019-12-10 CN CN201911259717.7A patent/CN110889393A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108345869A (en) * | 2018-03-09 | 2018-07-31 | 南京理工大学 | Driver's gesture recognition method based on depth image and virtual data |
| CN109657631A (en) * | 2018-12-25 | 2019-04-19 | 上海智臻智能网络科技股份有限公司 | Human posture recognition method and device |
| CN109948453A (en) * | 2019-02-25 | 2019-06-28 | 华中科技大学 | A kind of more people's Attitude estimation methods based on convolutional neural networks |
| CN109902628A (en) * | 2019-02-28 | 2019-06-18 | 广州大学 | A kind of seat Management System of Library of view-based access control model Internet of Things |
| CN109919245A (en) * | 2019-03-18 | 2019-06-21 | 北京市商汤科技开发有限公司 | Deep learning model training method and device, training equipment and storage medium |
| CN110334743A (en) * | 2019-06-10 | 2019-10-15 | 浙江大学 | A kind of progressive transfer learning method based on the long memory network in short-term of convolution |
| CN110472481A (en) * | 2019-07-01 | 2019-11-19 | 华南师范大学 | A kind of sleeping position detection method, device and equipment |
| CN110503077A (en) * | 2019-08-29 | 2019-11-26 | 郑州大学 | A kind of real-time body's action-analysing method of view-based access control model |
Non-Patent Citations (1)
| Title |
|---|
| SHENG WANG等: "Combined Top-Down/Bottom-Up Human Articulated Pose Estimation Using AdaBoost Learning", 《2010 20TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111626105A (en) * | 2020-04-15 | 2020-09-04 | 南京旷云科技有限公司 | Attitude estimation method and device and electronic equipment |
| CN111626105B (en) * | 2020-04-15 | 2024-02-20 | 南京旷云科技有限公司 | Gesture estimation method and device and electronic equipment |
| CN112800991A (en) * | 2021-02-02 | 2021-05-14 | 厦门美图之家科技有限公司 | Rapid human body posture estimation method and device with calculated quantity self-adaption capability and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113159056B (en) | Image segmentation method, device, equipment and storage medium | |
| CN111052128B (en) | Descriptor learning method for detecting and locating objects in video | |
| CN113326851B (en) | Image feature extraction method and device, electronic equipment and storage medium | |
| Selvan et al. | Uncertainty quantification in medical image segmentation with normalizing flows | |
| CN113743273B (en) | Real-time rope skipping counting method, device and equipment based on video image target detection | |
| CN110889393A (en) | Human body posture estimation method and device | |
| Gilbert et al. | Automated left ventricle dimension measurement in 2d cardiac ultrasound via an anatomically meaningful cnn approach | |
| Zhang et al. | Joint motion information extraction and human behavior recognition in video based on deep learning | |
| Fei et al. | Flow-pose Net: An effective two-stream network for fall detection | |
| Chen et al. | Brain–computer interface‐based target recognition system using transfer learning: a deep learning approach | |
| Mazzamuto et al. | Weakly supervised attended object detection using gaze data as annotations | |
| Wang et al. | Lightweight bilateral network for real-time semantic segmentation | |
| CN110197100B (en) | Object positioning method and device | |
| Pahwa et al. | A survey on object detection performance with different data distributions | |
| Yang et al. | Student Classroom Behavior Detection Based on YOLOv7+ BRA and Multi-model Fusion | |
| CN110910478B (en) | GIF map generation method and device, electronic equipment and storage medium | |
| Bühler et al. | Automated hand-raising detection in classroom videos: A view-invariant and occlusion-robust machine learning approach | |
| CN112241001A (en) | Radar human body action recognition method and device, electronic equipment and storage medium | |
| Iegawa et al. | Loop Closure Detection in Visual SLAM Based on Convolutional Neural Network | |
| Cheng et al. | Audio-visual salient object detection | |
| Abdulghani et al. | Discover human poses similarity and action recognition based on machine learning | |
| Zhang et al. | Arm Movement Analysis Technology of Wushu Competition Image Based on Deep Learning | |
| Wang et al. | An object detection algorithm based on the feature pyramid network and single shot multibox detector | |
| Mao et al. | A deep learning approach to track Arabidopsis seedlings’ circumnutation from time-lapse videos | |
| Fan et al. | Real-time and accurate meal detection for meal-assisting robots |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200317 |